Copyright

Copyright © 2018 by Ben Orlin Jacket design by Headcase Design Jacket

copyright © 2018 by Hachette Book Group, Inc.

Hachette Book Group supports the right to free expression and the value of
copyright. The purpose of copyright is to encourage writers and artists to
produce the creative works that enrich our culture.

The scanning, uploading, and distribution of this book without permission is a

theft of the author’s intellectual property. If you would like permission to use
material from the book (other than for review purposes), please contact
[email protected]. Thank you for your support of the author’s rights.

Black Dog & Leventhal Publishers Hachette Book Group

1290 Avenue of the Americas
New York, NY 10104
www.hachettebookgroup.com
www.blackdogandleventhal.com

First Edition: August 2018

Black Dog & Leventhal Publishers is an imprint of Running Press, a division of

Hachette Book Group. The Black Dog & Leventhal Publishers name and logo
are trademarks of Hachette Book Group, Inc.

The publisher is not responsible for websites (or their content) that are not
owned by the publisher.

The Hachette Speakers Bureau provides a wide range of authors for speaking
events. To find out more, go to www.HachetteSpeakersBureau.com or call (866)
376-6591.

Library of Congress Control Number: 2018930088

ISBNs: 978-0-316-50903-9 (hardcover); 978-0-316-50902-2 (ebook) E3-

ISBNs: 978-0-316-50903-9 (hardcover); 978-0-316-50902-2 (ebook) E3-
20180719-JV-PC
 CONTENTS

COVER
TITLE PAGE
COPYRIGHT
DEDICATION
INTRODUCTION

I
HOW TO THINK LIKE A MATHEMATICIAN

1. Ultimate Tic-Tac-Toe
2. What Does Math Look Like to Students?
3. What Does Math Look Like to Mathematicians?
4. How Science and Math See Each Other
5. Good Mathematician vs. Great Mathematician

II
DESIGN: THE GEOMETRY OF STUFF THAT WORKS

6. We Built This City on Triangles

7. Irrational Paper
8. The Square-Cube Fables
9. The Game of Dice
10. An Oral History of the Death Star

III
PROBABILITY: THE MATHEMATICS OF MAYBE

11. The 10 People You Meet in Line for the Lottery

12. Children of the Coin
13. What Does Probability Mean in Your Profession?
14. Weird Insurance
15. How to Break the Economy with a Single Pair of Dice

IV
STATISTICS: THE FINE ART OF HONEST LYING

16. Why Not to Trust Statistics

17. The Last .400 Hitter
18. Barbarians at the Gate of Science
19. The Scoreboard Wars
20. The Book Shredders

V
ON THE CUSP: THE POWER OF A STEP

21. The Final Speck of Diamond Dust

22. Bracketology
23. One State, Two State, Red State, Blue State
24. The Chaos of History

ACKNOWLEDGMENTS
ENDNOTES
NEWSLETTERS
For Taryn
 INTRODUCTION

This is a book about math. That was the plan, anyway.

Somewhere, it took an unexpected left turn. Before long, I found myself
without cell phone reception, navigating a series of underground tunnels. When I
emerged into the light, the book was still about math, but it was about lots of
other things, too: Why people buy lottery tickets. How a children’s book author
swung a Swedish election. What defines a “Gothic” novel. Whether building a
giant spherical space station was really the wisest move for Darth Vader and the
Empire.
That’s math for you. It connects far-flung corners of life, like a secret system
of Mario tubes.
 If this description rings false to you, it’s perhaps because you’ve been to a
place called “school.” If so, you have my condolences.
When I graduated from college in 2009, I thought I knew why mathematics
was unpopular: It was, on the whole, badly taught. Math class took a beautiful,
imaginative, logical art, shredded it into a bowl of confetti, and assigned students
the impossible, mind-numbing task of piecing the original back together. No
wonder they groaned. No wonder they failed. No wonder adults look back on
their math experiences with a shudder and a gag reflex. The solution struck me
as obvious: Math required better explanations, better explainers.
Then I became a teacher. Undertrained and oozing hubris, I needed a brutal
first year in the classroom to teach me that, whatever I knew about math, I didn’t
yet understand math education, or what the subject meant to my students.
One day that September, I found myself leading an awkward impromptu
discussion of why we study geometry. Did grown-ups write two-column proofs?
Did engineers work in “no calculator” environments? Did personal finance
demand heavy use of the rhombus? None of the standard justifications rang true.
In the end, my 9th graders settled on “We study math to prove to colleges and
employers that we are smart and hardworking.” In this formulation, the math
itself didn’t matter. Doing math was a weightlifting stunt, a pointless show of
intellectual strength, a protracted exercise in résumé building. This depressed
me, but it satisfied them, which depressed me even more.
The students weren’t wrong. Education has a competitive zero-sum aspect, in
which math functions as a sorting mechanism. What they were missing—what I
was failing to show them—was math’s deeper function.
Why does mathematics underlie everything in life? How does it manage to
link disconnected realms—coins and genes, dice and stocks, books and baseball?
The reason is that mathematics is a system of thinking, and every problem in the
world benefits from thinking.
Since 2013, I’ve been writing about math and education—sometimes for
publications like Slate, the Atlantic, and the Los Angeles Times, but mostly for
my own blog, Math with Bad Drawings. People still ask me why I do the bad
drawings. I find this odd. No one ever wonders why I choose to cook mediocre
food, as if I’ve got a killer chicken l’orange that, on principle, I decline to serve.
It’s the same with my art. Math with Bad Drawings is a less pathetic title than
Math with the Best Drawings I Can Manage; Honestly, Guys, I’m Trying, but in
my case they are equivalent.
I suppose my path began one day when I drew a dog on the board to illustrate
a problem, and got the biggest laugh of my career. The students found my
ineptitude shocking, hilarious, and, in the end, kind of charming. Math too often
feels like a high-stakes competition; to see the alleged expert reveal himself as
the worst in the room at something—anything—that can humanize him and,
perhaps, by extension, the subject. My own humiliation has since become a key
element in my pedagogy; you won’t find that in any teacher training program,
but, hey, it works.

Lots of days in the classroom, I strike out. Math feels to my students like a
musty basement where meaningless symbols shuffle back and forth. The kids
shrug, learn the choreography, and dance the tuneless dance.
But on livelier days, they see distant points of light and realize that the
basement is a secret tunnel, linking everything they know to just about
everything else. The students grapple and innovate, drawing connections, taking
leaps, building that elusive virtue called “understanding.”
Unlike in the classroom, this book will sidestep the technical details.You’ll
find few equations on these pages, and the spookiest ones are decorative
anyway. (The hard core can seek elaboration in the endnotes.) Instead, I want to
focus on what I see as the true heart of mathematics: the concepts. Each section
of this book will tour a variety of landscapes, all sharing the underground
network of a single big idea: How the rules of geometry constrain our design
choices. How the methods of probability tap the liquor of eternity. How tiny
increments yield quantum jumps. How statistics make legible the mad sprawl of
reality.
Writing this book has brought me to places I didn’t expect. I hope that
reading it does the same for you.
—BEN ORLIN, October 2017
 I
HOW TO THINK LIKE A MATHEMATICIAN

To be honest, mathematicians don’t do much. They drink coffee, frown at

chalkboards. Drink tea, frown at students’ exams. Drink beer, frown at proofs
they wrote last year and can’t for the life of them understand anymore.
It’s a life of drinking, frowning, and, most of all, thinking.

You see, there are no physical objects in math: no chemicals to titrate, no

particles to accelerate, no financial markets to destroy. Rather, the verbs of the
mathematician all boil down to actions of thought. When we calculate, we turn
one abstraction into another. When we give proofs, we build logical bridges
between related ideas. When we write algorithms or computer programs, we
enlist an electronic brain to think the thoughts that our meat brains are too slow
or too busy to think for themselves.
Every year that I spend in the company of mathematics, I learn new styles of
thought, new ways to use that nifty all-purpose tool inside the skull: How to
master a game by fussing with its rules. How to save thoughts for later, by
recording them in loopy Greek symbols. How to learn from my errors as if they
were trusted professors. And how to stay resilient when the dragon of confusion
comes nibbling at my toes.
In all these ways, mathematics is an action of the mind.
What about math’s vaunted “real-world” usefulness? How can spaceships
and smartphones and godforsaken “targeted ads” emerge from this fantasy
skyline of pure thought? Ah, patience, my friend. All that comes later. We’ve
got to begin where all mathematics begins: That is to say, with a game…
 Chapter 1
ULTIMATE TIC-TAC-TOE

WHAT IS MATHEMATICS?

Once at a picnic in Berkeley, I saw a group of mathematicians abandon their

Frisbees to crowd around the last game I would have expected: tic-tac-toe.
As you may have discovered yourself, tic-tac-toe is terminally dull. (That’s a
medical term.) Because there are so few possible moves, experienced players
soon memorize the optimal strategy. Here’s how all my games play out:
 If both players have mastered the rules, every game will end in a draw,
forever and ever. The result is rote gameplay, with no room for creativity.
But at that picnic in Berkeley, the mathematicians weren’t playing ordinary
tic-tac-toe. Their board looked like this, with each square of the original nine
turned into a mini-board of its own:
As I watched, the basic rules became clear:
 But it took a while longer for the most important rule of the game to emerge:
You don’t get to pick which of the nine mini-boards to play on. Instead,
that’s determined by your opponent’s previous move. Whichever square she
picks on the mini-board, you must play next in the corresponding square on
the big board.
(And whichever square you pick on that mini-board will determine which
mini-board she plays on next.)
 This lends the game its strategic element. You can’t just focus on each mini-
board in turn. You’ve got to consider where your move will send your opponent,
and where her next move will send you—and so on, and so on.
(There’s only one exception to this rule: If your opponent sends you to a
mini-board that’s already been won, then congratulations—you can go anywhere
you like, on any of the other mini-boards.)
The resulting scenarios look bizarre, with players missing easy two-and
three-in-a-rows. It’s like watching a basketball star bypass an open layup and
fling the ball into the crowd. But there’s a method to the madness. They’re
thinking ahead, wary of setting up their foe on prime real estate. A clever attack
on a mini-board leaves you vulnerable on the big board, and vice versa—it’s a
tension woven into the game’s fabric.
From time to time, I play Ultimate Tic-Tac-Toe with my students; they enjoy
the strategy, the chance to defeat a teacher, and, most of all, the absence of
trigonometric functions. But every so often, one of them will ask a sheepish and
natural question: “So, I like the game,” they’ll say, “but what does any of this
have to do with math?”
I know how the world sees my chosen occupation: a dreary tyranny of
inflexible rules and formulaic procedures, no more tension-filled than, say,
insurance enrollment, or filling out your taxes. Here’s the sort of tedious task we
associate with mathematics:
 This problem can hold your attention for a few minutes, I guess, but before
long the actual concepts slip from mind. “Perimeter” no longer means the length
of a stroll around the rectangle’s border; it’s just “the two numbers, doubled and
added.” “Area” doesn’t refer to the number of one-by-one squares it takes to
cover the rectangle; it’s just “the two numbers, multiplied.” As if competing in a
game of garden-variety tic-tac-toe, you fall into mindless computation. There’s
no creativity, no challenge.
But, as with the revved-up game of Ultimate Tic-Tac-Toe, math has far
greater potential than this busywork suggests. Math can be bold and exploratory,
demanding a balance of patience and risk. Just trade in the rote problem above
for one like this:
 This problem embodies a natural tension, pitting two notions of size (area
and perimeter) against one another. More than just applying a formula, solving it
requires deeper insight into the nature of rectangles. (See the endnotes for
spoilers.)
Or, how about this:
 A little spicy, right?
In two quick steps, we’ve traveled from sleepwalking drudgery to a rather
formidable little puzzle, one that stumped a whole school’s worth of bright-eyed
6th graders when I tossed it as a bonus on their final exam. (Again, see the
endnotes for solutions.)
As this progression suggests, creativity requires freedom, but freedom alone
is not enough. The pseudopuzzle “Draw two rectangles” provides loads of
freedom but no more spark than a damp match. To spur real creativity, a puzzle
needs constraints.
Take Ultimate Tic-Tac-Toe. Each turn, you’ve got only a few moves to
choose from—perhaps three or four. That’s enough to get your imagination
flowing, but not so many that you’re left drowning in a sea of incalculable
possibilities. The game supplies just enough rules, just enough constraints, to
spur our ingenuity.
That pretty well summarizes the pleasure of mathematics: creativity born
from constraints. If regular tic-tac-toe is math as people know it, then Ultimate
Tic-Tac-Toe is math as it ought to be.

You can make the case that all creative endeavors are about pushing against
constraints. In the words of physicist Richard Feynman, “Creativity is
imagination in a straitjacket.” Take the sonnet, whose tight formal restrictions—
Follow this rhythm! Adhere to this length! Make sure these words rhyme!
Okay… now express your love, lil’ Shakespeare!—don’t undercut the artistry but
heighten it. Or look at sports. Humans strain to achieve goals (kick the ball in the
net) while obeying rigid limitations (don’t use your hands). In the process, they
create bicycle kicks and diving headers. If you ditch the rulebook, you lose the
grace. Even the wacky, avant-garde, convention-defying arts—experimental
film, expressionist painting, professional wrestling—draw their power from
playing against the limitations of the chosen medium.
Creativity is what happens when a mind encounters an obstacle. It’s the
human process of finding a way through, over, around, or beneath. No obstacle,
no creativity.
But mathematics takes this concept one step further. In math, we don’t just
follow rules. We invent them. We tweak them. We propose a possible constraint,
play out its logical consequences, and then, if that way leads to oblivion—or
worse, to boredom—we seek a new and more fruitful path.
For example, what happens if I challenge one little assumption about parallel
lines?

Euclid laid out this rule regarding parallel lines in 300 BCE; he took it for
granted, calling it a fundamental assumption (a “postulate”). This struck his
successors as a bit funny. Do you really have to assume it? Shouldn’t it be
provable? For two millennia, scholars poked and prodded at the rule, like a piece
of food caught between their teeth. At last they realized: Oh! It is an assumption.
You can assume otherwise. And if you do, traditional geometry collapses,
revealing strange alternative geometries, where the words “parallel” and “line”
come to mean something else entirely.
 New rule, new game.
As it turns out, the same holds true of Ultimate Tic-Tac-Toe. Soon after I
began sharing the game around, I learned that there’s a single technicality upon
which everything hinges. It comes down to a question I confined to a
parenthetical earlier: What happens if my opponent sends me to a mini-board
that’s already been won?
These days, my answer is the one I gave above: Since that mini-board is
already “closed,” you can go wherever you want.
But originally, I had a different answer: As long as there’s an empty space on
that mini-board, you have to go there—even though it’s a wasted move.
 This sounds minor—just a single thread in the tapestry of the game. But
watch how it all unravels when we give this thread a pull.
I illustrated the nature of the original rule with an opening strategy that I
dubbed (in a triumph of modesty) “The Orlin Gambit.” It goes like this:
 When all is said and done, X has sacrificed the center mini-board in exchange
for superior position on the other eight. I considered this stratagem pretty clever
until readers pointed out its profound uncleverness. The Orlin Gambit wasn’t
just a minor advantage. It could be extended into a guaranteed winning strategy.
Rather than sacrifice one mini-board, you can sacrifice two, giving you two-in-a-
rows on each of the other seven. From there, you can ensure victory in just a
handful of moves.
Embarrassed, I updated my explanation to give the rule in its current version
—a small but crucial tweak that restored Ultimate Tic-Tac-Toe to health.
New rule, new game.
This is exactly how mathematics proceeds. You throw down some rules and
begin to play. When the game grows stale, you change it. You pose new
constraints. You relax old ones. Each tweak creates new puzzles, fresh
challenges. Most mathematics is less about solving someone else’s riddles than
about devising your own, exploring which constraints produce interesting games
and which produce drab ones. Eventually, this process of rule tweaking, of
moving from game to game, comes to feel like a grand never-ending game in
itself.
Mathematics is the logic game of inventing logic games.
The history of mathematics is the unfolding of this story, time and time
again. Logic games are invented, solved, and reinvented. For example: what
happens if I tweak this simple equation, changing the exponents from 2 into
another number, like 3, or 5, or 797?

Oh! I’ve turned an ancient and elementary formula, easily satisfied by

plugging in whole numbers (such as 3, 4, and 5), into perhaps the most vexing
equation that humans have ever encountered: Fermat’s last theorem. It tormented
scholars for 350 years, until the 1990s, when a clever Brit locked himself in an
attic and emerged nearly a decade later, blinking at the sunlight, with a proof
showing that whole number solutions can never work.
Or, what happens if I take two variables—say, x and y—and create a grid that
lets me see how they relate?

Oh! I’ve invented graphing, and thereby revolutionized the visualization of

mathematical ideas, because my name is Descartes and that’s why they pay me
the big bucks.
 Or, consider that squaring a number always gives a positive answer. Well,
what if we invented an exception: a number that, when squared, is negative?
What happens then?

Oh! We’ve discovered imaginary numbers, thereby enabling the exploration

of electromagnetism and unlocking a mathematical truth called “the fundamental
theorem of algebra,” all of which will sound pretty good when we add it to our
résumés.
In each of these cases, mathematicians at first underestimated the
transformative power of the rule change. In the first case, Fermat thought the
theorem would yield to a simple proof; as centuries of his frustrated successors
will attest, it didn’t. In the second case, Descartes’s graphing idea (now called
“the Cartesian plane” in his honor) was originally confined to the appendix of a
philosophical text; it was often dropped from later printings. And in the third
case, imaginary numbers faced centuries of snubs and insults (“as subtle as they
are useless,” said the great Italian mathematician Cardano) before being
embraced as true and useful numbers. Even their name is pejorative, originating
from a dis first uttered by none other than Descartes.
It’s easy to underestimate new ideas when they emerge not from somber
reflection but from a kind of play. Who would guess that a little tweak to the
rules (new exponent; new visualization; new number) could transform a game
beyond imagination and recognition?
I don’t think the mathematicians at that picnic had this in mind as they
huddled over a game of Ultimate Tic-Tac-Toe. But they didn’t need to. Whether
we’re cognizant of it or not, the logic game of inventing logic games exerts a
pull on us all.
 Chapter 2
WHAT DOES MATH LOOK LIKE TO STUDENTS?

Alas, this will be a short, bleak chapter. I’d apologize for that, but I’ll be too
busy apologizing for other things, like the often soul-pureeing experience of
math education.
You know what I’m talking about. To many students, “doing math” means
executing the prescribed sequence of pencil maneuvers. Mathematical symbols
don’t symbolize; they just dance across the page in baffling choreography. Math
is a tale told by an abacus, full of “sin” and “θ,” signifying nothing.
 Allow me to offer two brief apologies:
First, I apologize to my own students for any times I’ve made you feel this
way about math. I’ve tried not to; then again, I’ve also tried to stay on top of
email, cut back on ice cream, and avoid four-month lapses between haircuts.
Please enter my plea as “merely human.”
Second, I apologize to mathematics, for any violence you have suffered at my
hands. In my defense, you’re an intangible tower of quantitative concepts bound
together by abstract logic, so I doubt I’ve left any lasting scars. But I’m not too
proud to say I’m sorry.
That’s it for this chapter. I promise the next one will have more explosions,
like any good sequel should.
 Chapter 3
WHAT DOES MATH LOOK LIKE TO MATHEMATICIANS?

It’s very simple. Math looks like language.

A funny language, I’ll admit. It’s dense, terse, and painstaking to read. While
I zip through five chapters of a Twilight novel, you might not even turn the page
in your math textbook. This language is well suited to telling certain stories (e.g.,
the relations between curves and equations), and ill suited to others (e.g., the
relations between girls and vampires). As such, it’s got a peculiar lexicon, full of
words that no other tongue includes. For example, even if I could translate
into plain English, it wouldn’t make sense to someone
unfamiliar with Fourier analysis, any more than Twilight would make sense to
someone unfamiliar with teenage hormones.
But math is an ordinary language in at least one way. To achieve
comprehension, mathematicians employ strategies familiar to most readers. They
form mental images. They paraphrase in their heads. They skim past distracting
technicalities. They draw connections between what they’re reading and what
they already know. And—strange as it may seem—they engage their emotions,
finding pleasure, humor, and squeamish discomfort in their reading material.
Now, this brief chapter can’t teach fluent math any more than it could teach
fluent Russian. And just as literary scholars might debate a couplet by Gerard
Manley Hopkins or the ambiguous phrasing of an email, so mathematicians will
disagree on specifics. Each brings a unique perspective, shaped by a lifetime of
experience and associations.
That said, I hope to offer a few nonliteral translations, a few glimpses into the
strategies by which a mathematician might read some actual mathematics.
Consider it Squiggle Theory 101.
 A common question I get from students: “Does it matter whether I multiply
by 11 or 13 first?” The answer (“no”) is less interesting than what the question
reveals: that in my students’ eyes, multiplication is an action, a thing you do. So
one of the hardest lessons I teach them is this: Sometimes, don’t.
You don’t have to read 7 × 11 × 13 as a command. You can just call it a
number and leave it be.
Every number has lots of aliases and stage names. You could also call this
number 1002 – 1, or 499 × 2 + 3, or 5005/5, or Jessica, the Number That Will
Save Planet Earth, or plain old 1001. But if 1001 is how this number is known to
its friends, then 7 × 11 × 13 isn’t some quirky and arbitrary moniker. Rather, it’s
the official name you’d find on the birth certificate.
7 × 11 × 13 is the prime factorization, and it speaks volumes.
Some key background knowledge: Addition is kind of boring. To wit, writing
1001 as the sum of two numbers is a truly dull pastime: you can do it as 1000 +
1, or 999 + 2, or 998 + 3, or 997 + 4… and so on, and so on, until you slip into a
boredom coma. These decompositions don’t tell us anything special about 1001,
because all numbers can be broken up in pretty much the same way. (For
example, 18 can be written as 17 + 1, or 16 + 2, or 15 + 3…) Visually, this is
like breaking a number up into two piles. No offense, but piles are dumb.
Multiplication: now that’s where the party’s at. And to join the festivities,
you need to deploy our first math-reading strategy: forming mental images.
As the picture on the previous page shows, multiplication is all about grids
and arrays. 1001 can be seen as a giant structure of blocks, 7 by 11 by 13. But
that’s just getting started.
You can visualize this as 13 layers of 77 each. Or, if you tilt your head
sideways, it’s 11 layers of 91 each. Or, tilting your head a different sideways,
seven layers of 143 each. All of these ways to decompose 1001 are immediately
evident from the prime factorization… but virtually impossible to discern from
the name 1001 without laborious guesswork.
The prime factorization is the DNA of a number. From it, you can read all the
factors and factorizations, the numbers that divide our original and the numbers
that don’t. If math is cooking class, then 7 × 11 × 13 isn’t the pancake recipe. It’s
the pancake itself.

To casual fans, π is a mystic rune, a symbol of mathematical sorcery. They

ponder its irrationality, memorize thousands of its digits, and commemorate Pi
Day on March 14 by combining the most glorious of mankind’s arts (dessert
pies) with the least glorious (puns). To the general public, π is an object of
obsession, awe, even something approaching worship.
And to mathematicians, it’s roughly 3.
That infinite spool of decimal places that so captivates laypeople? Well,
mathematicians aren’t that bothered. They know math is about more than
precision; it’s about quick estimates and smart approximations. When building
intuition, it helps to streamline and simplify. Intelligent imprecision is our next
crucial math-reading strategy.
Take the formula A = πr2, which many students have heard so often the mere
phrase “area of a circle” triggers them to scream, “Pi r squared!” like sleeper
agents programmed by brainwashing. What does it mean? Why is it true?
Well, forget the 3.14159. Let your mind go fuzzy. Just look at the shapes.
r is the radius of our circle. It’s a length.
r2, then, is the area of a little square, like the one pictured.
Now, the π-dollar question: How does the area of the circle compare to the
area of the square?
Clearly, the circle is bigger. But it’s not quite four times bigger (since four
squares would cover the circle and then some). Eyeballing it, you might
speculate that the circle is a little more than three times bigger than the square.
And that’s exactly what our formula says: Area = a little more than 3 × r2.
If you want to verify the precise value—why 3.14-ish and not 3.19-ish?—
then you can use a proof. (There are several lovely demonstrations; my favorite
involves peeling the circle like an onion and stacking the layers to make a
triangle.) But mathematicians, whatever they insist, don’t always prove
everything from first principles. Like everyone from carpenters to zookeepers,
they’re happy to employ a tool without knowing precisely how it was
constructed, so long as they have a sense of why it works.

“Go graph these equations” is a familiar homework assignment. I’ve given it

myself. It’s also the seed of a vicious myth: that graphs are an end unto
themselves. In fact, graphing is not like solving an equation or carrying out an
operation. A graph is not an end; it is always, always a means.
A graph is a tool for data visualization, a picture that tells a story. It
represents another powerful math-reading strategy: turning the static into the
dynamic.
Take the equation above: . Here, x and y stand for a pair of numbers
obeying a precise relationship. Here are some examples of possible pairs:

A few patterns are already peeking out. But our vision is only as powerful as
our technology, and tables are not a fancy gadget. Of the infinite x-y pairs that
satisfy this equation, the table can show only a handful at a time, like the
scrolling text of a stock ticker. We need a superior visualization tool: the
mathematical equivalent of a television screen.
Enter the graph.
By treating x and y as a sort of longitude and latitude, we transform each
intangible pair of numbers into something satisfyingly geometric: a point. The
infinite set of points becomes a unified curve. And from this, a story emerges, a
tale of motion and change.

• When x grows small, approaching zero ( …), then y balloons to

enormous values (25, 3600, 1 million…).
• When x grows large (20, 40, 500…), then y shrinks to tiny proportions (
…).
• While x can assume negative values (-2, -5, -10), y never does. It remains
positive.
• And neither variable ever adopts a value of zero.
 Okay, maybe it’s not the juiciest plotline, but mental maneuvers like this
mark the difference between the novice’s experience of mathematics (as a
paralyzing stream of meaningless symbols) and the mathematician’s (as
something coherent and communicative). Graphs imbue lifeless equations with a
sense of motion.

There’s a psychological phenomenon known by the unfortunate name of

chunking. More than just a way to cleanse the system after too many beers,
chunking also happens to be a potent mental technique, indispensable to
mathematicians. It’s our next math-reading strategy.
When we “chunk,” we reinterpret a set of scattered, hard-to-retain details as a
single unit. The above equation offers a simple example. A good chunker
disregards the minutiae on the left side. Is it x or y, 5 or 6, + or -? Don’t know;
don’t care. Instead, you see only a pair of chunky factors, forming a skeletal
equation that reads: chunk × chunk = 0.
Now, if you’re familiar with multiplication, you know that zero is a peculiar
result to come by.
6 × 5? Not zero.
18 × 307? Not zero.
13.91632 × 4,600,000,000,000? No need to fire up your calculator app: This,
too, is not zero.
Zero is a singular number in the world of multiplication. In contrast to, say, 6,
which can be produced a variety of ways (3 × 2, 1.5 × 4, 1200 × 0.005…), zero
is special and elusive. In fact, there’s precisely one way that two numbers
multiply to give you zero: if one of the original numbers was, itself, zero.
This is where our chunking pays off: Since their product is zero, one of the
chunks must be, too. If it was the first one (x - 5), then x must be 5. And if it was
the latter (x - 7), then x must be 7.
Equation solved.
Chunking purifies not just the contents of our stomachs but the contents of
our minds. It makes the world digestible. And the more you learn, the more
aggressively you are able to chunk. A high schooler might chunk a whole row of
algebra as “find the area of the trapezoid.” An undergraduate might chunk
several dense lines of calculus as “compute the volume of the solid of
revolution.” And a graduate student might chunk half a page of formidable
Greek-lettered technicalities as “compute the Hausdorff dimension of the set.”
For each level up, you’ve got to learn subtle new details: What are trapezoids?
How do integrals behave? What was Hausdorff smoking, and where can we get
some?
But we don’t learn details for details’ sake. We learn details so as to ignore
them later, to focus on the big, chunky picture instead.
 Switch two symbols. What happens?
Well, in the eyes of the novice, nothing. You’ve interchanged scribbles,
swapped syllables in a gibberish language. Who cares? But in the eyes of a
mathematician, it can be like switching sea and sky, or mountain and cloud, or
bird and fish (much to the consternation of both). Switching two symbols can
change everything.
For example, take the two expressions above and imagine that x is 10.
Now, 102 is a big number. It means 10 × 10. That’s 100. It’s a reasonable
number of students to teach in a given year, or miles to drive to a theme park, or
dollars to pay for a used television. (It’s a suspicious number of dalmatians to
own.)
But 210 is a much bigger number. It’s 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2.
That’s 1024. It’s a reasonable number of students to teach in a decade, or miles
to drive to the world’s greatest theme park, or dollars to pay for an awe-inspiring
television. (It’s a very suspicious number of dalmatians to own; this is why we
have animal cruelty laws.)
As we try out bigger x’s, the gap between the two expressions widens. In
fact, the word “widens” is too soft, like describing the Grand Canyon as “bit of a
crack in the ground.” As x grows, the gap between x2 and 2x explodes.
To wit, 1002 is quite big. It’s 100 × 100. That’s 10,000.
But 2100 is huge. It’s 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2 × 2
×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×
2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2
×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×
2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2×2
× 2 × 2 × 2. That’s 1,267,650,600,228,229,401,496,703,205,376, which is
roughly a billion billion trillion.
If we’re talking about pounds, then the former is the weight of a pickup truck
carrying a load of bricks. Heavy, to be sure, but the latter is in a different weight
class altogether.
It’s the size of a hundred thousand Earths.
x2 and 2x don’t look so different to the untrained eye. But the more math you
experience and the more fluent you become in this language of squiggles, the
more dramatic the difference begins to feel. Before long it becomes visceral,
tactile; it starts to enlist your emotions, which is our final crucial strategy. You
read lines of mathematics with a full spectrum of feelings, from satisfaction to
sympathy to shock.
 Eventually, mixing up x2 and 2x becomes as absurd as imagining a pickup
truck towing a hundred thousand planets behind it.
 Chapter 4
HOW SCIENCE AND MATH SEE EACH OTHER

1. TWINS NO MORE
Back in 9th grade, my friend John and I looked oddly similar: pensive round-
faced brown-haired boys who spoke about as often as furniture. Teachers called
us by the wrong names; older students thought we were one person; and in the
yearbook, we’re mislabeled as each other. For all I remember, we sparked a
friendship just to mess with people.
Then, with time, we thinned out and grew up. John is now a broad-chested
six foot two and looks like a Disney prince. I’m five nine and have been
described as “a cross between Harry Potter and Daniel Radcliffe.” The
“practically twins” stage of our friendship is long gone.
And as it is with me and John, so it is with mathematics and science.
 Back when science and math still had their baby fat, they didn’t just look the
same. They were the same. Isaac Newton didn’t worry whether history would
classify him as a scientist or a mathematician: he was both, inextricably. The
same went for his intellectual elder siblings—Galileo, Kepler, Copernicus. For
them, science and math were interwoven, inseparable. Their core insight was
that the physical cosmos followed mathematical recipes. Objects obeyed
equations. You couldn’t study one without the other, any more than you could
eat a single ingredient out of a baked cake.
Science and math have diverged since then. Just look at how they’re taught:
separate classrooms, separate teachers, separate (if equally mind-numbing)
textbooks. They’ve thinned out, added muscle, aged beyond their wide-eyed 9th-
grade innocence.
But still, people get them confused. Any fool can see that I’m not John, and
John isn’t me, but the citizen on the street has more trouble telling the
differences between math and science—especially at a layman’s distance.
 Perhaps the easiest way to tell them apart is to answer this question: What do
science and math look like, not to the layperson, but to each other?

2. IN EACH OTHER’S EYES

Looking through science’s eyes, the answer is pretty clear. Science sees
mathematics as a tool kit. If science is a golfer, then mathematics is the caddy,
proffering the right club for each situation.
 This view casts mathematics in a subservient role. It’s not my favorite
dynamic, but, hey, I sympathize. Science is trying to make sense of reality,
which—as you’ll know if you have experience with reality—is pretty darn hard.
Things are born. Things die. Things leave maddeningly partial fossil records.
Things exhibit qualitatively different behaviors at quantum and relativistic
scales. Reality is kind of a mess.
 Science quests to make sense of all this. It aims to predict, classify, and
explain. And in this endeavor, it sees mathematics as a vital assistant: the Q to its
James Bond, supplying helpful gadgets for the next adventure.
 Now, let’s spin the camera 180° and change perspectives. How does
mathematics see science?
You’ll find that we’re not just alternating camera angles. We’re changing
film genres altogether. Whereas science sees itself as the protagonist of an action
movie, mathematics sees itself as the auteur director of an experimental art
project.
That’s because, on a fundamental level, mathematicians do not care about
reality.
 I’m not talking about the odd habits of mathematicians—muttering to
themselves, wearing the same trousers for weeks, forgetting their spouse’s
names on occasion. I’m talking about their work. Despite the aggressive ad
campaign about its “real-world usefulness,” mathematics is pretty indifferent to
the physical universe.
What math cares about are not things but ideas.
Math posits rules and then unpacks their implications by careful reasoning.
Who cares if the resulting conclusions—about infinitely long cones and 42-
dimensional sausages—don’t resemble physical reality? What matters is their
abstract truth. Math lives not in the material universe of science but in the
conceptual universe of logic.
 Mathematicians call this work “creative.” They liken it to art.
That makes science their muse. Think of a composer who hears chirping
birds and weaves the melody into her next work. Or a painter who gazes at
cumulus clouds drifting through an afternoon sky, and models her next
landscape on that image. These artists don’t care if they’ve captured their
subjects with photorealistic fidelity. For them, reality is nothing more or less
than a fertile source of inspiration.
That’s how math sees the world, too. Reality is a lovely starting point, but the
coolest destinations lie far beyond it.

3. THE PARADOX OF MATHEMATICS

Math sees itself as a dreamy poet. Science sees it as a supplier of specialized
technical equipment. And herein we find one of the great paradoxes of human
inquiry: These two views, both valid, are hard to reconcile. If math is an
equipment supplier, why is its equipment so strangely poetic? And if math is a
poet, then why is its poetry so unexpectedly useful?
 To see what I mean, take the twisted history of knot theory.
This branch of mathematics, like many, was inspired by a scientific problem.
Before the discovery of atoms, some scientists (including Lord Kelvin)
entertained the idea that the universe was filled with a substance called ether,
and matter was made of knots and tangles in it. Thus, they sought to classify all
the possible knots, creating a periodic table of tangles.
Before long, science lost interest, lured away by the shiny new theory of
atoms (which had the unfair advantage of being right). But mathematics was
hooked. It turns out that classifying knots is a delightful and devilish problem.
Two versions of the same knot can look wildly different. Totally different knots
can taunt you with their resemblance. It was perfect fuel for mathematicians,
who soon developed an exquisite, complex theory of knots, unperturbed that
their clever abstractions appeared to have no practical purpose whatsoever.
The centuries rolled along.
Then, science ran into a real snake of a problem. As you know, every cell
inscribes its precious information on DNA molecules, which are fantastically
long. If straightened out and laid flat, the DNA from one of your cells would
stretch for six feet—a hundred thousand times the length of the cell itself. This
makes DNA a long string stuffed into a small container. If you’ve ever shoved
earbuds into your pocket or removed Christmas lights from their box, you know
what this scenario creates: maddening tangles. How do bacteria manage this?
Can we learn their tricks? Can we perhaps disable cancer cells by tangling their
DNA?
Biology was flummoxed. It needed help. “Ooh,” Mathematics cried. “I know
just the thing!”

Here, then, is a brief biography of knot theory. It was born from a practical
need. Soon, it grew into something deliberately impractical, a logic game for
poets and philosophers. And yet somehow this mature creature, which had
barely spared a thought for reality over the years, became profoundly useful in a
field far removed from the one of its birth.
This is no isolated case. It’s a basic pattern of mathematical history.
Remember the strange alternate geometries from Chapter 1? For centuries,
scholars saw them as mere novelties, a poet’s fancy. They bore no
correspondence to reality, which was understood to follow Euclid’s assumption
about parallel lines.
Then, along came a young patent clerk named Einstein. He realized that these
wacko alternatives aren’t just thought experiments; they underlie the structure of
the cosmos. From our tiny perspective, the universe appears Euclidean, much in
the way that the curved Earth appears flat. But zoom out, shedding the
prejudices of the surface dweller, and you’ll see a different picture entirely, a
shifting landscape of strange curvatures.
Those “useless” geometries turned out to be pretty darn useful.
Perhaps my favorite example concerns logic itself. Early philosophers like
Aristotle developed symbolic logic (“If p, then q”) as a guide for scientific
thinking. Then mathematical theorists got their hands on it, and turned logic into
something bizarre and abstract. Reality fell away. By the 20th century, you had
folks like Bertrand Russell writing Latin-titled tomes that aimed to “prove” from
elementary assumptions that 1 + 1 = 2. What could be more useless, more
irredeemable?
One logician’s mother nagged him: C’mon, honey, what’s the point of all this
abstract mathematics? Why not do something useful?
That mother was named Ethel Turing. And as it turns out, her son Alan was
kind of on to something: a logic machine we now call “the computer.”
I can’t blame her for doubting. Who would have guessed that her son’s
abstract research into logical systems would help to define the next century? No
matter how many examples I encounter, this historical cycle of useful to useless
to useful again remains a wonder and a mystery to me.
My favorite description of this phenomenon is a phrase coined by physicist
Eugene Wigner: “the unreasonable effectiveness of mathematics.” After all,
bacteria don’t know any knot theory, so why should they follow its rules? The
space-time continuum hasn’t studied hyperbolic geometry, so why does it
execute its theorems so perfectly? I’ve read some philosophers’ answers to this
question, but I find them tentative and conflicting, and none have worked to
blunt my astonishment.
So, how best to understand the relationship between the poet we call Math
and the adventurer known as Science? Perhaps we ought to see them as a
symbiotic pair of very different creatures, like an insect-eating bird perched on
the back of a rhino. The rhino gets its itchy problems solved. The bird gets
nourished. Both emerge happy.
When you visualize math, picture something dainty and elegant astride the
wrinkled gray mass of reality below.
 Chapter 5
GOOD MATHEMATICIAN VS. GREAT MATHEMATICIAN

It’s a lot of fun to bust myths. Just look at the carefree explosions and ear-to-ear
smiles of television’s MythBusters, and you can see it’s a career with high job
satisfaction.
What’s trickier is tweaking myths. Lots of the culture’s prevailing views
about mathematics aren’t flat-out wrong—just crooked, or incomplete, or
overemphasized. Is computation important? Sure, but it’s not all-important. Does
math require attention to detail? Yes, but so do knitting and parkour. Was Carl
Gauss a natural genius? Well, yeah, but most beautiful math comes not from
depressive German perfectionists but from ordinary folks like you and me.
Before we close this section, this chapter offers one final exploration of how
to think like a mathematician, a chance to revise and annotate some popular
myths. Like most myths, they’ve got a basis in truth. And, like most myths, they
miss the flux, uncertainty, and struggle toward understanding that make us
human—and that make us mathematicians.

A few years ago, when I lived in England, I taught a boy named Corey. He
reminded me of a soft-spoken 12-year-old Benjamin Franklin: quiet and
insightful, with long ginger hair and round spectacles. I can totally picture him
inventing bifocals.
 Corey poured his heart into every homework assignment, drew lucid
connections across topics, and packed up his papers at the period’s end with such
care and patience that I always fretted he’d be late for his next lesson. So it’s no
surprise that on the first big test in November, Corey nailed every question.
Well… every question that he’d had time to answer.
The bell rang with the last quarter of his test still blank. He scored in the low
70s and came to me the next day with a furrowed brow. “Sir,” he said, because
England is an amazing land where clumsy 29-year-old teachers get fancy
honorifics, “why are tests timed?”
I figure honesty is the best policy. “It’s not because speed is so important. We
just want to see what students can do by themselves, without anyone’s help.”
“So why not let us keep working?”
“Well, if I held the class hostage for a whole day, it might annoy your other
teachers. They want you to know about science and geography, because of their
nostalgic attachment to reality.”
I realized that I had never seen Corey like this: jaw clenched, eyes dark. He
was radiating frustration. “I could have answered more,” he said. “I just ran out
of time.”
I nodded. “I know.”
There wasn’t much else to say.
Intentionally or not, school mathematics sends a loud, clear message: Speed
is everything. Tests are timed. Early finishers get to start their homework. Just
look how periods end: with a ringing bell, as if you’ve just finished a round in a
perverse, compulsory, logarithm-themed game show. Math comes to feel like a
race, and success becomes synonymous with quickness.
All of which is supremely silly.
Speed has one fabulous advantage: It saves time. Beyond that, mathematics is
about deep insight, real understanding, and elegant approaches, none of which
you’re likely to find when moving at 600 miles per hour. You learn more
mathematics by thinking carefully than by thinking fast, just as you learn more
botany by studying a blade of grass than by sprinting like the dickens through a
wheat field.
Corey understands this. I only hope that teachers like me don’t manage,
against our own best intentions, to persuade him otherwise.
 My wife, who is a research mathematician, once pointed me toward a funny
pattern in mathematical life.

• Step #1: There’s a tricky and exciting question on the loose, an important
conjecture in need of proof. Many try to tame the beast, without success.
• Step #2: Someone finally proves it, via a long and convoluted argument
that’s full of insight but very difficult to follow.
• Step #3: Over time, new proofs are published, growing shorter and
simpler, shorter and simpler, until eventually the original proof is relegated
to “historical” status: an inefficient Edison-era lightbulb made obsolete by
sleeker, more modern designs.

Why is this trajectory so common?

Well, the first time you arrive at the truth, you’ve often come by a jagged and
meandering path. To your credit, it takes patience to survive all the twists and
turns. But it takes a more profound patience to keep thinking afterward. Only
then can you sort the necessary steps from the superfluous ones, and perhaps boil
that sprawling 120-page proof down to a seamless 10-pager.
 In 1920, of all the branches of mathematics, algebra was perhaps the dullest.
To do algebra was to enter a bog of petty specifics, a thornbush of unwieldy
technicalities, a discipline of details.
Then, in 1921, a mathematician named Emmy Noether published a paper
titled “Theory of Ideals in Ring Domains.” It was the dawn, a colleague later
said, of “abstract algebra as a conscious discipline.” Noether wasn’t interested in
unpacking particular numerical patterns. In fact, she laid the whole idea of
“number” to the side. All that mattered to her were symmetry and structure. “She
taught us to think in simple, and thus general, terms,” remembered another
colleague years later. “She therefore opened a path to the discovery of algebraic
regularities where before these regularities had been obscured.”
To do good work, you’ve first got to engage with nitty-gritty details. Then, to
do great work, you’ve got to move beyond them.
For Noether, abstraction was more than an intellectual habit; it was a way of
life. “She often lapsed into her native German when she was bothered by some
idea,” said a colleague. “She loved to walk. She would take her students off for a
jaunt on a Saturday afternoon. On these trips, she would become so absorbed in
her conversation on mathematics that she would forget about the traffic, and her
students would need to protect her.”
Great mathematicians don’t bother with trivialities like crosswalks and traffic
flow. They’ve got their mind’s eye on something bigger.
 In 1998, Sylvia Serfaty got hooked on a question. It dealt with how certain
vortices evolve through time. She had literally written the book on the subject
—Vortices in the Magnetic Ginzburg-Landau Model—but found herself
stumped by this puzzle.
“A lot of good research,” she later said, “actually starts from very simple
things, elementary facts, basic bricks… Progress in math comes from
understanding the model case, the simplest instance in which you encounter the
problem. And often it is an easy computation; it’s just that no one had thought of
looking at it this way.”
You can attack the castle at the main gate, fighting the defensive forces head-
on. Or, you can seek to better understand the castle itself—and perhaps uncover
an easier way in.
The mathematician Alexander Grothendieck offered a different metaphor for
this matter of perspective. Think of a problem as a scrumptious hazelnut, with
“nourishing flesh protected by the shell.” How do we get at it?
There are two basic approaches. First, the hammer and chisel: whack the
shell with a sharp edge until you’ve cracked it. That’s likely to work, but it’s
violent and effortful. Or, second, you can immerse the nut in water. As
Grothendieck put it, “From time to time, you rub so the liquid penetrates better,
and otherwise you let time pass. The shell becomes more flexible through weeks
and months—when the time is ripe, hand pressure is enough; the shell opens like
a perfectly ripened avocado!”
For two decades, progress came to Serfaty in fits and starts, as she and
collaborators let the nut soak. Finally, in 2015, she found the right angle of
attack, the perfect perspective. The problem yielded in a matter of months.
 Every mathematical field has its Holy Grail. For many statisticians, that grail
was the Gaussian correlation inequality.
“I know of people who worked on it for 40 years,” said a Penn State
statistician named Donald Richards. “I myself worked on it for 30.” Many
scholars gave valiant attempts—hundred-page calculations, sophisticated
geometric framings, new developments drawn from analysis and probability
theory—but none could capture the grail. A few came to suspect it might even
be false, a myth.
Then, one day in 2014, Richards received an email from a German retiree
named Thomas Royen. It contained a Microsoft Word attachment. That was
weird—virtually all serious mathematicians type up their work with a program
called LaTeX. Why was this former pharmaceutical employee reaching out to a
leading researcher in statistics?
Well, as it so happens, this fellow had proved the Gaussian correlation
inequality. He had done it using arguments and formulas that any graduate
student could follow. The insight had popped into his head while he was
brushing his teeth.
“When I looked at it,” said Richards, “I knew instantly that it was solved.”
Richards felt humbled and frustrated to have missed such a simple argument, but
he was delighted nevertheless. “I remember thinking to myself that I was glad to
have seen it before I died. Really, I was so glad I saw it.”
Royen’s tale echoes a piece of wisdom I learned from a beloved physics
teacher: “If you want to kill flies, you don’t need bazookas.” For
mathematicians, there is elegance in restraint.
 November 2010, Oakland, California. I am 23 years old and spending the
morning demonstrating to my trigonometry students the sheer number of ways
that one can unsuccessfully explain de Moivre’s theorem.
“Okay, let me start again!” I say, sweating. “You’re going to be raising this
number to the nth power anyway, right? And so you can go ahead and add 2kπ/n
to the angle, because the zoombini’s other fraggle will repatriate the fleen. You
get it now?”
“No!” a roomful of teenagers cried, hands over their ears. “Stop! You’re
making it worse!”
A student named Vianney raised her hand. “Can I try to see if I understand?”
“Please,” I sighed. “Be my guest.”
“Okay, we’re going to be dividing this angle in half, right?”
The room calmed a little.
“And adding 360° to the angle just puts you back in the same place, right?
Like, 90° and 450° are no different; you’ve just done an extra rotation.”
Students straightened in their chairs.
“But when you divide the angle in half, that extra 360° becomes an extra
180°. You wind up on the opposite side of the circle.”
I watched as mental lightbulbs flashed all over the room, like a crowd of
photographers.
“So,” Vianney concluded, “that’s why there are two solutions. Right?”
I thought for a moment. The whole room leaned forward. At last I nodded.
“Yes. Well said.”
The room erupted into applause, as an avalanche of high fives and shoulder
thumps buried Vianney. Even I could see the superiority of her approach. Where
I had been trying to explain the theorem backward, and to cover all values of n
simultaneously, Vianney turned the problem right side out and focused on the
case where n = 2.
History’s greatest mathematicians have left their mark not only by feats of
intellectual strength but by blazing trails and teaching others to follow. Euclid
consolidated past insights into an irreplaceable textbook. Cantor distilled his
fresh understanding of infinity into clean, easy-to-follow arguments. Stein
mentored generations of harmonic analysts, becoming adviser to mathematicians
as great as himself.
It’s not that Vianney understood de Moivre’s theorem better than I did. It’s
that she could cash out her knowledge into clear language, whereas my insight
remained locked inside a thick skull, attached to a clumsy tongue. A
mathematician who can’t convey his thinking will suffer the same fate that I did
that day: to be a lonely island of thought, whose ideas never reach other shores,
while the mathematician who can share her truth enjoys a hero’s welcome from
the grateful crowd.

I doubt you’ve heard of Ngô B u Châu, unless (a) You got bored one day and
decided to memorize every winner of the Fields Medal, math’s most prestigious
prize, or (b) You’re from Vietnam, where Châu is a national celebrity and
household name.
(By the way, huge kudos to Vietnam for that. America’s closest
approximation to a celebrity mathematician is Will Hunting: not even our most
famous Matt Damon character.)
Châu is a fierce competitor, who craved as a youngster to be the best of the
best. For a time, he was. Winning back-to-back gold medals at the International
Mathematical Olympiad, he became the pride of his school and the envy of his
peers, the Simone Biles of Vietnamese math.
But his undergraduate years brought a slow quicksand agony, as he gradually
realized that he didn’t understand the math he was learning. “My professors
thought I was a fantastic student,” he remembers. “I could do all the exercises. I
did well on exams. But I didn’t understand anything.” Achievement came to feel
like hollow orb: a brittle shell of accolades, soon to shatter and reveal the
horrible vacuum inside.
Then came a turning point. He stopped trying to be the best, and began to
learn from them.
He shovels credit on his PhD adviser Gérard Laumon. “I had one of the best
advisers in the world,” Châu says, glowing. “I would come to his office every
week. He would read with me one or two pages every time.” They went line by
line, equation by equation, settling for nothing less than full comprehension.
Châu soon began work on the famous Langlands program. Think of it as the
transcontinental railroad of modern mathematics: a sweeping vision for how to
connect several distant branches of the discipline. The project has drawn
generations of ambitious mathematicians like Châu into its orbit, and Châu
found himself attracted to a particularly vexing piece of the Langlands program:
proving the “fundamental lemma.”
Was it the Olympiad all over again? Rival mathematicians jockeying for
primacy, racing to be the first to prove it?
No, says Châu.
“I was helped a lot by people in my field,” he says. “Many people
encouraged me to do so, in a very sincere way. I asked them for advice, and they
would tell me what to learn. It was very open. I did not feel competition.” With
help from these collaborators, Châu managed to prove the fundamental lemma.
It was the work than won him his Fields.
For such an exceptional scholar, Châu’s tale is pleasantly ordinary. People
who thrive in the tournament atmosphere of school—with its clear rankings,
easy lateral comparisons, and steady diet of rewards—find they need a new
attitude as they advance into the open-ended world of scholarship. Drawn in as
competitors, they evolve into collaborators.
 II
DESIGN

THE GEOMETRY OF STUFF THAT WORKS

Here’s a tough-love life lesson: just because you set your mind to something
doesn’t mean you can do it.
Say you’re creating a square, and you want its diagonal to be the same length
as its sides. Well, I hate to break it to you, but that’s not how squares work.
 Or say you’re tiling a floor, and you want to use a pattern of regular
pentagons. I’m sorry to report that this will never pan out. There will always be
gaps between your tiles.

Or say you’re building an equilateral triangle, and you fancy a bespoke angle
size—say 70°, or 49°, or 123°. I’m afraid the cosmos doesn’t care: Those angles
are going to be 60°, no more and no less, or else your triangle won’t be
equilateral.
 Human laws are flexible, subject to repeal and renegotiation. The drinking
age is 21 in the US, 18 in Britain, 16 in Cuba, “never” in Afghanistan, and “step
right up” in Cambodia. Any country can stiffen or relax its drinking laws at a
whim (and of course, the stiffer your drink and the laxer your laws, the more
whim-prone you become). Geometry’s laws are not like that. There’s no wiggle
room: no president to issue pardons, no jury to acquit, no officer to let you off
with a warning. Math’s rules are self-enforcing, unbreakable by their nature.
Yet as we’ve seen, and shall see many times again, that’s not a bad thing.
Restrictions birth creativity. The laws on what shapes can’t do come packaged
with case studies illuminating what they can. In design projects ranging from
sturdy buildings to useful paper to planet-destroying space stations, geometry
inspires even as it constrains.
So forget the sentiment that “anything is possible!” It is very sweet but
deeply unnatural, like most things we feed to children. The reality is harsher—
and more wondrous.
 Chapter 6
WE BUILT THIS CITY ON TRIANGLES

I’d like you to meet this chapter’s star: the triangle.

It’s not your typical protagonist. Snooty literary types may dismiss it as two-
dimensional. Yet this atypical hero will embark on a typical hero’s journey:
rising from humble origins, learning to harness an inner strength, and ultimately
serving the world in a time of crisis.
Now, if your mind is so narrow that it cannot encompass the idea of a valiant
polygon, then by all means, read no further. Don your blindfold of prejudice.
Just make sure to squeeze your eyelids tight, for only the deepest darkness can
guard your shuttered thoughts from the radiant truth, the penetrating light, of
planar geometry. Don’t you remember? We built this city. We built this city on
triangles…
 1. TWELVE KNOTS IN AN EGYPTIAN ROPE
Welcome to ancient Egypt: a prosperous kingdom, a teeming bureaucracy, rigid
of faith and right-angled of elbow. It shall endure for millennia as the sun rises
and sets on empires with lesser headgear.
Come, let’s stroll. The year is 2570 BCE, and the Great Pyramid of Giza is
halfway built. Three and a half million tons of brick rise out of the desert, with 3
million more to follow. The heaviest blocks outweigh two bull elephants. The
square base is 756 feet per side, the length of three New York City blocks.
Already, it is the tallest structure in the world. Upon its completion a decade
from now, when its workforce of 80,000 can finally rest for lemonade, it will
stand 481 feet tall. Five millennia from today, it will still be here, history’s most
enduring skyscraper, the greatest triumph of triangular architecture.

Except it isn’t.
I mean, it is still standing, last I checked. But it is not a victory for triangles.
If you want to see the triangle at work, forget the Great Pyramid, and walk with
me to a vacant lot nearby. There, we find a small team of surveyors carrying a
peculiar loop of rope, in which they have tied 12 equally spaced knots.
 What for? Just watch. After a few steps, three men each grab a knot (#1, #4,
and #8, respectively) and pull the rope tight. As if by magic, it forms a right-
angled triangle. A fourth worker marks out the right angle. They let the triangle
relax once again into a knotty loop. The scene repeats again and again, until the
whole expanse is subdivided into perfect sections of equal size.

If you’ve ever stayed awake through a geometry lesson (and even if you
haven’t), the scene may call to mind Pythagoras’s theorem. This rule states that
if you draw a square on each side of a right-angled triangle, the two smaller
areas taken together will equal the larger one. Or, in modern algebraic terms: a2
+ b2 = c2.

There is an endless roster of such triangles. For example, the sides can have
lengths 5, 12, and 13; or 7, 24, and 25; or 8, 15, and 17; or my personal favorite,
20, 99, and 101. The Egyptians wisely chose the simplest example, the triangle
with side lengths 3, 4, and 5. Hence, the 12 knots.
But this chapter is not about Pythagoras and “his” rule (which earlier
civilizations found without his help). It is about a simpler and more fundamental
property of triangles, a hidden elegance we shall soon discover. The story of the
triangle begins not in a Pythagorean temple, nor atop the Great Pyramid, but
here, in a vacant field. A slack rope transforms into a surveyor’s tool. It’s our
first hint of a power so great it will make the pyramids look crude by
comparison.

2. THREE PLANKS, ONE SELF

Now this tale enters its psychological phase. The triangle must peer into its own
soul and ask the essential question: Who am I?
Am I a shape like any other, indistinguishable except by the tally of my sides
and the number of my corners? The music swells, and the triangle begs the
cosmos for a sign, a vision, a purpose. What, it cries, am I truly made of? Then it
comes, a thunderous voice deep from within.
I am made of three sides.
Okay, maybe not so revelatory, as moments of self-discovery go. It’s a bit
like a therapy patient noticing the couch. But there are hidden depths here. New
truths emerge when we see triangles not as holistic shapes but as assembled
three-part composites.
For example: to make a triangle, not just any three sides will work. Take the
lengths 10 centimeters, 3 centimeters, and 2 centimeters. These three segments
form only a gap-toothed nonshape—an untriangle, if you will. The long side is
too long; the short ones are too short. I dub it a “T. rex triangle,” because the
stubby arms can’t reach across the body.

It’s a universal truth: a triangle’s longest side must be shorter than the other
two put together.
This law is self-evident to a fly walking the perimeter of the triangle. It
knows that a straight path (from A to B) will always be shorter than a detour
(from A to C to B). Thus, the two shorter sides, taken together, must be longer
than the third side.
 This law has a companion, even deeper and more powerful: If three sides do
form a triangle, then they will form exactly one. Given three sides, there is no
room to embellish or improvise. There is a single mold to follow.
For example, let’s agree in advance upon three sides (such as 5 meters, 6
meters, and 7 meters), then build our personal triangles in separate rooms. I
guarantee that we will emerge with identical creations.
Watch: I’ll lay my longest side flat along the ground, then lean the other two
against each other to meet at their tips. Done! Slide the corner to the left, and one
side sticks out; to the right, and the other does. The solution is what
mathematicians call unique. Even without anticipating your method, I know
you’ll arrive at the same solution, because there is no other.

This truth belongs to the triangle alone. No other polygon can claim the
same.
Try it with the triangle’s closest cousin, the four-sided quadrilateral. I lay one
side flat. I stand up the next two vertically. And I lay my final side across the
top, fastening the corners with duct tape for good measure. But then, a wind
begins to blow. My square sways. The uprights tilt, hinging at the corners, and
the whole apparatus starts to collapse like a folding chair. Every moment brings
a new shape, from “square” to “almost square” to “kind of diamondy” to “skinny
rhombus of super pointiness.”

The four sides haven’t yielded a unique shape. Rather, they’ve given us an
infinite family of possibilities. Any member can be rearranged into another by
applying a little pressure.
Thus, we see the triangle’s hidden magic, its secret identity: not its mere
three-sided-ness, but the rigidity it confers.
The Egyptian rope pullers knew this power. By pulling their 12-knot rope
tight, they summoned a Pythagorean triangle into existence, conjuring a right
angle from rope. You might try instead to summon a square, with its four right
angles, but beware: a whole family of unwanted shapes will answer that call.
Even pulled tight, the quadrilateral will twist at its corners, shifting forms,
refusing definition. The same is true of pentagons, hexagons, heptagons, and all
the other polygon cousins. None can do what the triangle can.
The pyramids, being solid, did not take advantage of this special power.
Cubes, cones, frustums—any of these would have suited the pharaohs’ purpose
just as well. The blunt language of stone doesn’t much care what shape it
articulates.
Now, I don’t mean to slight the pyramids. For one, “slight” is not the right
word for a brick pile weighing 20 billion pounds. I admire their otherworldly
precision: sides equal to within 8 inches, facing the cardinal directions with an
error below 0.1°, with corners less than 0.01° from right-angled perfection.
Those Egyptian cats knew their math.
But I must point out that these are a surveyor’s triumphs, not an engineer’s.
The Great Pyramid remains, fundamentally, a big stack of blocks. That’s cool
for a towering symbol of your pharaoh’s immortality, but not great for a building
you’re actually hoping to, y’know, use. The pyramid’s modest chambers and
slivers of passageway occupy less than 0.1% of its internal volume. Imagine if
the Empire State Building were solid steel except for a single floor with a 2-foot
ceiling, and you too will begin aspiring to a more efficient building plan.
In centuries to come, architects would seek new poetries of structure. They
would build bridges wider than the sky, towers taller than Babel. And for this,
they would need a shape of extraordinary resilience, a shape of unique and
unyielding character: a three-sided, three-cornered hero.

3. THE BENDING RAFTERS OF THE BURDENED

WORLD
At this point, our story intersects another—the 10-millennium saga of human
architecture. A quick recap of what you’ve missed:

1. “Outside” is a bad place to live. It can get very cold, there’s nowhere to
keep your stuff, and sometimes bears show up. That’s why humans
invented “inside.”
2. To create “inside,” you make a big empty shape and then live in the
middle of it.
3. If your shape is friendly and made of the right things, then it will be nice
to live inside and won’t fall down on you. This is called “architecture.”

Okay, now that we’re up to speed, I can introduce you to a crucial supporting
character in the triangle’s story: the beam. If you’re an architect aiming to avoid
both (a) pyramid-like monoliths and (b) collapsing floors, then beams will likely
factor in your design.
 The beam’s effect is to take vertical forces and move them horizontally. For
example, picture a plank stretched over a ditch. When you stand on the beam,
your weight pushes it downward. But the real support isn’t below: it’s off to the
sides, where the plank meets the dirt. The beam takes a force in the middle and
moves its impact to the sides.
There’s just one problem: beams are inefficient.
Architecture, like life itself, is all about stress management. Whereas life
offers many types of stresses (deadlines, child-rearing, low phone battery, etc.),
structures experience just two: pushes and pulls. Pushes create compression,
when an object is squeezed tighter. Pulls create tension, when an object is
stretched apart. Each type has its own peculiar character, and different materials
cope in different ways. Concrete can withstand fantastic degrees of compression,
but it crumbles under tension. At the other extreme, steel cables can endure
incredible amounts of tension, but they buckle under the slightest compression.
Now, picture a beam sagging under a load. It curves into a smile (or, more
fitting, a grimace). What’s the nature of its strain: tension or compression?

The answer is “both.” Look at the upper surface of the beam: like a runner on
the inside lane of a track, it has a shorter distance to curve. Thus, its material
squeezes together, creating compression. Now, move your gaze to the bottom:
like a runner on the outside lane of a track, it has farther to curve, so its material
stretches and experiences tension.

So far, nothing to worry about: many materials, such as wood, cope well with
both pushes and pulls. The problem isn’t that the beam experiences both
stresses; it’s that a large fraction of the beam experiences neither.

Look to the center. Halfway between the compression on top and the tension
on bottom, the beam’s middle experiences no strain whatsoever. Its curve is the
carefree smile of a guy who’s not helping. The middle is wasted substance, no
better than the useless bulk of a pyramid. A generic beam squanders half of its
strength, like a student coasting on 50% effort.
Every teacher knows the next two words: That’s unacceptable. In
architecture, every ounce counts, whether you’re building a sky-tickling tower, a
canyon-spanning bridge, or a life-affirming roller coaster.
Rest assured: architects are no fools. They’ve got a plan.

4. THE SHAPE OF THE RESISTANCE

Did I say architects are no fools? I may have to walk that back once you hear
their solution. Since the beam’s top and bottom take all of the strain while the
middle freeloads on their efforts, the brilliant solution devised by architects was
to build beams without middles.
No need to shout; I get it. A beam with no middle is otherwise known as
“two separate beams” and isn’t much of a solution.
Unless… you keep just a little bit of the middle. While you move most of the
substance to the extremes, you leave a thin connective layer in between. The
resulting shape’s cross-section resembles a capital letter I; hence, the term “I-
beam.”

This is a good start. But we’ve still got wasted material in the center. Thus,
we activate phase two of the architects’ plan: start punching holes in the I-beam.
Every gap in the material saves precious resources while costing us almost
nothing in strength. The emptier the space, the greater our savings, which means
our best hope is to leave the I-beam’s middle a hole-riddled web, more vacant
space than solid substance.

But hang on. Before we start punching holes willy-nilly, we need a plan.
What patterns of gaps will minimize the material required while preserving the
structure’s strength and rigidity? Where can we turn for a design that’s simple
and resilient, not to mention well suited to the flat, nearly two-dimensional realm
of the I-beam’s middle?
There is but one shape that can answer this call. The weak-willed square
twists at its corners. The cowardly pentagon collapses under pressure. And don’t
get me started on that spineless turncoat known as the hexagon. Only a
polygonal Superman can absorb the strain with a stoic and unyielding form.
Get me Triangle on the phone.
When you connect triangles into a single structural unit, you create a truss
(from the French word for “bundle”). In a truss, every member experiences
tension or compression. Trusses waste no material, like hunters using every part
of the animal.
In ancient Egypt, the triangle labored in a vacant lot, offering a nifty trick to
surveyors while the spotlight shone elsewhere. Then, millennia later and oceans
away, the triangle moved from the background scenery to center stage.

5. WE BUILT THIS CITY

In the 19th and early 20th centuries, the people of North America tamed a vast
continent. As this continent turned out to be rather bumpy, the project demanded
bridges of every kind, from modest pedestrian footpaths to enormous railways.
These bridges demanded trusses. And what do trusses demand? Triangles, of
course.
The Pratt truss, devised by two brothers in 1844, consists of rows of right-
angled triangles. It swept the United States, staying popular for decades.

The Warren truss, born in 1848, deployed equilateral triangles.

The Baltimore and Pennsylvania trusses, Pratt variants with nested triangles,
became common for rail bridges.

The K truss combined various triangles (while hoping no one will be

reminded of the KKK).

The Bailey truss arrived with the military needs of World War II. Its
standardized modular triangles could be dismantled, shipped, and reassembled to
meet the shifting urgencies of wartime.

It isn’t just bridges. Triangular roofs employ trusses. So do the skeletons of

high-rise buildings. Heck, the standard bicycle frame is nothing but a simple
two-triangle truss. To navigate a modern city is to stroll beneath triangles, to be
supported by them, even to ride upon them.
 Architects are hemmed in by myriad constraints: budgets, building codes,
physical laws. They turn to triangles not in the spirit of artists or interior
decorators, but because there are no other qualified applicants. The marriage
between architecture and triangles isn’t one of love. It’s convenience at best,
desperation at worst. So you might expect the resulting structures to strike us as
last-ditch compromises, crude eyesores.
And yet, they’re lovely. It is a funny paradox of design: utility breeds beauty.
There is elegance in efficiency, a visual pleasure in things that just barely work.
This is, I think, the same pleasure I draw from math. A good mathematical
argument, like a well-built truss, just barely works. Remove one foundational
assumption, and the whole artifice collapses. There is an undeniable grace there:
the minimalism, the mutually supporting elements, the utter strength without an
ounce of excess.
I can’t explain why I find things beautiful. (Case in point: ’90s pop rock.) But
I know there is something gorgeous in the narrative of the triangle. Its three-
sided-ness makes it unique; its uniqueness makes it strong; and its strength
makes it essential to modern architecture. Perhaps it’s a stretch to claim that the
triangle “saved the world,” but if you ask me, it did one better. The triangle
allowed the world to become what it is.
 Chapter 7
IRRATIONAL PAPER

When I moved to England, I was ready to confront the shortcomings of my

American upbringing. Instead of the scientific Celsius, I used the antiquated
Fahrenheit. Instead of the tidy kilometer (each composed of 1000 meters), I used
the idiosyncratic mile (each composed of 5280 feet). Instead of tea, I drank
Starbucks-branded buckets of caramel. So I knew there’d be a rocky adjustment
as I learned the customs of the civilized world.
But there was one culture shock I didn’t anticipate: paper.
Like any Yank, I grew up using “letter paper.” It’s 8.5 inches wide and 11
inches tall, which is why you sometimes hear it called by the catchy name—
really killer branding here—“eight and a half by eleven.” If I’d thought about it,
I’d have recognized that other countries use centimeters instead of inches, and
thus letter paper was unlikely to suit them. As clunky as I find the name “eight
and a half by eleven,” it pales in awfulness compared to “twenty-one and three-
fifths by twenty-seven and nineteen-twentieths.”

Still, when I saw their stuff—which goes by the even less appealing name of
“A4”—I developed a hot and immediate loathing. It’s slightly too skinny, like a
trendy pair of jeans. I was long accustomed to a loose-fit boot-cut style, and my
whole being chafed at this svelte European nonsense. Letter paper has a length
that’s about 30% greater than its width; clearly this ratio was different. And, just
as clearly, worse.
So I went to look up its measurements. I figured it’d be 22.5 centimeters by
28 centimeters, or perhaps 23 centimeters by 30 centimeters. Something nice and
tidy for these metric-minded folk, right?
Nope. It’s 21 centimeters by 29.7 centimeters.
What the heck?
I divided 29.7 by 21 to figure out the simplified ratio: roughly 1.41. As a
math teacher, I recognized this number immediately: It’s (approximately) √2,
a.k.a. “the square root of 2.” And in that instant, my befuddlement gave way to a
snarling, smoking outrage.
√2 is irrational: a word that literally means “not a ratio.”
The paper makers had chosen a ratio that—I can’t put too fine a point on this
—is not a ratio.

We tend to run across two kinds of numbers in life: (1) whole numbers, as in
“I have 3 children,” “My children consume 5 entire boxes of cereal each
morning,” and “The stains on my children’s clothing range over 17 different
colors”; and (2) ratios of whole numbers, as in “We spend 1/4 of our
disposable income on Legos,” “Houses with children are 17½ times more likely
to be the target of Magic Marker graffiti,” and “Hey, when did 2/3 of my hair go
gray?”
(Your everyday decimals, I should point out, are just ratios in disguise. For
example, $0.71 is simply 71/100 of a dollar.)
But some wild and exotic numbers don’t fit into either category. They are not
only unwhole, but unholy: They cannot be written as ratios. (My 12-year-old
student Aadam, who has a far more brilliant mind than I, dubbed them the
“disintegers.”) No fraction or decimal that you can write down will ever quite
nail these numbers. They’ll always slip between the cracks.
And √2—the number that, when multiplied by itself, yields 2—turns out to
be just such a number. Take a look:

There’s no nice decimal that quite equals √2, and there’s no nice ratio, either.
7/5? Close. 141/100? Even closer. 665,857/470,832? So close you can almost
taste it. But never quite right. Never quite √2.
√2 isn’t merely irrational. It is, alongside π, one of the most famously
irrational numbers in all of mathematics. Legend has it that the members of the
ratio-worshipping cult of Pythagoras were so dismayed to find that √2 couldn’t
be written as a fraction that they drowned the mathematician who reported the
discovery.
If European paper aspires to √2, it has chosen a goal that it can never attain.
To put it in a language my British colleagues will understand: That’s bloody
shortsighted, innit?
For a few days, I lived in a state of heightened irritation. To touch this foolish
paper was as loathsome as touching poison ivy or underdesk gum. I joked darkly
about it, the kind of quip that’s intended as charming and world-weary but
contains such concentrated bitterness that people recoil to hear it.
And then, I realized I was wrong.
I didn’t realize it all by myself, of course. I never do. Instead, someone
pointed out to me the wonderful property that A4 paper has.
It belongs to a team.
It’s exactly double the size of A5, quadruple the size of A6, and eight times
the size of the adorably tiny A7. Meanwhile, it’s exactly half the size of A3, a
quarter the size of A2, and one-eighth the size of the impressively ginormous
A1.

As befits the individualistic culture that employs it, letter paper is an island.
Our 8½" × 11" size bears no particular relationship to the smaller and larger
paper types in circulation. It’s a one-off.

By contrast, global paper is as interconnected as the globe itself. A4 belongs

to a unified series of papers: all different sizes but all exactly the same
proportion.
 If you are a serious user of paper—like, say, a math teacher—this is
delightful. You can shrink down two pages to fit perfectly on a single sheet. Or
you can put two pages together to fill a sheet of A3. For those unfeeling and
indifferent people who get their fun at beaches, nightclubs, and French
restaurants, this is perhaps not very stimulating. But for a stationery aficionado,
this is a transcendent thrill. This paper makes sense.
And once I understood this, I saw that the blunder was no blunder at all. It
was inevitable: the only way to make this magical Russian nesting doll system of
paper work properly.
To see why, let us imagine the scene.

INT. PAPER LABORATORY - NIGHT

Gwen and Sven, two beautiful paper scientists, are working on a top-
secret research project. It is code-named “The Prince and the Paper,” or
perhaps “Paper Caper”—whichever sounds cooler in their vague, foreign
accents. It is late. They are exhausted but devoted to their work.

GWEN: All right, Sven. I may not know our precise nationalities, but I
know one thing: The fate of civilization depends on our ability to create a
series of papers where folding one size in half gives you the next size in
the series.

SVEN: The stakes could not be higher. But… what dimensions will such
paper have?

GWEN: There’s only one way to find out.

With a decisive motion, Gwen folds a piece of paper in half, and marks
three lengths: “long” (the length of the original paper), “medium” (the
width of the original paper) and “short” (the width of the half-sized
paper).

GWEN (continuing): Now, what’s the ratio between “long” and

“medium”?

SVEN: That’s what we’re trying to find out.

GWEN: Well, what’s the ratio between “medium” and “short”?

SVEN: Blast it, Gwen! We know that’s the same ratio, but we still don’t
know what it is.

A moment passes, filled with romantic tension.

GWEN: Okay. Let’s say that the “medium” is r times longer than the
“short.”
SVEN: But what’s r?

GWEN: I don’t know yet. I only know that it’s more than one but less
than two, because the “medium” is longer than the “short,” but it’s not
twice as long.

SVEN: All right. And I suppose that makes the “long” also r times longer
than the “medium.”

GWEN: So if you want to turn “short” into “long,” you’d have to multiply
by r (to get “medium”) and then multiply by r again. That’s r squared.
 SVEN (slapping the table): You genius with perfect pores and
inexplicable high heels—Gwen, you’ve done it!

GWEN: Have I?

SVEN: The “long” is r2 times longer than the “short.” But look: it’s also
double the length of the short!

GWEN: My word… you’re right… which means…

SVEN: Yes. r2 is 2.

GWEN: So r is the square root of 2! That’s the secret ratio that will end
suffering and unify humanity!

SVEN (suddenly changing accents): All right, Gwen. Hand over the ratio.

GWEN: Sven? Why are you holding that gun?

Contrary to initial appearances, the makers of A4 paper didn’t choose this ratio
to upset me personally. Nor did they choose it in deference to arbitrary fashions,
or in obstinate defiance of American hegemony, or even for the sadistic pleasure
of selecting an irrational ratio.
In fact, they didn’t really choose it at all.
What they chose was to create a system of paper where folding one size in
half yields the next size in the series. That’s a pretty cool and hard-to-fault
feature. But once they committed to this path, the decision was out of their
hands. There is only one ratio that achieves this property, and it just so happens
to be the famously irrational √2.
Now, I know we all like to imagine paper designers as unshackled fantasists,
bound only by the limits of their imagination. The reality is actually more
interesting. Designers move through a space of possibilities that are governed by
logic and geometry. This landscape has immovable features: some numbers are
rational, some aren’t, and there’s nothing a designer can do about it. Instead, the
designer must navigate around these obstacles—or, better yet, turn them into
assets, like an architect whose building is in harmony with the surrounding
environment.
Long rant made short: I’ve come around on A4 paper. Now that I know why
they’re aiming for √2, the fact that they’re doomed to miss the target
microscopically doesn’t much bother me. To be honest, A4 paper doesn’t even
look wrong to me anymore. Now it’s letter paper that looks off: a little stout and
old-fashioned.
It seems that I’ve completed the transition from one type of insufferable
American to another. From jingoistic defender of my own arbitrary national
customs, I’ve become a fervent evangelist for arbitrary foreign customs. I’m
even drinking fewer caramel buckets these days, although—much like letter
paper—I’m sure I’ll never give it up altogether.
 Chapter 8
THE SQUARE-CUBE FABLES

JUST-SO STORIES OF MATHEMATICAL SCALING

Fables and math have a lot in common. Both come from dusty, moth-eaten
books. Both are inflicted upon children. And both seek to explain the world
through radical acts of simplification.
If you want to reckon with the full idiosyncrasy and complexity of life, look
elsewhere. Ask a biologist, or a painter of photorealistic landscapes, or someone
who files their own taxes. Fable tellers and math makers are more like
cartoonists. By exaggerating a few features and neglecting all the rest, they help
explain why our world is the way it is.
This chapter is a brief collection of mathematical fables. They show how
diverse settings, from baking to biology to the financing of the arts, are governed
by the constraints of geometry. At these tales’ collective heart is a single
fundamental idea, a moral so simple that even Aesop neglected to spell it out:
size matters.
A big statue is not just a big version of a small statue. It is a different object
entirely.

1. WHY BIG PANS MAKE BETTER BROWNIES

You and I are baking. We stir the batter with pride, bettering mankind via
chocolate miracle. But when the oven is preheated, the cupboards defy us. The
only available pan has double the dimensions of the one suggested by the
cookbook.
What do we do?

Needing to fill a double-sized pan, we are tempted to double the recipe. But
this would be a mere half measure. Look carefully and you will see: We need to
quadruple the recipe.
What happened? Well, the flat pan has two dimensions: length and width. By
doubling the length, we doubled the pan’s area. And by doubling its width, we
doubled the area again. That means the area has been double-doubled. Hence, a
multiplier of four.
This happens any time that you grow a rectangle. Triple the sides? Nine times
the area. Quintuple the sides? Twenty-five times the area. Multiply the sides by
9 bajillion? That’ll yield 81 bajillion bajillion times the area.
Or, more precisely: Multiply the lengths by r, and you multiply the area
by r2.
 It’s not just rectangles. The same principle holds for all two-dimensional
shapes: trapezoids, triangles, circles, and any other vessel into which you might
pour the sacred brownie batter. As the length grows, the area grows much faster.
Back in the kitchen, we have just finished quadrupling our recipe when a
forgotten cupboard reveals the pans we’d been looking for all along. We blame
each other, then laugh, because who can stay mad when chocolate glory is so
close at hand?
We now face a choice: Shall we bake the brownies in the one big pan, or in
four small ones?
This is a fable, so we shall ignore the details. Forget oven temperature,
cooking times, heat flow, and minimizing the dishes to clean. Focus instead on
one matter: size itself.
As a brownie pan grows, its exterior (the length of the edges, which is one-
dimensional) gets larger. But its interior (which is two-dimensional) gets larger
faster. This means that small shapes tend to be “edge-heavy” while large shapes
are “interior-heavy.” In our case, the four small pans will have the same area as
the single big one, but they’ll have double the edge length.
The small pans will maximize our quota of edge brownies, while the big pan
will minimize it.
 No matter how I try, I cannot fully embrace the humanity of people who
prefer edge brownies. Who would forsake a fudgy wonder for a chewy chore, a
crispy mistake? I can only imagine that they would prefer bones over meat,
crumbs over crackers, side effects over pain relief. Such people lie beyond
explanation and exculpation. Go big pan, or go home.

2. WHY THE AMBITIOUS SCULPTOR WENT

BANKRUPT
About 2300 years ago, the Greek island of Rhodes repelled an attack from
Alexander the Great. In a mood of self-congratulation, the people commissioned
local sculptor Chares to build a magnificent commemorative statue. Legend tells
that Chares originally planned a 50-foot bronze sculpture. “What if we supersize
it?” said the Rhodians. “You know, double the height? How much would that
cost?”
“Double the price, of course,” said Chares.
“We’ll take it!” said the Rhodians.
But as construction unfolded, Chares found his funds drying up. The material
expenses astounded him; they far exceeded his budget. Bankruptcy loomed.
According to some, Chares chose to escape financial ruin by taking his own life,
and never saw the masterpiece finished. But perhaps, before he died, he
understood his mistake.
He was too shy about jacking up the price.
To see why, forget all the details. Pay no mind to the labor market for Greek
construction workers, or the pricing of wholesale bronze. Heck, forget artistry
altogether: imagine Chares was just building a giant bronze cube. We must focus
on a single overpowering question: size.
What happens when you double the dimensions of a 3D shape?

Well, you’ve doubled the length; that doubles the volume. You’ve doubled
the height, which doubles the volume again. And you’ve doubled the depth,
which doubles the volume a third time. That’s a triple double, although not in the
Russell Westbrook sense: here, double-double-double amounts to “multiply by
eight.”
The results are as clear as they are startling: Volumes get big fast. Triple the
sides of a cube? That’s 27 times the volume. Multiply the sides by 10? You’ll
multiply the volume by an absurd factor of a thousand. And what’s true of cubes
is true of every shape: pyramids, spheres, prisms, and (unlucky for Chares)
exquisite statues of the sun-god Helios. In precise terms: Multiply the length by
r, and you multiply the volume by r3.
 If Chares had created a one-dimensional artwork (The Colossally Long String
of Rhodes), then the aesthetics might have suffered but his price scheme would
have worked fine: doubling the length would double the required material. Or
suppose he’d been commissioned to create a two-dimensional painting (The
Colossal Portrait of Rhodes). He’d still be lowballing the contract, but less so:
doubling the height of a canvas quadruples its area, demanding four times the
paint. Alas, Chares had the misfortune to work in a full three dimensions.
Doubling the height of his statue multiplied the bronze requirements by eight.
When a 1D length grows, a 2D surface grows faster, and a 3D volume grows
faster still. The Colossus of Rhodes—one of the wonders of the ancient world—
doomed its own creator for the simple reason that it was three-dimensional.
 3. WHY THERE AREN’T GIANTS
King Kong, the three-story ape. Paul Bunyan, the lumberjack whose footsteps
carved out lakes. Shaquille O’Neal, the mythical seven-foot one, 325-pound
basketball player who could do everything except make free throws. You know
these stories, and you know just as well that they are fantasies, legends, wide-
eyed fictions. There’s no such thing as giants.
Why? Because size matters.
Suppose we take exemplary human specimen Dwayne Johnson and double
his dimensions. Now with double the height, double the width, and double the
depth, Dwayne’s total body mass has grown by a factor of eight.

So far, so good. But take a glance at his legs. To keep our man standing, his
bones will need eight times the strength. Can they muster it?
I doubt it. They’ve undergone two helpful doublings (in width and depth) but
one useless doubling: length. Just as you don’t reinforce a column by making it
taller, you can’t bolster a leg by making it longer. Extra length confers no extra
strength, just extra strain, as the bone’s base must now support the greater mass
above.
 Dwayne’s leg bones won’t keep pace with the demands placed on them: A
factor of four can’t match a factor of eight. If we keep growing Dwayne
Johnson, doubling and tripling and quadrupling his size, then eventually he will
reach a critical breaking point. His limbs will buckle and splinter beneath the
overwhelming weight of his torso.
We’re dealing with a process called isometric scaling: growing a shape while
preserving its proportions. (Iso- means “same”; -metric means “measurement.”)
It’s a lousy method for making big animals. Instead, we need allometric scaling:
growing a shape while altering its proportions accordingly.
When an animal’s height grows by 50%, its legs can keep pace with the
strain only by growing 83% thicker. That’s why cats can survive on slender
limbs, whereas elephants need pillars to stand on.
 A constraint upon Dwayne Johnson is a constraint upon us all, and thus,
giants shall always belong to the realm of myth. Paul Bunyan’s tibiae would
shatter with every lake-punching step. King Kong’s muscles (their strength
growing with the square of his height) could never carry his bulk, which grows
“with the cube”: he’d merely sit there, a giant monkey puddle suffering
perpetual heart failure. And Shaquille O’Neal? Well, his tale is so implausible
that I doubt anyone truly believes it.

4. WHY NO ANT FEARS HEIGHTS

Ants are horrifying. They lift objects 50 times their body mass, work together in
flawless coordination, and thrive on every corner of the planet. This global army
of weightlifting pincer-jawed telepaths outnumbers humans by a factor of
millions. Visions of their alien faces would keep me up at night if not for one
redeeming fact:
Ants are very, very small.
It’s time now to solidify the pattern of our earlier fables. When a shape’s
length grows, its surface area grows faster, and its volume grows faster still.
 This means that big shapes (like humans) are “interior-heavy.” We have lots
of inner volume per unit of surface area. Small shapes (like ants) are the
opposite: “surface-heavy.” Our formic nemeses have lots of surface area relative
to their tiny interior volume.
What is it like being surface-heavy? First, it means you never need to fear
heights.
When you fall from a great height, two forces play tug-of-war: downward
gravity and upward air resistance. Gravity works by pulling on your mass, so its
strength depends on your interior. Air resistance works by pushing against your
skin, so its strength depends on your surface.
In short, your mass speeds up your fall, while your surface area slows it
down. That’s why bricks plummet and paper flutters, why penguins can’t fly and
eagles can.
 You and I are like penguins: lots of mass, not much surface area. When
falling, we accelerate to a top speed of nearly 120 miles per hour, which is a
rather disagreeable way to hit the ground.
By contrast, ants are like paper eagles: lots of surface area, not much mass.
Their terminal velocity is just 4 miles per hour. In theory, an ant could leap from
the top of the Empire State Building, land with all of its six feet on the
pavement, and walk away singing “Ants Marching.”

So is being surface-heavy all fun, games, and parachute-free skydiving? Of

course not. Ants have woes, too, and by “woes” I mean “a debilitating and
wholly justified fear of water.”
The trick is surface tension. Water molecules love sticking together, and
they’re willing to defy small amounts of gravity to do it. Thus, when any object
emerges from a bath, it carries with it a thin layer of water, about half a
millimeter thick, held in place by surface tension. For us, that’s a trifle: a pound
or so, less than 1% of our body weight. We towel off and go on living.
Compare that to the ordeal of a bathing mouse. The water layer’s thickness
remains the same, about half a millimeter, but our rodent friend finds this a far
greater burden. The surface-heavy mouse steps out of the tub carrying bathwater
that equals its own body weight.
And for an ant, the situation is dire. The clinging water outweighs its body by
an order of magnitude; a single wetting can be fatal. That’s why ants fear water
the way I fear them.

5. WHY BABIES NEED BLANKIES

Although the cutting edge of parenting advice evolves minute by minute, some
principles never change: cuddles are good; head trauma is bad; and swaddling
your little one is a must. We’ve been wrapping our young since the Paleolithic,
and thousands of years from now, I’m sure that the survivors of the zombie
apocalypse will still be swaddling their traumatized infants.
Babies need blankets because—please forgive the technical jargon—babies
are small.

Once again, ignore the details: the tiny toothless mouth, the eensy wiggling
toes, the small bald head that smells so amazing. Think of a baby the way you’d
think of any organism: as a homogenous bundle of chemical reactions. Every
activity of the body is built on such reactions; in some sense, the reactions are
the creature. That’s why animals are so temperature-sensitive: Too cold, and the
reactions slow to a halt; too hot, and some chemicals deform, making key
reactions impossible. You’ve got to keep a close eye on the thermostat.
Heat is created by reactions in each cell (i.e., in the interior). And heat is lost
through the skin (i.e., at the surface). This creates a familiar tug-of-war: interior
vs. surface.
 Bigger animals, being more interior-heavy, will have an easy time keeping
warm. Smaller ones, being surface-heavy, will struggle. That’s why you’re most
vulnerable to cold in your surface-heavy extremities: fingers, toes, and ears. This
also explains why cold climates support only big mammals: polar bears, seals,
yaks, moose, mooses, meeses, and (depending on your zoology professor) the
Sasquatch. A surface-heavy mouse wouldn’t stand a chance in the Arctic. Even
at moderate latitudes, mice cope with heat loss by eating a quarter of their body
weight in food each day.
A baby isn’t a mouse, but it’s definitely not a yak. Its tiny body expends heat
like governments spend money. And to stifle that heat loss, there’s no cuddlier
option than a blankie.

6. WHY AN INFINITE UNIVERSE IS A BAD IDEA

Once you start looking for them, square-cube fables are everywhere. Geometry
governs every design process, and no one can defy its rules—not sculptors, not
chefs, not the forces of natural selection.
Not even the cosmos itself.
Perhaps my favorite square-cube fable of all is the “paradox of the dark night
sky.” It dates to the 16th century. Copernicus had just laid out the notion that
Earth was not the center of existence, that ours is an ordinary planet orbiting an
ordinary star. Thomas Digges, importing Copernicus’s work to England, took a
further step. He proposed that the universe must go on forever, a sparse and
ageless cloud of far-flung stars, extending into infinity.
Then Digges realized that if that were true, the night sky should be a
blinding, scorching white.
To see why, we’re going to need a magical pair of sunglasses. I call ’em
“Digges shades.” These sunglasses have an adjustable dial and a wondrous
property: they block out all the light from beyond a certain distance. For
example, set them to “10 feet,” and most of the world goes dark. Sunlight,
moonlight, streetlamps—all of these vanish, leaving only the light sources within
10 feet of your current position. A reading lamp; an iPhone screen; perhaps
nothing more.

Let’s tune our Digges shades to “100 light-years.” Gazing up at the night sky,
we say goodbye to Rigel, Betelgeuse, Polaris (a.k.a. the North Star), and many
other familiar orbs. The remaining stars, those within a hundred-light-year
radius, number about 14,000—a sparser, dimmer heavens than the usual.
Altogether, these stars will have some total brightness. We’ll give a lowball
estimate according to the following equation:

Turn the dial on your Digges shades, doubling the distance: 200 light-years.
Stars reappear. The sky brightens. But by how much?
 Well, the visible sky forms a 3D hemisphere around Earth. By doubling its
radius, we’ve octupled its volume. Assuming (as Digges did) that the stars
follow an even distribution, like trees in a forest, then we’re now seeing eight
times as many stars as before. Their population leaps from 14,000 to more than
100,000.

But these new stars are farther, which means they’ll be dimmer. The
question, once again, is: By how much?
Each star appears in the night sky as a tiny circle. The bigger the circle, the
more light hits our eye. Since this is a fable, we can ignore personality
differences among stars—their temperatures, their colors, whether they have
cool names like Betelgeuse and Polaris. We’ll assume, like unapologetic star-
racists, that all stars are identical. The only thing that matters is their distance.
If Star A is twice the distance of Star B, then its circle will be half the height
and half the width. That leaves it with a quarter of the area—and thus, a quarter
of the brightness.
 So, what can we make of our new and enlarged night sky? The number of
visible stars has been multiplied by eight; their minimum brightness has been
divided by four; and (correct me if I’m wrong) eight divided by four is two,
meaning that the total brightness, according to our simple equation, has doubled.
Double the radius on your Digges shades? Double the brightness of the sky.

Because “number of stars” is 3D and “minimum brightness” is 2D, the sky

becomes brighter and brighter the more we spin our dial. Triple the radius, to
300 light-years? You’ll triple the original brightness. Multiply the radius by a
thousand? Our night sky is now a thousand times brighter than before.
You can see where this is going—or rather, you can’t, because before long
you are literally blinded by the stars. The night sky can grow a million, a billion,
a trillion, even a googol times brighter than the initial level. Given a large
enough radius, the stars outshine the day and overpower the sun, turning the sky
into a scorching miasma of infinite temperature. Compared to this nonstop
vaporizing blast of unfathomable brightness, the dinosaurs got off easy with a
measly asteroid.
Take off your Digges shades, and the totality of the night sky will brighten
you to death.
If the universe is infinitely large, why isn’t our night sky infinitely bright?
This paradox endured for centuries. In the 1600s, it nagged at Johannes Kepler.
In the 1700s, it spurred Sir Edmund Halley. In the 1800s, it inspired an Edgar
Allan Poe prose poem that the author called his greatest work. (Critics
disagreed.)
Only in the 1900s did the paradox find a happy resolution. The deciding
factor is not size, but age. Whether or not the universe is infinitely big, we know
for certain that it is finitely old, born less than 14 billion years ago. Thus, any
settings on your Digges shades beyond “14 billion light-years” are useless. The
light hasn’t had time to reach us, so there is nothing at that distance to block out.
This square-cube fable is more than just a clever argument. It was an early
line of evidence for the big bang theory. To me, this is the apotheosis of the
square-cube spirit. By thinking about the simplest aspects of our universe—
what’s 2D, what’s 3D—we can reach sweeping levels of understanding.
Sometimes it takes an act of radical simplification to see the world as it truly is.
 Chapter 9
THE GAME OF DICE

FOR 1 TO 7,500,000,000 PLAYERS

Thank you for purchasing the Game of Dice! This fun pastime for the whole
civilization is suitable for ages from “stone” to “digital,” and beloved by gamers
from the commoner to the tyrant. Don’t take my word for it! Just ask these
Roman emperors:

This instruction manual will introduce you to the basic rules. The Game of
Dice is equal parts theory and practice, stretching the mind and the fingertips
alike. Let’s play!

THE GOAL OF THE GAME:

Enthrall and entertain humankind with a device that generates
outcomes beyond its control.

We begin with a class of characters known as “people.” These creatures love

being in control, which is why they invented cars, guns, governments, and
central air-conditioning. But they also obsess over what’s out of their control:
traffic, weather, their children, and the success of famous guys who play sports
for money.
In their hearts, they wish to confront fortune, to hold their own powerlessness
in their palms. This is where dice come in. They are handheld pieces of fate.
In 6000 BCE, ancient Mesopotamians tossed stones and seashells. Ancient
Greeks and Romans rolled the knucklebones of sheep. The first Americans threw
beaver teeth, walnut shells, crows’ claws, and the pits of plums. In the Sanskrit
epics of ancient India, kings cast handfuls of hard bahera nuts. These natural
dice enabled games of chance, the telling of fortunes, the dividing of prizes, and
(no doubt) other customs ranging from the sacred to the mundane. Like desserts
or naps, the idea of dice was so obvious and beautiful that every society came to
it on its own.
Nowadays, only a few die-hard traditionalists play Monopoly with beaver
teeth. Civilizations have advanced from “found” dice to designed ones. This is
where the true Game of Dice begins.

RULE #1: GOOD DICE PLAY FAIR.

When you roll a die, each face should have an equal chance of coming up.
Otherwise, people get cranky and suspicious and stop inviting you to their
backgammon parties.
A useful starting point: congruence. Two shapes are congruent if one can be
superimposed upon the other. Congruent shapes are identical twins, in that their
angles and edges all match. Thus, your first idea for a fair die might go
something like: make sure the faces are congruent.
This sounds great… until you meet the snub disphenoid.

This snout-faced mole rat of a polyhedron dashes our hopes. Its 12 faces are
equilateral triangles, all identical. Yet it does not make a fair die. At some
corners, four triangles meet. At others, five do. When you roll the little monster,
some faces come up more than others. I’m afraid congruence just isn’t enough.
We need symmetry.
In ordinary speech, “symmetry” refers to a vague, pleasing uniformity. Its
mathematical meaning is more specific: a geometric action that transforms an
object without really changing it. For example, a square table has eight
symmetries:
 The symmetries leave the table unchanged, though on careful inspection, they
scramble the locations of the corners. For example, the 180° rotation swaps pairs
of opposite corners: #1 and #3 trade places, as do #2 and #4. Contrast this with
the diagonal reflection, which swaps #2 and #4, but leaves #1 and #3 where they
are. Symmetries on dice work in much the same way: they rearrange the faces
while leaving the overall shape unchanged.
Symmetry offers a surefire path to fair dice. Just pick a shape with enough
symmetries that every face can be swapped with every other.
One example: the dipyramid. Take two identical pyramids and glue their
bases together. The right combination of rotations can swap any triangular face
with any other, meaning the faces are geometric equivalents, and so the die is
fair.
 Another example: the trapezohedron, which looks like a dipyramid with
some nifty carving at its equator, to turn the three-sided triangles into four-sided
kites.
 With a dipyramid or a trapezohedron, you can create any even-sided die:
eight-sided, 14-sided, 26-sided, 398-sided. In theory, every single one will be
fair, each face appearing with equal likelihood. You might think that we’ve
solved the problem. The Game of Dice is over, yes?
Not so fast! People are fickler than that. It’s not enough for dice to be fair…

RULE #2: GOOD DICE LOOK FAIR.

We’ve met dice candidates that (1) are easy to define, (2) produce fair outcomes,
and (3) come adorned with awesome Greek and Latinate names. Yet these
promising models prove historical duds, the failed presidential nominees of the
randomization world. To my knowledge, no dice-rolling culture has embraced
the dipyramid, and only one welcomes the trapezohedron: the shadow society
known as “Dungeons & Dragons players,” whose 10-sided die (or d10) is
trapezohedral.
 Why so finicky, humankind? How can you spurn good shapes, cast fair dice
aside?
Roll a skinny dipyramid, and you’ll see the problem. It doesn’t tumble.
Instead, stabilized by its two pointy ends, it just sort of rolls, like a crooked
paper-towel tube, rocking a little back and forth between the two clusters of
faces. Its stops are precarious; a jokester’s breath could tip it from one face to
another. That’s a recipe not for a fun game of Parcheesi, but for a family-
shredding afternoon of recriminations.

The best dice need more than symmetric faces. They need symmetric
everything. And if you’re a polyhedron aficionado, you know what that means.
Platonic solids!
Of all the straight-edged 3D shapes, the Platonic solids are the most perfect.
Their symmetries can swap any two faces, any two corners, or any two edges—
symmetries so magnificent that even a hardened cynic cannot doubt their
eminent fairness.
There are five Platonic solids: no more, no less. And each member of this
geometric pantheon has graced Earth in the form of a die.

1. The tetrahedron, a pyramid of equilateral triangles. In 3000 BCE,

Mesopotamians rolled tetrahedra to play the Royal Game of Ur, an early
precursor to backgammon.
2. The cube, a prism of squares. Simple, sturdy, and easy to manufacture, it
remains history’s most popular die. Our oldest specimen, a cube of baked
clay excavated in northern Iraq, dates to 2750 BCE.

3. The octahedron, a special dipyramid of equilateral triangles. Egyptians

sometimes entombed such dice alongside their dead.
4. The dodecahedron, a pleasing gemstone of 12 pentagonal faces. They
helped tell fortunes in 1500s France. Astrologers today enjoy their
correspondence to the 12 signs of the Zodiac.

5. The icosahedron, a figure made of 20 equilateral triangles. Though

 ubiquitous in Dungeons & Dragons, it enjoys greatest favor among
fortune-tellers. That includes Magic 8 Balls, which are icosahedra
floating in water. If you’ve ever shaken one, then you’ve asked a Platonic
solid about your future.

In the Game of Dice, the Platonic solids are the trump cards. One cannot
picture them without heavenly choirs piping up in the background. But, as an
exclusive pantheon of five, they are all too rare. They let us randomize among 4,
6, 8, 12, or 20 outcomes… but no others.
It’s worth trying to break out of the box. Why not a paradigm-busting design,
a fresh and innovative way to randomize any number of outcomes?
Spoiler: It’s harder than it sounds.
 RULE #3: GOOD DICE WORK WELL ANYWHERE.
One alternative is “long dice.” Instead of worrying about every face having an
equal chance, build a long prism instead.

These dice work not because all faces come up with equal likelihood, but
because two of them never come up at all. Long dice play fair, look fair, and
allow you to randomize among any number of outcomes. So why aren’t they
more popular?
Well… they roll too much.
 Whereas Platonic solids treat the table like a dance floor, leaping this way
and that, long dice roll in a single direction. You’ve got to clear a whole
bowling-alley path for them. What kind of self-important die demands a red
carpet?
Back to the chalkboard we go, to recruit another mathematical principle:
continuity.
Grab your long die from wherever it rolled to a stop. I’ll wait. (And wait.
And wait.) Now, as we can see, the two bases almost never come up. But
imagine shortening the die, to make a “less long” variant. The shorter it grows,
the more probable the two bases become. After a while, you’re left with a die so
short that it’s basically a coin, at which point the roles reverse. Now, almost
every toss lands on a base.
Somewhere in the middle, there is a moment of crossing over: a length at
which the bases and the lateral faces enjoy the same likelihood. Somewhere in
the middle is a fair die.
 In theory, you can pull this trick with any polyhedron, achieving shapes that
are funky-looking yet fair. So where are they? Why aren’t novelty shops selling
nifty trick dice that look wild but play well—the bizarro antithesis of loaded
dice?
It’s because the adjustments are too sensitive. Fair on hardwood? Unfair on
granite. Fair at one size? Unfair when built double-sized. Fair on one roll?
Unfair at a different strength, or with a different spin. Change any condition, no
matter how incidental, and you alter the physics. Such dice would be chained to
the hyperspecific circumstances of their birth. People want portable, durable dice
—not fussy prima donnas.

RULE #4: THEY’RE SIMPLE TO ROLL.

Say we want a random letter from the 26 in the alphabet. The icosahedron has
too few faces. The dipyramid wobbles funny. The long die rolls into the distance
like a runaway meatball. Ruling out those options, we’re stuck. Is there no way
to accomplish the simple task of picking a random letter?
Sure there is. Just flip five coins.
There are 32 outcomes, all equally likely. Assign a letter to each of the first
26, and interpret the leftovers as “Start again.”

This procedure works for any randomization scenario. Say we want to pick a
random word from the Lord of the Rings trilogy. There are about 450,000 to
choose from. So flip 19 coins, yielding more than 500,000 possible outcomes.
Assign a word to each. If you land on one of the unassigned outcomes, just flip
all 19 coins again.
Heck, you don’t even need 19. Just flip a single coin 19 times.
 By this logic, there is nothing dice can accomplish that a single coin cannot
match. And yet it’s hard to picture a Vegas casino drawing in crowds with a
coin-only craps table or a penny-based roulette wheel.
The problem is obvious: these systems are too complicated. It’s an awful lot
of trouble to record a sequence of coin flips, look up the result in an index, and
then (if necessary) repeat the whole process. You want a single roll. No sides
wasted. No user manual required.
This principle nixes some clean mathematics. For example, you could
randomize among four outcomes with a simple cubical die. Just label two of the
faces “roll again.” But this approach bugs people. It feels inelegant to waste
faces. When sharing a cake among four friends, you would never cut six pieces
and then throw out the two extras.
 I suspect that’s why Dungeons & Dragons players roll a four-sided die
instead. I read this as a sign of desperation, because of all the Platonic solids, the
tetrahedron is history’s least popular die. Again, any human can see why: It
lands with a face down but not a face up. This feels wrong, like asking someone
to guess the number that you’re not thinking of.

Across the millennia, humans have shunned the tetrahedron in favor of dice
that have parallel pairs of opposite faces, so that every “down” face corresponds
to an “up.” Mathematics doesn’t care. But humans are the ones calling the shots.
 RULE #5: THEY’RE HARD TO CONTROL.
Remember the whole purpose of the die? It puts the human body in contact with
higher forces: random chance, karmic fate, the will of the gods. This enables
brainstorming, games of luck, the telling of fortunes, and other profound
expressions of our humanity.
So of course, people try to cheat.
One path: manipulate the exterior of the die. For example, extend it by an
undetectable increment, forming a brick shape. Or create faces with slight bulges
(making them less likely) or indentations (making them likelier). Or you can cap
some faces with a bouncy material, or sandpaper them down so the die prefers
them. These tricks are as old as ruins. I mean that literally: loaded dice with
shaved corners have been excavated in Pompeii.
 A second approach: manipulate the interior of the die. “Trappers” are dice
with two hidden compartments; the right motion will move a heavy blob of
liquid mercury from one compartment to the other, shifting the probabilities. (If
poison metals aren’t your cup of tea, then use wax that melts just below body
temperature.) Another scheme: Back when wooden dice were popular, some
cheaters grew small trees with pebbles embedded in the branches. Then they’d
carve a die from the wood, leaving that invisible rocky weight within. This con
job demands not just extraordinary patience but above-average botanical skill.
 A third approach: renumber the faces. A normal die features opposite faces
that sum to seven. (The pairs are 1 and 6, 2 and 5, and 3 and 4.) In cheat dice
known as “taps,” some numbers are duplicated, yielding (for example) opposite
pairs of 6 and 6, 5 and 5, and 4 and 4. From any angle, your victims will see only
three faces, so nothing appears to be wrong.

Though all of these cheating methods target cubical dice, it’s not that cubes
are extra vulnerable. They’re just extra popular. Evidently, there’s more money
in craps than in D&D.
 RULE #6: GOOD DICE FEEL NICE.
The Game of Dice is like most games, in that nobody really needs it. We live in
the 21st century. I jetpack to work and take vacations by flying car. Okay, I
don’t, but I’ve got half the world in my pocket in the form of a 5-ounce
computer. Technology is making us all obsolete, and that includes the blue-
collar workers called dice. Watch: I’m going to stop writing and go simulate a
million rolls of a cubical die in Microsoft Excel. I’ll let you know how long it
takes.
Okay, done. Took about 75 seconds. Here are the results:
 Not only is computer randomization faster and easier than throwing a plastic
cube across a table, it’s more random, too. Casinos could eliminate craps tables
and roulette wheels tomorrow, and their digital replacements would outperform
the old dinosaur randomizers by leaps and bounds.
But what fun would that be?
 Dice are meant to be held. The first time I played Dungeons & Dragons
(okay, the only time), I encountered something more alluring than any of the
orcs and mages in the game itself: the Zocchihedron, the 100-sided die. Can you
imagine? A hundred sides! A die that rolls for 30 seconds before coming to a
stop! I knew that two d10s (one for the tens digit, one for the units) served better
and was fairer than Lou Zocchi’s lumpy shaved golf ball. I didn’t care. I wanted
to roll that d100.

Greeks must have felt that same allure when tossing the sheep knuckle they
called the astragalos. They valued its four sides at strange values (1, 3, 4, and 6)
and threw them by the handful. A throw of all 1’s was “the dogs,” the worst
throw imaginable. The best throw (either all 6’s or one of each face, depending
who you ask) was called “Aphrodite.” The astragali weren’t fair; they were
something better. They were skeletons held in a living hand to foretell fate.
When Julius Caesar crossed the Rubicon, tipping history toward the end of the
Roman Republic and the dawn of empire, his comment was “Alea iacta est”:
The die is cast.
 The Game of Dice, I suspect, will never end. These objects speak to
something deep in our bones. Just remember to follow those six rules:

Good dice play fair.

Good dice look fair.
Good dice work well anywhere.
They’re simple to roll.
They’re hard to control.
And good dice feel nice.
 Chapter 10
AN ORAL HISTORY OF THE DEATH STAR

MEMORIES OF THE GALAXY’S MOST FAMOUS SPHERE

Perhaps the greatest construction project in the history of geometry is the Death
Star. Before being destroyed by a blond desert-boy in the tragic finale to the film
Star Wars, it was pure terror. It was sheer beauty. It was a near-perfect sphere, a
hundred miles across, armed with a planet-vaporizing laser. And yet even this
behemoth, designed to compel the obedience of an entire galaxy, could not help
but obey a higher master in turn: geometry.
Geometry yields to no one, not even evil empires.
I convened the team responsible for creating the Death Star to discuss the
geometry behind history’s most controversial solid. They brought up several
considerations involved in building a tremendous spherical space station:
• Its manifold symmetries
• Its near-perpendicular surface relative to the direction of travel • Its
 • Its near-perpendicular surface relative to the direction of travel • Its
gravitational properties relative to naturally arising spheres • Its personnel
capacity as a function of surface area • Its very slight curvature at large
scale
• Its uniquely low surface-area-to-volume ratio
Even these powerful minds—architects, engineers, and grand moffs among
them—could not dictate terms to geometry. Rather, they had to apply their
ingenuity to work within its constraints. Here—in their own words—is how they
did that.
Take note: For improved readability, I have edited out the loud, ominous
breathing noises.

1. TO FRAME THY FEARFUL SYMMETRY

GRAND MOFF TARKIN

Our aim: To cow the galaxy with the greatest structure it had ever seen. Our
funding: virtually infinite, thanks to the Emperor’s unflinching approach to
taxation. The sky was—quite literally—the limit. So our first question was:
What should this thing look like?

IMPERIAL GEOMETER
I was told to find a simple, elemental design. Something so unnerving and
spectacular that it would make droids weep and bounty hunters lose bladder
control. Compared to peer-reviewing articles, it was a tough assignment.

GRAND MOFF TARKIN

That poor geometer spent a month brainstorming, filling notebooks with
sketches and design ideas… all of which Lord Vader HATED.

DARTH VADER
A hexagonal prism? Was this to be an empire of honeybees?

IMPERIAL GEOMETER
Even in my deepest valleys of self-doubt, I was grateful for Lord Vader’s
feedback. Like a lot of visionaries, he is an exacting manager. But I know it’s all
meant as constructive criticism.

DARTH VADER
Imbecile.
 IMPERIAL GEOMETER
Eventually, we identified a single surpassing goal: symmetry.
Most people use the term casually, but in mathematics, “symmetry” has a
precise meaning: it’s a thing you can do to a shape that leaves it looking the
same.
For example, a Wookiee face has a single symmetry: you can reflect it in a
vertical mirror. That’s it. If you do anything else—say, rotate it 90°, or reflect it
in a horizontal mirror—then you rearrange the face entirely, after which the
Wookiee may undertake to rearrange YOUR face.

By contrast, the face of a dianoga—a tentacle-festooned swamp dweller,

sometimes found in trash compactors—has three symmetries: two reflections,
and a rotation of 180°.
 GRAND MOFF TARKIN
Why the obsession with symmetry? Well, symmetry is the essence of beauty.
Take faces. No one’s is perfectly symmetrical. One ear is a little higher; one
eye is a little larger; the nose is a bit crooked. But the more symmetrical the face,
the more beautiful we find it. It’s a strange fact of psychology: mathematics is
the better part of beauty.
With the Death Star, we sought to create a force of awe and intimidation as
powerful as the face of a supermodel.

IMPERIAL GEOMETER
One day, Lord Vader swept my illustrations from the table and roared, “MORE
SYMMETRY!” We’d been considering an icosahedron, which has 120
symmetries. How could I top that? But then, in the proudest moment of my
career—heck, of my LIFE—it came to me: the maximally symmetric shape.

DARTH VADER
Why did he not think of a sphere to begin with? One loses so much time to the
glacial workings of feeble minds.

GRAND MOFF TARKIN

 GRAND MOFF TARKIN
The headache wasn’t over. We needed to install the planet-destroying laser array
in the northern hemisphere, and thereby tarnish the symmetry. That ruffled some
feathers.

IMPERIAL GEOMETER
I still maintain it was a mistake to include that weapon. I mean, which is more
effective for intimidation: a little laser light show, or an INFINITE number of
symmetries?

2. THROWING AERODYNAMICS TO THE WIND

GRAND MOFF TARKIN

Right away, we encountered a problem. Ran into a headwind, you might say.
Back then, everybody was into Star Destroyers: these sleek, angular designs.
They were galactic steak knives, ready to pop a star like a helium balloon. I
thought the appeal was aesthetic, but when I learned the design was functional,
too, it spelled trouble for our sphere.

IMPERIAL PHYSICIST
Imagine you’re flying an airplane. No matter how good a pilot you are, you’re
going to have A LOT of collisions. I’m referring, of course, to air molecules.
Best-case scenario? The air molecules travel PARALLEL to your surface.
Then, they won’t impact you at all. They’re like passing traffic in the
neighboring lane. The worst-case scenario is that the air molecules hit
PERPENDICULAR to your surface, at 90-degree angles. Then, your vessel
bears the full force of the impact. That’s why you don’t build airplanes with big,
flat fronts: it’d be like trying to weasel through a crowd while wearing a giant
sandwich board on your torso.
Hence, the tapered design of the Star Destroyer. When it’s traveling through
atmosphere, the air molecules mostly glance off the sides—not quite parallel to
the surface, but close enough. The Death Star, by contrast, gave us all
aerodynamics nightmares. There’s a huge surface hitting the air at near-perfect
right angles.
 IMPERIAL ENGINEER
Imagine your friends are launching paper planes, and instead of joining in, you
decide to throw a desk across the room. It’s going to cost a lot more energy, and
the flight won’t be pretty.

GRAND MOFF TARKIN

Originally, we intended the Death Star to visit planets directly. It would enter the
atmosphere, vaporize a continent or two, and blast “The Imperial March”
through loudspeakers.
That dream died the moment we chose a sphere. Thwarted by aerodynamics,
we were forced to keep our station up in the vacuum of space. No air resistance
—but no music, either.

DARTH VADER
It was a harsh sacrifice. But leadership is no place for the fainthearted.

3. TOO BIG TO FAIL, TOO SMALL TO SPHERE

GRAND MOFF TARKIN

Soon, another obstacle leapt into our path: our physicist kept insisting that the
Death Star really ought to be lumpy and asteroid-shaped.

IMPERIAL PHYSICIST
Glance around the galaxy. Where do you see spheres? It’s the big, heavy stuff.
Stars, planets, a few larger moons. Now look at the smaller, less dense objects:
asteroids, comets, dust clouds. You find a lot of weird potato shapes.
That’s not a coincidence. It’s a fact of gravity. I told them from the start: The
Death Star is too small to become a sphere.

GRAND MOFF TARKIN

You should have seen Lord Vader’s face.

DARTH VADER
I was about to break ground on the most ambitious construction project in the
history of evil, and a lab coat with glasses was telling me it was TOO SMALL.
Was I angry? You tell me.

IMPERIAL PHYSICIST
Look, I’m no propaganda expert, but the physics is pretty clear here. All matter
attracts all other matter. More matter, more attraction. That’s gravity.
So, toss a bunch of ingredients together in the mixing bowl of space, and
every bit is mutually drawn toward every other bit. They congregate around a
kind of 3D balancing point: the center of mass. Over time, the outlying clumps
and more distant protrusions are drawn toward this center, until it reaches the
final equilibrium shape: a perfect sphere.
But that’s only if you’ve got enough matter. Otherwise, the attraction is too
weak to conquer those lumpy outcroppings. Hence, big planets go spherical, but
small moons stay potatoes.
 DARTH VADER
I wondered: Are all physicists this insolent? Should I drain the life out of every
scientist right now, to spare myself future encounters?

IMPERIAL PHYSICIST
The magic size, where you’re big enough to go spherical, depends on what
you’re made of. Ice will go spherical at a diameter of about 400 kilometers,
because it’s pretty malleable. Rock is much stiffer and needs more gravitational
persuasion—so it won’t go spherical until the diameter hits 600 kilometers. For a
material like imperial steel, designed to withstand tectonic-level forces, it’d be
even larger. Maybe 700 or 750 kilometers.
And the Death Star? It was only 140 kilometers across. A pebble.

GRAND MOFF TARKIN

Just as Lord Vader began to Force-choke the physicist, I was able to strike a
compromise. “Gentlemen,” I said, “this is a GOOD thing! It means that people
seeing our artificial sphere will unconsciously assume that it’s far bigger than it
is. In effect, we’ll triple our diameter without an extra dime for construction!” I
also pointed out that we control the gravity anyway—as well as the physicist’s
air supply. At last, the physicist saw the wisdom.

IMPERIAL PHYSICIST
You could see Lord Vader’s face light up. Well… I think you could. His mask
looked smiley. Anyway, he clearly loved the idea that a spherical shape could
imitate and evoke larger celestial bodies.

DARTH VADER
There is a fearsome sea creature that inflates itself into a sphere in order to
intimidate its enemies. That is why, in my journals, I call the Death Star “the
Pufferfish of the Skies.”

GRAND MOFF TARKIN

I pleaded with him to stop using that “Pufferfish of the Skies” line, but you know
how Lord Vader is. Once he latches on to an idea…
 4. WEST VIRGINIA, FLOATING IN SPACE

GRAND MOFF TARKIN

Ever seen footage of the Death Star? You get the sense it was teeming with
stormtroopers. Densely packed as a submarine.
Ha. In reality, the Death Star was the loneliest, emptiest place I’ve ever been.

IMPERIAL CENSUS TAKER

There were about 2.1 million people on the Death Star; that’s counting droids.
Meanwhile, with a radius of 70 kilometers, it had a surface area of almost 62,000
square kilometers. Now, assuming that you bring everybody to the surface level,
you’ll have a population density of about 30 people per square kilometer. That’s
five soccer fields per person.
To put that in perspective, it’s about the same size, population, and
population density as West Virginia. Want to picture social life on the Death
Star? Imagine West Virginia floating in space.
 STORMTROOPER
Gosh, it got lonely sometimes. You’d be patrolling a sector all day with nobody
in sight. Not even a droid.

IMPERIAL CENSUS TAKER

Of course, it was even worse than that. Not everybody spent their time on the
surface! The habitable zones of the station went 4 kilometers down. At 4 meters
per level, that’s 1000 levels.
So, picture your floating Space West Virginia blanketed with coal mines,
with people living up to 2.5 miles below the surface. On a given level, that’s a
population density of one person per 40 square kilometers. The only comparable
place on Earth? Greenland.

STORMTROOPER
And the ultimate insult? They slept us 60 guys to a room. Each dorm had three
toilets and two showers. Forget atomizing planets; if you want to give people
nightmares, just show them the lines outside those bathrooms every morning.
The Emperor can afford a multi-quadrillion-dollar space orb, and here I am
on board, sleeping in a triple-decker bunk bed and walking half a kilometer to
pee? Ugh. It still makes me mad.
 5. CURVES YOU CAN’T SEE

GRAND MOFF TARKIN

Fundamentally, the Death Star wasn’t about personnel. All the humans, all the
machines—reactor core, sublight engine, hyperdrive—they were merely a
support system for the station’s real purpose: the superlaser.

IMPERIAL GEOMETER
Ah, that’s another reason that a sphere was the perfect shape! For a given
volume, it’s got the lowest possible surface area. If you’re building a casing for
something huge—say, a planet-demolishing laser gun—then a sphere will get it
done with the minimum of material.
 IMPERIAL ENGINEER
I know the geometer will tell you that the spherical shape saved us money,
because a cube would have required 24 percent more steel. That’s a typical
mathematician for you: all theory, no practicality.
There’s a reason we don’t tend to build spherical ships: Curves are a pain!
Ever tried to arrange furniture in a curved room? Good luck positioning the sofa.
Now, the Death Star was too big for you to notice the curvature on that level.
No problem for the interior decorators. But on a construction level, that
curvature gave me a multiyear migraine. You needed the steel beams to bend by
about zero degrees, zero minutes, and 3 seconds of arc per meter—that’s less
than 1 degree of curvature per kilometer.
What a mess. The curvature was low enough that you couldn’t see it with the
naked eye, but high enough that every single part had to be custom-built.

GRAND MOFF TARKIN

The curved beams… Ugh, don’t remind me. I remember one subcontractor
shipped us straight beams for a whole year, thinking we wouldn’t notice. Which,
to be fair, we didn’t, until one day the emperor is approaching by shuttle and
says, “What’s that funny bulge?” That set us back MONTHS. At least the
subcontractor suffered worse than we did.
 DARTH VADER
Foolish subcontractors. If you’re going to stiff a customer, don’t make it an evil
empire.

6. MAYBE WE DID OUR JOBS TOO WELL

IMPERIAL GEOMETER

Like I said, a sphere minimizes surface area. But I’ll admit that it had other
drawbacks.

GRAND MOFF TARKIN

Waste removal was a perpetual torment.

IMPERIAL SANITATION WORKER

The fundamental solution to trash in space is simple: jettison it. However,
minimal surface area means the bulk of your station is far removed from the
surface. We’d have needed trash chutes dozens of kilometers long. I launched a
recycling campaign, but people just dumped everything—food, steel beams, live
rebel prisoners—into the trash compactors. You couldn’t help feeling that
sustainability wasn’t the priority.

IMPERIAL ENGINEER
The problem that still haunts me is the heating. It’s outer space, right? Cold. You
want to retain heat, and a sphere is great for that. Minimum surface area means
minimum heat loss. But apparently, we did our jobs TOO well, because early
simulations showed that the station would be prone to overheating.

IMPERIAL ARCHITECT
We had to dispose of some excess heat, so I put in thermal vents. Nothing big. A
few meters wide. Release the heat into space; problem solved.
I didn’t think…
I mean, when I heard that the rebels had destroyed the station by exploiting a
thermal vent…

ATTORNEY FOR THE IMPERIAL ARCHITECT

Let the record show: The investigatory panel found that the Death Star was
destroyed by a faulty reactor core, which my client did NOT design. My client’s
contribution was the thermal vents. These succeeded in their goal of transmitting
heat out of the station.

DARTH VADER
They also succeeded in transmitting a proton torpedo INTO the station.

GRAND MOFF TARKIN

From shaky start to bitter end, the Death Star was a child of compromise. Was it
expensive? Yes. Inefficient? No doubt. Destroyed by a ragtag band of rebel
scum? Hard to deny.
And yet… I couldn’t be prouder of that big, glorious sphere.
From a distance, it looked like a moon with an impact crater. Then you got
closer, and the geometric perfection started to overwhelm you: the myriad
symmetries of that crater, the right angles of the surface channels, the dark
circular scar of the equatorial band…

IMPERIAL GEOMETER
The Death Star achieved this eerie fusion: so large it must have been natural, but
so perfect it couldn’t be. It was disturbing, compelling, horrifying. That’s the
power of geometry.

DARTH VADER
As often happens, our critics supplied our finest slogan. When my old frenemy
Obi-Wan said, “That’s no moon,” we knew we had the tagline for the
advertisements.
 III
PROBABILITY

THE MATHEMATICS OF MAYBE

Ever flipped a coin? Unless you’re too poor to have touched one, or too rich to
have bothered, I’m going to guess “yes.” I’ll also guess that, even though the
odds are 50/50, your most recent flip did not show a 50/50 mix of outcomes. It
showed “heads” or “tails.” All or nothing.
That’s how life is: full of random one-off events. Unexpected train delays.
Come-from-behind victories. Magical out-of-nowhere parking spots. In our
storm-tossed world, anything can happen, and fate never RSVPs.
 But if you could flip a trillion coins, you’d find yourself approaching a
different world altogether: a well-groomed land of long-term averages. Here,
half of all coins land on heads, half of all newborns are boys, and one-in-a-
million events happen a millionth of the time, give or take. In this blue-skied
theoretical realm, quirks and coincidences don’t exist. They vanish against the
totality of all possibilities, like stones thrown into the sea.
 Probability bridges these two realms. The wild, woolly world we know is
what makes probability necessary. The calm, measured world we can never quite
reach is what makes probability possible. The probabilist is a dual citizen,
aiming to understand every breaking headline and celebrity fiasco as a single
card drawn from an infinite deck, a cup poured from a bottomless pitcher. As
mortal creatures, we’ll never set foot in the land of eternity—but probability can
offer us a glimpse.
 Chapter 11
THE 10 PEOPLE YOU MEET IN LINE FOR THE LOTTERY

Ah, the lottery ticket. It’s a certificate of optimism, a treasury bond from the
Department of Hope. Why cling to a drab, dog-eared $1 bill when you could
exchange it for a thrilling mystery bill worth anywhere from zero to $50 million?
If that sounds unappealing to you, well, you and humanity will just have to
agree to disagree.
I should confess that I’ve spent less on lottery tickets in my life ($7) than I
spent on croissants this month ($don’t ask). Nevertheless, each year, roughly
half of US adults play the lottery. It’s not the half you might guess. People
earning at least $90,000 are likelier to play than those earning below $36,000.
Those with BAs participate more than those without. The state with the highest
rate of lottery spending is my home state of Massachusetts: a haven of wealthy,
overeducated liberals, who spend $800 per person each year on the lottery.
Playing the lottery is like watching football, suing your neighbors, or butchering
the national anthem: an American pastime, pursued for a variety of reasons.
Come, join me in line for a ticket, and let’s investigate the multifaceted
appeal of turning your money into commodified chance.
 1. THE GAMER
Behold! It’s the Gamer, who buys lottery tickets for the same reason I buy
croissants: not for sustenance but for pleasure.
Take the Massachusetts game entitled $10,000 Bonus Cash. It’s a genius
name. Put “10,000” and “bonus” in front of any word and there’s no way you
can go wrong. Plus, the graphics on the $1 tickets for this game look like rave
night at the color printer. On the back, you’ll find the following complicated
odds of victory:
 What is this ticket worth? Well, we don’t know yet. Maybe $10,000; maybe
$5; maybe (by which I mean “very probably”) nothing.
It’d be nice to estimate its value with a single number. So imagine that we
spent not a mere $1, but $1 million. This lets us escape the rowdy dance floor of
the short-term world into the peace and quiet of the long-term one, where each
payout arrives in the expected ratio. In our million tickets, a one-in-a-million
event will occur roughly once. A 1-in-100,000 event will occur roughly 10
times. And a 1-in-4 event will occur 250,000 times, give or take.
Sifting through our stacks upon stacks of overstimulating paper, we’d expect
the results to look something like this:
 About 20% of our tickets are winners. Totaling up all of the prizes, our $1
million investment yields a return of roughly $700,000… which means that we
have poured $300,000 directly into the coffers of the Massachusetts state
government.
Put another way: on average, these $1 tickets are worth about $0.70 each.
Mathematicians call this the ticket’s expected value. I find that a funny name,
because you shouldn’t “expect” any given ticket to pay out $0.70, any more than
you would “expect” a family to have 1.8 children. I prefer the phrase long-run
average: It’s what you’d make per ticket if you kept playing this lottery over and
over and over and over and over…
Sure, it’s $0.30 less than the price you paid, but entertainment ain’t free, and
the Gamer is happy to oblige. Poll Americans about why they buy lottery tickets,
and half will say not “for the money” but “for the fun.” These are Gamers.
They’re why, when states introduce new lottery games, overall sales rise.
Gamers don’t see the new tickets as competing investment opportunities (which
would lead to a corresponding drop in sales for old tickets), but as fresh
amusements, like extra movies at the multiplex.

What’s the precise attraction for the Gamer? Is it the gratification of victory,
the adrenaline rush of uncertainty, the pleasant buzz of watching it all unfold?
Well, it depends on each Gamer’s appetites.
I’ll tell you what it isn’t: the financial benefit. In the long run, lottery tickets
will almost always cost more than they’re worth.
 2. THE EDUCATED FOOL
Wait… “Almost always”? Why “almost”? What state is dumb enough to sell
lottery tickets whose average payout is more than their cost?
These exceptions emerge because of a common rule by which big-jackpot
lotteries sweeten the deal: if no one wins the jackpot in a given week, then it
“rolls over” to the next week, resulting in an even bigger top prize. Repeat this
enough times, and the ticket’s expected value may rise above its price. For
example, in January 2016, the UK National Lottery draw offered a juicy
expected value of more than £4, compared to a ticket price of only £2. Weird as
they appear, such schemes typically spur more than enough sales to justify their
cost.
In the queue for lotteries like this, you’ll meet a very special player, a treat
for gambling ornithologists like us. See it there, high in the branches, grooming
itself? It’s the Educated Fool, a rare sap-brained creature who does with
“expected value” what the foolish always do with education: mistake partial
truth for total wisdom.
“Expected value” distills the multifaceted lottery ticket, with all its prizes and
probabilities, down to a one-number summary. That’s a powerful move. It’s also
simplistic.
Take these two tickets, each costing $1.

Spend $10 million on A, and you expect to earn back only $9 million,
amounting to a $0.10 loss per ticket. Meanwhile, $10 million spent on B should
yield $11 million in prizes, and thus a profit of $0.10 per ticket. So to those
enamored with expected value, the latter is a golden opportunity, while the
former is pyrite trickery.
And yet… would $11 million bring me any greater happiness than $9
million? Both values exceed my current bank account many times over. The
psychological difference is negligible. So why judge one a rip-off and the other a
sweet bargain?
Even simpler, imagine that Bill Gates offers you a wager: for $1, he’ll give
you a 1 in a billion chance at $10 billion. Calculating the expected value, you
start to salivate: $1 billion spent on tickets would yield an expected return of $10
billion. Irresistible!
Even so, Educated Fool, I beg you to resist. You can’t afford this game.
Scrape together an impressive $1 million, and the 99.9% likelihood is still that
rich man Gates walks away $1 million richer while you walk away broke.
Expected value is a long-run average, and with Gates’s offer, you’ll exhaust your
finances well before the “long run” ever arrives.
The same holds true in most lotteries. Perhaps the ultimate repudiation of
expected value is the abstract possibility of $1 tickets like this:

If you buy 10 tickets, you’re likely to win $1. That’s pretty terrible: only
$0.10 per ticket.
If you buy 100 tickets, you’re likely to win $20. (That’s 10 of the smallest
prize, and one of the next smallest.) Slightly less terrible: now $0.20 per ticket.
If you buy 1000 tickets, you’re likely to win $300. (That’s a hundred $1
prizes, ten $10 prizes, and a single $100 prize.) We’re up to $0.30 per ticket.
Keep going. The more tickets you buy, the better you can expect to do. If you
somehow buy a trillion tickets, the likeliest outcome is that you will win $1.20
per ticket. A quadrillion tickets? Even better: $1.50 per ticket. In fact, the more
tickets you buy, the greater your profit per ticket. If you could somehow invest
$1 googol dollars, you’d be rewarded with $10 googol in return. With enough
tickets, you can attain any average return you desire. The ticket’s expected value
is infinite.
But even if you trusted a government to pay out, you could never afford
enough tickets to glimpse that profit. Go ahead and spend your life savings on
these, Educated Fool. The overwhelming likelihood is bankruptcy.
We humans are short-run creatures. Better leave the long-run average to the
immortals.

3. THE LACKEY
Ooh, look who just stepped into the queue. It’s the Lackey!
Unlike most folks here, the Lackey’s windfall is guaranteed. That’s because
the Lackey isn’t keeping the tickets, just pulling down minimum wage to stand
in line, purchase them, and pass them off to someone else. Modest
compensation, but ultrareliable.
 Who would pay such a Lackey? Well, that’s a question for our next player…

4. THE BIG ROLLER

At first glance, this character looks an awful lot like the Educated Fool: the same
gleaming eye, the same scheming grin, the same obsessive focus on expected
value. But watch what happens when he encounters a lottery with positive
expected value. Whereas the Fool flails, buying a hapless handful of tickets and
rarely winning, the Big Roller masterminds a simple and nefarious plan. To
transcend risk, you can’t just buy a few tickets. You’ve got to buy them all.
How to become a Big Roller? Just follow these four steps, as elegant as they
are loony.
Step #1: Seek out lotteries with positive expected value. This isn’t as rare
as you’d think. Researchers estimate that 11% of lottery drawings fit the bill.
Step #2: Beware the possibility of multiple winners. Big jackpots draw
more players, increasing the chance that you’d have to split the top prize. That’s
an expected-value-diminishing disaster.
Step #3: Keep an eye on smaller prizes. In isolation, these consolation
prizes (e.g., for matching four out of six numbers) aren’t worth much, but their
reliability makes them a valuable hedge. If the jackpot gets split, small prizes
can keep the Big Roller from losing too much.
Finally, Step #4: When a lottery looks promising, buy every possible
combination.
Sounds easy? It’s not. The Big Roller needs extraordinary resources: millions
of dollars of capital, hundreds of hours to fill out purchase slips, dozens of
Lackeys to complete the purchases, and retail outlets willing to cater to massive
orders.
To understand the challenge, witness the dramatic tale of the 1992 Virginia
State Lottery.
That February’s drawing offered the perfect storm of circumstances.
Rollovers had driven the jackpot up to a record $27 million. With only 7 million
possible number combinations, that gave each $1 ticket an expected value of
nearly $4. Even better, the risk of a split prize was reassuringly small: only 6%
of previous Virginia lotteries had resulted in shared jackpots, and this one had
climbed so high that it would yield profit even if shared three ways.
And so the Big Roller pounced. An Australian syndicate of 2500 investors,
led by mathematician Stefan Mandel, made their play. They worked the phones,
placing enormous orders at the headquarters of grocery and convenience store
chains.
The clock labored against them. It takes time to print tickets. One grocery
chain had to refund the syndicate $600,000 for orders it failed to execute. By the
time of the drawing, the investors had acquired only 5 million of the 7 million
combinations, leaving a nearly 1 in 3 chance that they would miss out on the
jackpot altogether.
Luckily, they didn’t—although it took them a few weeks to surface the ticket
from among the stacks of 5 million. After a brief legal waltz with the state
lottery commissioner (who vowed “This will not happen again”), they claimed
their prize.
 Big Rollers like Mandel enjoyed profitable glory days in the 1980s and early
1990s. But those heady times have vanished. As tricky as the Virginia buyout
was, it was a logistical breeze compared to the daunting prospect of buying out
Mega Millions or Powerball (each with more than 250 million possible
combinations). Factor in post-Virginia rules designed to thwart bulk ticket
purchases, and it seems the Big Roller may never find the right conditions for a
return.

5. THE BEHAVIORAL ECONOMIST

For psychologists, economists, probabilists, and various other -ists of the
university, nothing is more intriguing than how people reckon with uncertainty.
How do they weigh danger against reward? Why do some risks appeal while
others repel? But in addressing these questions, researchers encounter a problem:
Life is super complicated. Ordering dessert, changing jobs, marrying that good-
looking person with the ring—to make these choices is to roll an enormous die
with an unknown number of irregular faces. It’s impossible to imagine all the
outcomes or to control for all the factors.
Lotteries, by contrast, are simple. Plain outcomes. Clear probabilities. A
social scientist’s dream. You see, the behavioral economist is not here to play,
just to watch you play.
The scholarly fondness for lotteries goes back centuries. Take the beginnings
of probability in the late 1600s. This age saw the birth of finance, with insurance
plans and investment opportunities beginning to spread. But the nascent
mathematics of uncertainty didn’t know what to make of those complex
instruments. Instead, amateur probabilists turned their eye to the lottery, whose
simplicity made it the perfect place to hone their theories.
More recently, the wonder duo of Daniel Kahneman and Amos Tversky
identified a powerful psychological pattern on display in lotteries. For that, I
introduce you to our next line-buddy…
 6. THE PERSON WITH NOTHING LEFT TO LOSE
In the spirit of behavioral economics, here’s a fun would-you-rather:

In the long run, it doesn’t matter. Choose B 100 times, and (on average)
you’ll get 90 windfalls, along with 10 disappointing zeroes. That’s a total of
$90,000 across 100 trials, for an average of $900. Thus, it shares the same
expected value as A.
Yet if you’re like most people, you’ve got a strong preference. You’ll take
the guarantee, rather than grab for a little extra and risk leaving empty-handed.
Such behavior is called risk-averse.
Now, try this one:
 It’s the mirror image of Question #1. In the long run, each yields an average
loss of $900. But this time, most people find the guarantee less appealing.
They’d rather accept a slight extra penalty in exchange for the chance to walk
away scot-free. Here, they are risk-seeking.
These choices are characteristic of prospect theory, a model of human
behavior. When it comes to gains, we are risk-averse, preferring to lock in a
guaranteed profit. But when it comes to losses, we are risk-seeking, willing to
roll the dice for the chance to avoid a bad outcome.
A crucial lesson of prospect theory is that framing matters. Whether you call
something a “loss” or a “gain” depends on your current reference point. Take
this contrast:
 The questions offer identical choices: (a) you walk away with $1500, or (b)
you flip a coin to determine whether you end with $1000 or $2000. But folks
don’t give identical responses. In the first case, they prefer the $1500
guaranteed; in the second, they favor the risk. That’s because the questions
create different reference points. When the $2000 is “already yours,” the thought
of losing it stresses you out. You’re willing to take risks to avoid that fate.
When life is hard, we roll the dice.
This line of research has a sad and illuminating implication for lottery ticket
purchasers. If you’re living under dire financial conditions—if every day feels
like a perpetual loss—then you’re more willing to risk money on a lottery ticket.
Think of how a basketball team, trailing late in the game, will begin fouling
its opponents. Or how a hockey team, down a goal with a minute to play,
replaces its goalie with an extra forward. Or how a political candidate, behind
with two weeks until the election, will go on the attack, hoping to shake up the
campaign. These ploys harm your expected value. You’ll probably lose by an
even larger margin than before. But by heightening the randomness, you boost
your chances at winning. In times of desperation, that’s all people want.
Researchers find that those who buy lottery tickets “for the money” are far
likelier to be poor. For them, the lottery isn’t an amusing way to exercise wealth,
but a risky path to acquire it. Yes, on average, you lose. But if you’re losing
already, that’s a price you may be willing to pay.
 7. THE KID WHO JUST TURNED 18
Hey, look at that fresh-faced young lottery player! Does the sight fill you with
nostalgia for your own youth?
Or does it make you think, “Hey, why are states in the business of selling a
product so addictive that they feel compelled to protect minors from buying it?”
Well, there’s someone else I’d like you to meet…

8. THE DUTIFUL TAXPAYER

Say hello to our next friend, a would-be civic hero: the Dutiful Taxpayer.
Nobody likes being hit up for money, even by the friendly neighborhood
government. States have tried taxing all sorts of things (earnings, retail, real
estate, gifts, inheritance, cigarettes), and none of it brings their citizens any
special joy. Then, in the 1970s and 1980s, these governments stumbled upon an
ancient mind trick.
Turn tax paying into a game, and people will line up down the street to play.
 The proposition “I’ll keep $0.30 from each dollar” prompts grumbles. Adjust
it to “I’ll keep $0.30 from each dollar and give the entire remainder to one
person chosen at random,” and you’ve launched a phenomenon. States don’t
offer lotteries out of benevolence; they do it to exploit one of the most effective
moneymaking schemes ever devised. State lotteries deliver a profit of $30 billion
per year, accounting for more than 3% of state budgets.
If you’re picking up a funny odor, I suspect it’s hypocrisy. It’s a bold move
to prohibit commercial gambling and then run your own gambling ring out of
convenience stores. Heck, the “Daily Numbers” game that states embraced
during the 1980s was a deliberate rip-off of a popular illegal game.
Is there an innocent explanation? To be fair, public lotteries are less
corruption-prone than private ones. Just ask post-USSR Russians, who found
themselves inundated with unregulated mob-run lotteries. Still, if we’re merely
aiming to shield gamblers from exploitation, why run aggressive advertising
campaigns? Why pay out so little in winnings? And why print gaudy tickets that
resemble nightclub fliers spliced with Monster Energy drinks? The explanation
is obvious: lotteries are for revenue, period.
To fend off attacks like the one I’m making now, states employ a clever ruse:
earmarking the funds for specific popular causes. “We’re running a lottery to
raise money” sounds better when you append “for college scholarships” or “for
state parks” than it does with “for us—you know, to spend!” This tradition goes
way back: Lotteries helped to fund churches in 15th-century Belgium,
universities like Harvard and Columbia, and even the Continental Army during
the American Revolution.
Earmarks soften the lottery’s image, helping to attract otherwise
nongambling citizens. These are your Dutiful Taxpayers: not necessarily
inclined toward games of chance, but willing to play “for a good cause.” Alas,
they’re getting hoodwinked. Tax dollars are fungible, and lottery revenue is
generally offset, dollar for dollar, by cuts in spending from other sources. As
comedian John Oliver quips: designating lottery money for a particular cause is
like promising “to piss in one corner of a swimming pool. It’s going all over the
place, no matter what you claim.”
So why do earmarks—and state lotteries, for that matter—endure? Because a
play-if-you-feel-like-it lottery sounds a lot better than a pay-or-you’ll-go-to-jail
tax.

9. THE DREAMER
Now, allow me to introduce you to a starry-eyed hopeful, who holds as special a
place in my heart as the lottery holds in theirs: the Dreamer.
 For the Dreamer, a lottery ticket isn’t a chance to win money. It’s a chance to
fantasize about winning money. With a lottery ticket in hand, your imagination
can go romping through a future of wealth and glory, champagne and caviar,
stadium skyboxes and funny-shaped two-seater cars. Never mind that winning a
lottery jackpot tends to make people less happy, a trend well documented by
psychologists. Daydreaming about that jackpot affords a blissful few minutes
while driving your disappointing regular-shaped car.
The prime rule of fantasy is that the top prize must be enough to change your
life, transporting you into the next socioeconomic stratum. That’s why instant
games, with their modest top prizes, draw in low-income players. If you’re
barely scraping together grocery money each week, then $10,000 carries the
promise of financial transformation. By contrast, comfortable middle-class
earners prefer games with multimillion-dollar jackpots—enough kindling for a
proper daydream.
If you’re seeking an investment opportunity, it’s crazy to fixate on the
possible payout while ignoring the probability. But if you’re seeking a license
for fantasy, then it makes perfect sense.
This Dreamer tendency helps explain the lottery’s odd economy of scale,
whereby big states fare better than small ones. Imagine a simplified lottery
game: Half of all revenue goes to the state, and the other half goes to the jackpot
winner. If a state sells $1 million in tickets, then each player has a 1-in-a-million
chance at a $500,000 jackpot. Meanwhile, if a state can sell $100 million in
tickets, then the odds drop to 1 in 100 million, and the jackpot leaps to $50
million.
 The expected value doesn’t change. In each state, an average ticket is worth
$0.50. But the psychology of the Dreamer will favor the megajackpot with the
miniscule odds.
 10. THE ENTHUSIAST FOR SCRATCHING THINGS
Ah, this citizen knows where it’s at. Forget the cash prizes and all that
probability mumbo-jumbo. The chance to rub a quarter across some cardboard is
winnings enough. The state lottery: it’s like a scratch-and-sniff for grown-ups.
 Chapter 12
CHILDREN OF THE COIN

If I’ve been stumped and impressed by one thing in my teaching career, it’s
human genetics. I’m not referring to that unfortunate semester in 2010 when I
was pressed into teaching 10th-grade biology. Rather, I refer to the job’s finest
long-term perk: getting to know families.

Every time you teach a pair of relatives—brother and sister, cousin and
cousin, aunt and equal-aged nephew—you renew your wonder at the scattershot
nature of biological inheritance. I’ve taught siblings who looked like twins, and
twins who looked like strangers. Parents’ nights always mess with my head.
Meeting after meeting, my mind performs a real-time face-mash of the two
adults before me, and I discover that their kid is a seamless Photoshop job: his
ears flanking her eyes, his hair on her head shape. All families are alike, and yet
no two families are alike in quite the same way.
The riddles of resemblance strike at the heart of biology. But as my students
can tell you, I am not a biologist. I have no DNA sequencer, no special
knowledge of introns or histones, and (more to the point) no clue. No, I’m a
mathematician. What I’ve got is a coin, a theorem, and a Brave Little Toaster–
esque faith in the power of figuring things out from first principles.
And maybe that’s enough.
This chapter is about two seemingly unrelated questions in the field of
probability. First is the question of genetic inheritance, which could fill grad-
school textbooks. Second is the question of coin flips, which feels so trivial it’s
barely worth stooping to pick up off the ground.
Can we bridge the two realms? Can the coin’s monosyllabic grunts give
expression to the complexity of the human race?
I want to say yes. And I’m pretty sure I got that from my dad.

Okay, let’s start with the easier of the two questions: What happens when you
flip a coin?
Answer: The two outcomes, heads and tails, occur with equal likelihood.
Solved!
Mmm, you feel that? It’s the pleasure of turning away from an urgent,
intractable real-world problem to focus instead on a puzzle that nobody cares
about. Savor the sensation. It’s why people become mathematicians.
Okay, so nothing much happens when we toss a single coin. But what if we
toss a pair? You’ll find that four outcomes are equally likely:
 A stickler for specifics will see the two middle outcomes as different: head
then tail vs. tail then head. (Imagine that one coin is a penny and the other is a
nickel.) But if you’re like most coin flippers, you won’t care. Your mind
collapses both into a single outcome (“one head, one tail”) that happens to have
double the probability.
Keep going: What if we toss three coins?

Eight outcomes in all. If you’re hoping for all heads, there’s only one path:
every coin must perform its duty, so the probability is 1 in 8. But if you’re
aiming for two heads and a tail, then you’ve got three ways to do it—the tail can
appear on the first, second, or third coin—giving a probability of 3 in 8.
 Now, what about four coins—or five, or seven, or 90?

What we’re exploring here is a family of probabilistic models known as the

binomial distribution. You take a single binary event (heads or tails, win or lose,
one or zero) and repeat it several times. How many of each outcome will occur?
That’s the binomial. And, pursuing a branch of this mathematical family, we
witness two clear trends.
First, as we add coins, the complexity blooms. An extra coin doubles the
number of possible results, from two to four to eight to 16. Yet the orchestra is
only tuning up. Make it 10 coins, and we’ll exceed 1000 possibilities. Reach 50
coins, and we’ll face 250 possible outcomes: more than a quadrillion. This
aggressive growth pattern is known as the combinatorial explosion. A linear
increase in the number of objects leads to an exponential increase in the number
of ways to combine them. A handful of coins, simple as it seems, gives rise to
unfathomable variety.
The second trend points the other direction: Even as the variety explodes, the
extremes vanish. Check out the four-coin case above, which has only two
“extreme” outcomes (all heads or all tails) against 14 “middle” outcomes (with
some of each). The more coins we toss, the less likely the extremes become and
the more we find ourselves mired in the swampy middle ground of “a pretty
even mix of heads and tails.”
 The reason is simple. Ordinary outcomes are ordinary precisely because there
are so many ways to achieve them. Meanwhile, extreme outcomes are extreme
precisely because they are so singular, unfolding in very few ways.
Suppose we flip 46 coins. There is only one way to achieve all heads.
But allow yourself just one tail among 45 heads, and the possibilities
multiply: the tail could come first, or second, or third, or fourth, or fifth… all the
way up to the tail coming last. That’s 46 combinations.
Make it two tails among 44 heads, and it’s even easier to achieve. The tails
could come first and second, or first and third, or first and fourth… or first and
last… or second and third, or second and fourth, or second and fifth… all the
way up to last and second to last. That’s more than 1000 possibilities in total,
and we’re just getting warm. The likeliest option—23 tails and 23 heads—can
occur in 8 trillion distinct ways.
I’m no psychic, but if you go flip 46 coins, I predict that the following will
happen:

1. You will get between 16 and 30 heads.

2. You will flip a specific sequence that no 46-coin flipper in history has
ever flipped: a one-time historical anomaly, unique in human experience.
 ever flipped: a one-time historical anomaly, unique in human experience.

Somehow, though, the first point (your outcome is middling) will

overshadow the second (your outcome is glorious and unique). To us, coins are
interchangeable, so any roughly balanced sequence will strike us as generic and
forgettable. A snowflake in a blizzard.
But imagine if we cared about the specific constitution of that snowflake.
Imagine if each heads and tails struck us as a potent twist of fate. What if we
worried about not just how many heads came up but which ones precisely?

Then, whichever of the 70 trillion possible combinations emerged would

strike us as so vanishingly unlikely, so egregiously improbable, that its
appearance would seem a miracle. It would enchant us like a star plucked from
the sky and lowered into our hands. Those 46 coins would be as precious as…
well, as a newborn baby.
And that brings us to the chapter’s hard question: genetics.
Every cell in your body contains 23 pairs of chromosomes. Think of them as
a 23-volume recipe book, the instructions for building you. You have two
versions of each volume: one from your mother, and one from your father.
 Of course, your mother and father each have two copies themselves: one
from your grandmother, and one from your grandfather. How did they decide
which one to pass on to you, their precious little pea plant? Well—and here I
simplify rather dramatically—they flipped a coin. Heads: you get Grandpa’s
chromosome. Tails: you get Grandma’s.
Your mother and father repeated this process 23 times each, once per volume,
and the result was… well, you.
 By this model, each couple can generate 246 distinct offspring. That’s 70
trillion. And unlike with coins, the specifics matter. I have my mother’s thick
hair and pleasure in reading, as well as my father’s gait and his love of clarity. If
I had a different mix of their traits—my father’s curls, say, and my mother’s
height (or lack thereof)—then I’d be a different person, a sibling to myself.
In genetic inheritance, a head and a tail is not the same as a tail and a head.
This model predicts the varying degrees of resemblance we see between
siblings. At one extreme, every coin flip for the younger sibling could come up
exactly as it did for the older one. Such siblings, despite being separated by
years, would be, in essence, identical twins.
At the other extreme, it is possible for none of the coin flips to come up the
same. Such siblings, as in a creepy Philip K. Dick story, would be genetic
strangers, sharing no closer a biological relation than did the parents who birthed
them.
 Of course, both extremes are wildly improbable. The far greater likelihood is
that two siblings will share roughly half of their 46 chromosomes in common—
perhaps a little more, perhaps a little less. According to our binomial model, the
vast majority will share between 18 and 28.

In fact, data on nearly 800 sibling pairs, compiled by Blaine Bettinger at his blog
The Genetic Genealogist, matches our crude prediction pretty darn well:
 There’s a caveat. I’ve omitted a factor so important that I can’t blame
biologists if they’re rage-tearing the pages of this book right now. That’s
crossing over, or recombination.
I’ve claimed that chromosomes are passed down undisturbed and intact. Like
many things I said while teaching biology, that’s false. Before a single version of
each chromosome is selected, the two experience a splicing-together—for
example, swapping the middle third of each. Thus, a given chromosome may
come mostly from Grandpa, but it’ll feature a handful of recipes exchanged for
the equivalents from Grandma.
 Crossing over happens roughly twice per chromosome. Thus, to enhance the
accuracy of our model, we can triple the number of coin flips (since two
crossovers will split a chromosome into three pieces).
How does this impact the number of possible offspring? Remember, linear
growth in the number of objects yields exponential growth in their combinations.
So the variety of offspring does far more than triple. In fact, it grows from 246
(or 70 trillion) to 2138 (which is a staggering 350 duodecillion).
I’m arguing, in short, that a newborn baby has a lot in common with a
handful of scattered pennies. That’s not to say you should revalue your infant’s
life at $0.46. To the contrary, you should admire $0.46 of spilled pocket change
as a miracle on par with human birth.

Getting to know families, you feel that they mix like liquids, blue and yellow
paint sloshing together to create green. But they don’t. Families mix like decks
of cards, with elements shuffled and reshuffled, tucked up one’s sleeve and then
brought back into play. Genetics is a game of combinations: the lovely and
elusive patterns of resemblance between siblings can all be traced back to the
combinatorial explosion. Flip enough coins, and the stark, distinct outcomes
(heads vs. tails) begin to blur and blend. The jagged edges of the graph fuse to
form something as liquid and flowing as humanity itself. In this way, we are the
offspring of combinatorics, the progeny of a shuffled deck, the children of the
coin.
 Chapter 13
WHAT DOES PROBABILITY MEAN IN YOUR
PROFESSION?

To call human beings “bad” at probability would be simplistic and mean-

spirited. Probability is a subtle branch of modern mathematics, booby-trapped
with paradoxes. Even an elementary problem can throw a stone-cold expert for a
loop. Dissing folks for their probabilistic failures is a bit like calling them bad at
flying, or subpar at swallowing oceans, or insufficiently fireproof.
No, it would be fairer to say that human beings are terrible at probability.
In their psychological research, Daniel Kahneman and Amos Tversky found
that people harbor persistent misconceptions about uncertain events. Time and
again, they overrate negligible possibilities, and underrate near certainties.

No big deal, right? I mean, does probability ever come up in the real world?
It’s not like we spend our lives clawing for intellectual tools that might offer the
slightest stability in the swirling miasma of uncertainty that surrounds us every
waking moment…
Well, just in case—this chapter is a handy guide on how various sorts of
human beings think about uncertainty. Just because it’s hard doesn’t mean we
can’t have some fun with it.

Hello! You are a person. You have two eyes, a nose, and sentience. You
dream, laugh, and urinate, not necessarily in that order.
Also, you live in a world where nothing is certain.
Take a simple fact: How many planets does your solar system have? These
days, you say “eight.” From 1930 until early 2006, you answered “nine”
(counting Pluto). In the 1850s, you wrote textbooks listing twelve: Mercury,
Venus, Earth, Mars, Jupiter, Saturn, Uranus, Neptune, Ceres, Pallas, Juno, and
Vesta. (You now call the last four “asteroids.”) In 360 BCE, you named seven:
Mercury, Venus, Mars, Jupiter, Saturn, the moon, and the sun. (You now call the
last two “the moon” and “the sun.”)
You keep changing your mind as new data and new theories emerge. That’s
classic you: When it comes to knowledge, you’ve got lots of good ideas but no
guarantees. Teachers, scientists, politicians, even your sense organs—any of
these could deceive you, and you know it.
Probability is the language of your uncertainty. It lets you quantify what you
know, what you doubt, what you doubt you know, and what you know you
doubt, expressing these nuances of confidence in a clear and quantitative
language. At least, that’s the idea…
 Hello! You are a political journalist. You write about elections coming up.
You write about elections going down. On rare special days, you even write
about things like “policy” and “governance.”
Also, you seem baffled by the occurrence of slightly unlikely events.
This wasn’t always true. Once upon a time, you saw elections as magical
moments of infinite possibility. You downplayed the likeliest outcome so as to
build suspense, to make it seem as if each race was won on a half-court buzzer-
beater. On election night 2004, when George W. Bush led Ohio by 100,000
votes with barely 100,000 left to count, you said the state was “too close to call.”
When probabilistic models gave Barack Obama a 90% chance to win in 2012,
you dubbed the race “a toss-up.”
Then, 2016 turned your world to pineapple upside-down cake. Donald Trump
defeated Hillary Clinton. Waking up the next day, you felt you’d experienced a
quantum singularity, no more foreseeable than a rabbit materializing out of thin
air. But to probabilist Nate Silver and like-minded others, this was only a
moderate surprise, a 1-in-3 event—like rolling a die that lands on 5 or 6.
 Hello! You are an investment banker. You bank investments. You invest in
banks. And your suits are worth more than my car.
As recently as the 1970s, your job was pretty boring. You were a funnel for
funds, pouring money into “stocks” (i.e., pieces of companies) or “bonds” (i.e.,
pieces of debt). Stocks were exciting; bonds were bland. You were a human
bond.
Now, your job is as thrilling as a roller coaster that has failed its safety
inspection. Starting in the 1970s and 1980s, you began inventing complex
financial instruments that no one fully understands, least of all the government
regulators. Sometimes, these devices earn you big bonuses. Sometimes, they
make your century-old firm collapse, leaving a vast crater across the economy as
if a dinosaur-killing asteroid has struck. Exciting times!
To be fair, allocating capital is a pretty important job in capitalism, and has
the potential to create lots of value. It annoys you when snarky math teachers
slight your profession. If so, take a moment to calculate the factor by which your
salary exceeds that of the snarky math teachers, and I think you’ll find your
spirits rising.
 Hello! You are a local news anchor. You have great hair, crisp enunciation,
and a long history of artificial banter with your coanchor.
Also, you are obsessed with very unlikely events.
You like to report on chilling dangers. Local murders. Airborne carcinogens.
Defective toys that attach themselves like the creature in Alien to your baby’s
chubby cheeks. Ostensibly, you do this to inform your audience. But be honest.
You do it to hold their attention. If you hoped to raise awareness about the
statistical dangers to children, you’d warn about household accidents caused by
guns and swimming pools. Instead, you paint vivid landscapes in which the only
danger more pressing than abduction is crocodile attacks. You know that we
can’t tear our eyes away from such horrors—especially when they’re leavened
with artificial banter.
 Hello! You are a weather forecaster. A televised cloud prophet. You gesture
with conviction, and end every conversation with “And now, back to you.”
Also, you fudge probabilities so people don’t get mad at you.
Of course, you try to be honest. When you say that there’s an 80% chance of
rain, you’re exactly right: It rains on 80% of such days. But when rain is less
likely, you inflate the numbers. You fear angry tweets from people who leave
their umbrellas at home and then blame you for the sky’s damp doings. So when
you forecast a 20% chance of rain, it rains only 10% of the time. You’re
increasing the odds so as to decrease your hate mail.
Perhaps, if people understood probability better, you could speak your truth.
People seem to act as though “10%” means “not going to happen.” If only they
embraced its true meaning (“happening one time in 10”) then you could loosen
your tongue and divulge the numbers of your inner heart. Until then, you remain
a merchant of half-truths.
And now, back to you.
 Hello! You are a philosopher. You read strange books, write even stranger
ones, and enjoy walking into a bar alongside a priest and a rabbi.
Also, you do not let probability intimidate you.
Leave “probable” to the empiricists, you say. You seek ideas that are
improbable. You ask questions no one else will ask and think thoughts no one
else will think—mostly because they are pedantic, technical, and probably false.
But this is to your credit! At your best, you birth new fields. Psychology, from
Aristotle to William James, has its ancestry in philosophy. Even at your worst,
your inquiries provoke and refresh us as no one else can.
 Hello! You are a Mission: Impossible agent. You dangle from ceilings into
locked vaults, cling via suction cup to skyscrapers, and unmask your true
identity to the double-crossers you’ve just triple-crossed.
Also, you do not seem to get what “impossible” means.
“Impossible” does not mean “as routine as a quarterly earnings report.” Nor
does it mean “very rare” or “rather difficult” or “whew, it’s a good thing that
knife blade halted a millimeter from Tom Cruise’s eye.” It means “not possible.”
And yet it keeps happening. Your film’s title is as dishonest as its theme song is
catchy.
In this, you are not alone. All fiction bends plausibility. One of my favorite
shows is Friday Night Lights, a humble exploration of life in a small Texas
town, depicting the human-scale struggles of ordinary people. Yet even this
paragon of plausibility concludes each football game with once-in-a-lifetime
dramatics: a 90-yard touchdown pass, a goal-line fumble, or a would-be field
goal bouncing off the crossbar. It’s enough to make me wonder: are we
projecting our improbable fantasies onto the TV screen, or is the TV screen
planting those fantasies in us to begin with?
 Hello! You are the captain of the Millennium Falcon. You are a ruffian, a
rascal, and a gold-hearted rogue. Your life partner is an 8-foot space dog
wearing a bandolier.
Also, you reject probability altogether.
You’re not a man of sober reflection and strategic consideration. You’re a
contraband smuggler, an empire toppler. You’re a quick-draw Greedo-killing
daredevil for whom doubt is fatal and hesitation is worse. There are no
probabilists in foxholes, and you have spent your life in foxholes. For you,
laborious probabilistic calculations would be as cumbersome as a neurotic
golden robot who keeps saying, “Oh my!” and “If I may say, sir…”
I like to think that there’s a bit of you in each of us. Probability serves us well
in times of cool and careful assessment, but sometimes we need a confidence
that’s unwarranted by the hard quantitative facts. In moments of instinct and
action, the soul leashed to plausibility may shrink from a necessary leap. At such
times, we need to forget the numbers and just fly.
 Chapter 14
WEIRD INSURANCE

COVERAGE SCHEMES FROM THE DEEPLY STRANGE TO

THE STRANGELY DEEP

Despite my best efforts and worst drawings, I have become a grown-up. I drink
unsweetened coffee. I wear unironic ties. Children call me “mister.” Most
discouraging of all, I spend a nonzero percentage of life thinking about
insurance: health, dental, home, auto. Before long, I’ll be insuring my ties
against coffee stains, all without a shred of irony.
Beneath the suffocating weight of adulthood, there’s something funny going
on here. Insurance companies don’t really sell products. They’re vendors of
psychology, offering peace of mind. They let short-run folks like me purchase a
bit of stability from the long run of eternity.
So, for this chapter, let us spit our coffee back in the face of adulthood and
reach for the chocolate milk instead. Let us gaze upon insurance with fresh eyes
and explore the mathematics of this quirky industry, not through its traditional
products, but through a series of peculiar and unconventional plans. Let us
investigate the promise of mutual gain, the danger of exploitation, and the slow
data-driven tug-of-war between the insurer and the insured—all while talking
about college football and aliens.
Even in the darkest depths of grown-up-ness, it’s worth remembering that
insurance is a strange and silly thing. Kind of like adulthood itself.

SHIPWRECK INSURANCE
Ever since there have been ships, there have been shipwrecks. In China five
millennia ago, merchants traded goods downriver in boats. If all went per usual,
then the cargo would arrive, no problem. But if rapids and rocks intervened,
you’d find your shipment rerouted to the riverbed, which can rarely be trusted to
pay up.
Fortune dealt in two cards: moderate success and utter disaster. The former
gave you reason to play the game, but the latter could ruin you.
What was an ancient Chinese trader to do?
Simple: redistribute the danger. Instead of loading one’s own boat with one’s
own goods, merchants would scatter their wares across each other’s ships so that
each vessel carried a mixed cargo. By coming together, they turned the small
likelihood of a high cost into the high likelihood of a small one.
 This kind of risk redistribution predates history. You see it in any tight-knit
community where people sacrifice to help their neighbors, knowing they might
one day draw on the same communal kindness. A minor guaranteed cost helps
eliminate an existential risk.
Today, “tight-knit community” has given way to “for-profit insurance
company.” Old-school social mechanisms of enforcement (e.g., not helping the
stingy people) have yielded to precise formula-driven ones (e.g., setting rates
based on risk factors). History has distilled a social arrangement into a
mathematical one.
 Still, in principle, insurance today works much as it did 5000 years ago. I find
fellow souls facing the same mortal risks. We exchange pieces of our livelihood.
And we float downriver knowing that now, no shipwreck can sink us.

INSURANCE TO MAKE SURE THE KING HAS

YOUR BACK
Quick time-travel jaunt. Set your watch to “Iran, 400 BCE; the New Year
holiday of Norouz.”
We arrive at the palace, where the monarch, flanked by bureaucrats and
notaries, greets a long line of visitors. Each presents a gift, which circulates
swiftly among the king’s accountants so it can be assessed and documented. We
are witnessing that most joyous of celebrations: insurance enrollment.
If you donate an item worth more than 10,000 darics, your name will land on
a special honorary ledger. In your time of need, you shall receive 20,000 darics.
Smaller gifts don’t merit that guarantee but may still receive generous payouts.
Take the humble fellow who donated a single apple and later received in return
an apple full of gold coins.
Deluxe fruity piñatas aside, the scene invites a more cynical way to
conceptualize insurance. It’s the rich selling a bit of resilience to the poor.
When bad things happen to rich people, they cope. If you’re a king, a
billionaire, or a grandchild of Sam Walton, then you pay the hospital bills, buy a
new car, and move on with life. But when bad things happen to poor people,
they struggle. They might forgo medical treatment, take on debt, or lose a job
because they can’t afford transportation to work. The rich rebound after
misfortune; the poor spiral.
Thus, in a step of topsy-turvy logic, insurance is born. The poor, unable to
afford risk, pay the rich to take it off their hands.

“OH NO, MY EMPLOYEES ALL WON THE

LOTTERY” INSURANCE

It’s a common scene: workers pooling their money to buy lottery tickets, hoping
to quit their jobs en masse. They are comrades unified by the dream of no longer
being unified. But suppose they win. What happens to their poor boss, awaking
one morning to the abrupt resignations of an entire department, like some kind of
unnatural disaster?
Fear not, forsaken boss. There’s a cure for what ails you.
In the 1990s, over a thousand British businesses invested in the National
Lottery Contingency Scheme. If your staff won the jackpot, insurance payouts
would help to cover recruitment, retraining, revenue losses, and temporary labor.
(I reckon the government, which runs these lotteries, should sell the
insurance plans, too. It’d be like Godzilla Insurance Co. offering fire-breath
protection.)
As you’d guess, the expected value favors the insurer. What might surprise
you is the margin. The odds of a single ticket winning the jackpot are 14 million
to 1, and yet the payout ratios max out at 1000 to 1. Smaller businesses fare even
worse. With two employees, the expected value remains unfavorable even if
each staffer buys 10,000 tickets.

Of course, insurers can’t just charge the expected value. They’ve got to
consider their own costs, not to mention the risk involved. An insurer that sets its
margins too narrow will go bankrupt the first time a slightly unusual number of
clients demand payouts.
As for the fearful bosses, is there a better option? Well, if you can’t insure
’em, join ’em. Tossing a little money into the pool is an insurance plan in itself:
a small investment that will pay out if (and only if) they win.

MULTIPLE-BIRTH INSURANCE
I view “one” as a formidable number of babies to care for. “Two” sounds like
quartering a child army. “Three” is unimaginable—a childbirth googol. So it’s
dizzying to learn that 1 in 67 pregnancies results in multiple births. In the UK,
expectant parents who share my trepidation can purchase “multiple-birth
insurance.”
You pop out twins, and the insurance company pays out £5000.
One analyst and personal finance adviser, David Kuo, scoffs at this plan,
saying: “You’re better off putting money on the horses.” I see his point: The
expected value here is pretty lousy. The lowest premium is £210, for a mother
with an age below 25 and no family track record of twins. If a hundred such
mothers buy the insurance, then the company will collect £21,000, and (most
likely) pay out just one sum of £5,000. That’s a comfy profit margin, which is
the polite way to say “their customers are paying through the nose.”
Parents with a greater likelihood of twins fare little better. One 34-year-old
mother who was a twin herself paid £700 for coverage. That’s a payout ratio of
just 7 to 1.
If you’ve had fertility treatments, or waited until after the 11-week
ultrasound, then you’re not eligible at all.

I find it helpful to distinguish between two types of insurance: financial and

psychological.
Most common insurance products—health, life, auto, etc.—hedge against
financial ruin. But psychological insurance hedges against more modest risks.
Take travel insurance. You can probably afford the plane tickets (heck, you’ve
already paid for them), but no one feels good canceling a vacation. By offsetting
the financial cost, insurance can ease the emotional one.
This isn’t a self-help finance book (Math with Bad Investment Advice). Still, I
shall offer one piece of counsel: Be wary of psychological insurance.
Imagine if the multiple-birth insurer offered a raft of other plans. Would you
pay £200 to insure against postpartum depression? Or the higher cost of raising
an autistic child? Or a child born with Down syndrome? Or a kid with a chronic
illness? Each risk feels worth protecting against. But buy enough plans at £200
each, and eventually, the cumulative cost of the premiums will swamp any
payout. If you can afford the insurance, then you can afford the outcome it’s
insuring against.
A more frivolous example: For $3 each, I’ll offer to insure any T-shirt worth
$20 or less. Eager to protect your wardrobe, you insure 15 beloved favorites.
Then, when your dogs and/or twins chew one up, I’ll ship a replacement.
However, if this happens once per year, you’re paying $45 for a service worth no
more than $20. Why not just keep the money and replace your own T-shirts?

It’s the basic problem of all psychological insurance. Why pay good money to
offload a risk that you could handle yourself?

ALIEN-ABDUCTION INSURANCE
In 1987, a fellow in Florida named Mike St. Lawrence rolled out a $19.95 alien-
abduction policy. It provided $10 million in coverage, paid out at a rate of $1 per
year (until completion or death, whichever comes first). It also entitled abductees
to “psychiatric care” and “sarcasm coverage” (“limited to immediate family
members”). The slogan: “Beam me up—I’m covered.” Applicants were rejected
if they answered yes to the question “Do you take this seriously?” or “Were your
parents related before they married?” It was hard to miss the winking.
Then a British company began selling abduction coverage in earnest. Soon,
they had sold 37,000 plans at £100 apiece—and (surprise!) not made a single
payout. That’s nearly £4 million in pure profit. Said one managing partner: “I’ve
never been afraid of parting the feeble-minded from their cash.”
The silly story points toward a real danger. Psychological insurance can
quickly turn predatory. When we buy protection against fear itself, we invite
companies to make us afraid.

FAILED-EXAM INSURANCE
During your mathematics PhD at UC Berkeley, you take a high-stakes qualifying
exam known as “the Qual.” This is a monster you face alone: three hours of oral
questioning from a trio of expert professors. Though preparation is exhausting,
the passing rate is high—at least 90%—because it isn’t scheduled until you seem
battle-ready. The expert professors desire a retake no more than you do.
Qual failure, then, fits the following criteria: (1) it’s rare; (2) it hurts; (3)
many individuals face it; and (4) it strikes more or less at random. That would
seem to make it ripe for insurance.
 Some PhD students have hashed out the idea. Each participant pays $5 into
the pool. If anyone fails their exam, then they collect the whole bundle. If
everyone passes, the money goes toward a round of celebratory beers.
I suppose every financial instrument designed by grad students must
culminate, one way or another, in beer. It’s psychological insurance of the best
kind.

HOLE-IN-ONE PRIZE INSURANCE

There’s nothing more American than a gaudy competition. The $10,000 half-
court shot. The $30,000 dice roll. The $1 million hole in one. When rolling out
such promotional gimmicks, there’s just a teeny-tiny risk.
Someone might win.
As it turns out, this market is a dream for insurance companies. As merchants
of probability, their solvency depends on accurate computation: 50-to-1 payouts
on a 100-to-1 event will keep you in the black; 100-to-1 payouts on a 50-to-1
event will bankrupt you. The actuarial intricacies of home, life, and health
insurance make it easy for an insurer to miscalculate.
But prize giveaways? No problem!
Take the hole in one. Amateur golfers succeed roughly once in every 12,500
attempts. So, for a $10,000 prize, the average golfer costs $0.80. Charge $2.81
per golfer (one company’s advertised premium), and you’ll turn a tidy profit.
The same company insured a $1 million hole-in-one prize for just $300. It’s a
fine deal on both ends: the charity offloads risk, and the expected-value cost to
the insurer is just $80.
The company also insured an NBA fan’s $16,000 half-court shot for $800.
Unless the fan’s odds of success exceed 1 in 50—about what NBA players
manage in game situations—it’s a square deal. Or consider insuring Harley-
Davidson against a $30,000 prize given out to a customer who rolls six lettered
dice to spell out H-A-R-L-E-Y. The expected value is just $0.64 per contestant.
The company charged $1.50. Actuarial math has never been easier.

That’s not to call this type of insurance riskless. In 2007, the Boston-area
retailer Jordan’s Furniture ran a glitzy promotion: buy any sofa, table, bed, or
mattress in April or May, and if the Red Sox go on to win the World Series in
October, then the price will be refunded. The sales approached 30,000 items—a
total of perhaps $20 million.
Then the Red Sox won. The Sox-loving folks at Jordan’s were thrilled;
they’d bought insurance.
The folks at the insurance company? Less delighted.
 “CHANGE OF HEART” WEDDING INSURANCE
Please join me now for a thought experiment. A stranger hands you a coin. You
flip it and hide the result. What’s the probability of heads?

• A random passerby: “I don’t know… 50%?”

• The deviant who gave you the coin and knows that it’s biased toward
heads: “70%!”
• You, having snuck a look at the outcome: “100%.”

No one here is wrong. Uncertainty is a measure of knowledge, and each

person has given a rational, well-calibrated response to the available
information. The fact is, every probability ought to come with a footnote:
“contingent on what I know now.”
This dynamic can become a nightmare for insurance companies. What if
they’re the random passerby, whereas the person buying the insurance has snuck
a peek at the coin?
The company Wedsure insures against all kinds of wedding fiascoes:
damaged bridal gowns, stolen gifts, no-show photographers, mass food
poisoning. But the most sensational coverage option they offer is also the most
problematic: the “Change of Heart” clause.

In 1998, a journalist asked Wedsure owner Rob Nuccio if the company

would consider reimbursing for weddings canceled because of cold feet. He
scoffed. “That would too easily set us up for fraud and collusion,” he said. “If
you’re asking the question, don’t get married.” By 2007, he had changed his
mind and begun offering to reimburse third-party financiers (but not the couples
themselves) for cancellations with at least 120 days’ notice. But according to
Nuccio, a problem arose: “We were getting claims from mothers of the bride
who… knew in advance that there was a problem.” Obviously, each woman had
mapped her daughter’s heart better than the insurance company could. They had
peeked at the coin. Nuccio widened the time window to 180 days, then 270, and
eventually 365.
It’s a quintessential problem for insurers. The person seeking coverage often
knows details that the provider cannot. The insurer has five possible solutions:

1. Set your profit margin even higher than normal.

2. Employ a detailed application process to equalize the knowledge level.
3. Offer a cheap low-coverage option, and a pricey high-coverage option.
The high-risk people will naturally be drawn to the latter.
4. Become so expert in the risk you’re insuring that you know more than
your customers.
5. Stop insuring that risk.

INSURANCE-COMPANY INSURANCE

Knowledge gaps scare insurers, but they don’t jolt them awake in a cold sweat.
Their true wide-eyed nightmare is a single whispered word: dependence.
Think back to the Chinese traders, who scattered their merchandise across a
hundred boats. What if shipwrecks don’t come one by one? What if on 99% of
days, no boats sink, and on 1% of days, a storm sinks them all? Then insurance
is useless. We’re not facing individual risks that can be mitigated by
redistribution. We’re all holding the same awful ticket to the same terrible
lottery. Trading pieces of fate does nothing. When we go down, we go down
together.
That’s dependence. For insurers, it’s doom. Hurricane Katrina, for example,
exacted its $41 billion in damage all at once, far exceeding the year’s $2 billion
or so in premiums. In such cases, the insurance company faces the same risk as
the people it’s insuring. The eggs remain all in one basket.
The solution is to scale up: insure the insurance company. This is called
“reinsurance.” Insurers from different regions trade assets, in effect swapping
customers. This lets a company in one location assemble a risk portfolio that’s
spread far and wide.
When it’s not enough to scatter your livelihood across a hundred boats, then
scatter it across a hundred rivers.

INSURANCE FOR COLLEGE FOOTBALL PLAYERS

IN CASE THEY GET WORSE
Time for some fantasy role-play: you are a college football star. Grown men
paint their bare chests in your team’s colors. You’ve got skills worth millions.
But you can’t collect on the payday while you’re in college—and devastating
injury can strike at any time. It’s like your body is a jackpot lottery ticket you’re
forbidden to cash in until you’ve run it through the washing machine a few
times. Better hope those winning numbers don’t fade.
What’s your move, kid?
You can insure against career-ending injury, but that’s not enough. What if
an injury hampers your career without ending it? What if it drops you from the
first round of the draft (for a contract worth $7 million) to the sixth round (for a
contract worth $1 million)? More than 80% of your payday has gone up in
smoke, and that pricey insurance plan (costing $10,000 per $1 million of
coverage) does nothing.
This is why top-tier players have begun purchasing an additional plan: “loss
of value” insurance. It’s not cheap; it costs an additional $4000 in premiums per
$1 million of coverage. But for high-value prospects, it’s worth it. Players like
cornerback Ifo Ekpre-Olomu and tight end Jake Butt have already collected.
More, no doubt, will follow.

HEALTH INSURANCE

I’ve saved for last the weirdest risk-reducing product of all: health insurance.
What’s so strange about it? First, the complexity. Deductibles, coverage
limits, preexisting conditions, co-pays, variable premiums… It’s not quite brain
surgery, but it is the financing of brain surgery, which is perhaps just as
daunting. And second, there’s the vexing question of whether, as our medical
knowledge and predictive powers grow, health insurance can continue to
function at all.
Here’s a simple model of healthcare. We each flip 10 coins. If you flip all
heads, then you contract the dreaded “ten-headed disease,” which will require
$500,000 in medical bills.
Each of us faces a roughly 1 in 1000 risk of disaster. But if we all chip in
$800 for insurance, then it works out. Per thousand people, the insurance
company collects in $800,000 and pays out just $500,000. They turn a profit; we
hedge against financial ruin. Everybody goes home happy.

But what if we know more? What if, before deciding whether to purchase the
plan, we can peek at the first five coins?
Now, from every 1000 people, roughly 970 of us glimpse a tail and sigh with
relief. We’re safe; no insurance necessary. But the remaining 30 people grow
nervous. One of them probably has the dreaded disease. Between them, those 30
can expect a collective $500,000 in expenses. Even if they redistribute the costs
equally, each remains on the hook for thousands and thousands of dollars.
Insurance no longer provides peace of mind; it drains it.
 When we looked at the coins, we reduced our uncertainty, and without
uncertainty, insurance collapses. If we know in advance who will suffer—whose
boat will sink, whose employees will win the lottery, who will suffer that NFL-
dream-crushing injury—then insurance becomes impossible. You can only pool
risk among people who share it.
In our medical system, this problem is unfolding year by year. Genetic testing
and improved statistics threaten the basic logic of insurance.
I see no easy solution. If you individualize the rates, then some folks pay
pennies while others face premiums almost as steep as the medical bills
themselves. But charging everyone the same amount turns a project of mutual
benefit, with each of us hedging against risk, into a project of collective charity,
with some folks subsidizing others. That’s a harder sell. It’s one reason why
American healthcare remains so contentious.
As a teacher, I’m inclined to think of all knowledge as a gift. But insurance
complicates that story. Our ignorance of fate can force us to collaborate against
it. We have built a democracy from our uncertainty—and new knowledge
threatens that balance, as sure as flood or fire.
 Chapter 15
HOW TO BREAK THE ECONOMY WITH A SINGLE PAIR
OF DICE

1. THE HAUNTED CAREER FAIR

In September 2008, I began my final year of the great free-pizza scavenger hunt
called “college.” Knowing, if not quite believing, that after college only wage
earners could acquire pizza, I decided to attend the annual career fair. Booths of
employers would fill the gymnasium to offer free giveaways and (secondarily)
job applications.
But when I arrived, I found a half-empty gymnasium, a ghost town. In abrupt
unison, the investment banks had all decided that maybe now wasn’t such a good
time to hire.
We knew why. In the prior month, the global financial system had frozen,
tossed up a blue screen of death, and was now refusing to reboot. Wall Street
had devolved into the final scene of a Shakespearean tragedy: century-old
institutions lay heaped in the dirt, swords sticking out of them, wheezing death
monologues. Journalists batted around terms like “worst,” “recession,” and
“since the Great Depression,” often consecutively. Even the pizza crusts tasted
of anxiety.
In this chapter, we come to our final lesson on probability, perhaps the
hardest. Many would-be probabilists fall for the seductive idea of independence,
choosing to imagine our world as an aggregation of isolated events. But if
probability is to confront the world’s uncertainty, it must face the world’s
interconnectedness: its narrative threads, its causal chains.
A simple illustration: What’s the difference between rolling two dice and
doubling a single die?
Well, in each case, the sums range from a minimum of 2 (snake eyes) to a
maximum of 12 (double sixes).
With two independent dice, extreme values can unfold only a few ways. (For
example, only two combinations will yield a 3). Middle values can unfold in
many ways; for example, six different combinations will yield a 7. Thus, middle
values are more likely.
 What about rolling just one die and doubling it? Now the “second” roll
depends entirely on the first; we’ve got a single event disguised as two. And so
the extreme outcomes are just as likely as the middle ones.
The difference is stark. Independence irons out extremes; dependence amplifies
them.

We can push it further. Let’s move from two dice to a million. Now, the
outcomes range from a minimum of 1 million (all ones) to a maximum of 6
million (all sixes).
What if each die acts alone, independent of the other 999,999? Then we find
ourselves in the long-run world of stable tendencies, where the exciting 6’s and
the disappointing 1’s arrive in equal proportions. The overwhelming likelihood
is that our total for all the dice will land smack in the middle, far from both
extremes. There’s a 99.9999995% probability we’ll land between 3.49 million
and 3.51 million. Getting the minimum of 1 million is all but impossible, its
probability less than 1 in a googol googol googol googol. (I could write the word
“googol” another 7000 times, but you get the idea.)
But what if we’re not rolling separate dice? What if we’re rolling a single die
and multiplying it by a million, reflecting it through a hall of mirrors? Well, then
we remain in the chaotic world of random chance. The later “rolls” depend
wholly on the first, providing no balance. Getting the minimum is no longer an
absurd when-pigs-fly proposition. Its probability is 1 in 6.

Insurance, portfolio diversification, and egg-in-basket placement all rely on

the same fundamental principle: Overcome risk by combining events. One stock
is a gamble; combine many of them into an index, and you’ve got an investment.
But all of this depends on independence. It’s pointless to scatter your eggs
across several baskets, then fasten the baskets together and load them into the
bed of the same pickup truck. A world of dependence is a world of feedback
loops and domino cascades, a world of extremes. It’s a world where career fairs
look like either festivals or funerals, with little in between, a world where banks
all thrive until they abruptly go down together.

2. EVERYTHING HAS A PRICE

Pop quiz! What is the basic activity of Wall Street banks?

A. Powering the world economy via the intelligent allocation of capital

B. Buying Italian suits with blood money snatched from the pockets of the
working class
C. Pricing things

If you answered A, then you work for Wall Street. (Hey, nice suit! Is that
Italian?) If you answered B, then I’m honored that you’re reading my book,
Senator Sanders. And if you answered C, then you’re already familiar with a key
theme of this chapter: that the basic function of the financial sector is to decide
what things are worth. Stocks, bonds, futures, Rainbow Shout® contracts,
standard Parisian barrier options, single-use monocular default swaps… whether
you’re buying, selling, or Googling to see which of those I made up, you want to
know what these things are worth. Your livelihood depends on it.
The problem, of course, is that pricing ain’t easy.

Take bonds. These are pieces of debt, promises that you’ll get paid back.
Consider somebody who borrowed money to buy a house and promises to pay
back $100,000 in five years.
What’s this IOU worth to you?
Well, we begin with Pricing Challenge #1: putting a value on time. In the
carpe diem logic of finance, a dollar today is worth more than a dollar tomorrow.
First, there’s inflation (which gradually lowers the value of a dollar), and second,
there’s opportunity cost (i.e., a wisely invested dollar today will be worth more
by next year). As a crude estimate, we’ll say a dollar today is worth $1.07 next
year. Carry that forward, multiplying year after year, and you’ll find that $1
today is equivalent to $1.40 in five years.

Receiving $100,000 in five years isn’t as glamorous as it sounds. It’s like

getting paid just $71,000 today.
Is that figure the bond’s true price? Can we call it a day and go wash that
icky Wall Street feeling off our hands? Alas, we’re just getting started. We must
also put a price on risk: Who owes us the debt, and can we count on them? If it’s
a two-income household with perfect credit history and sparkling teeth, then the
odds are with us. But if our debtor is someone shiftier—say, a pizza-addicted
recent college grad with a penchant for bad drawings—then there’s a real chance
our bond will amount to nothing.
How do we adjust the price?
Simple: expected value. If there’s a 90% chance of getting paid back, then
the bond is worth 90% as much.
 We’re still not done. Default isn’t binary, with debtors paying back all or
nothing. In reality, courts and lawyers will intervene to hammer out a deal in
which creditors receive some fraction of what they’re owed, ranging from
pennies per dollar to almost the whole sum. Our bond, then, is like a bizarro
high-stakes lottery ticket. How do you put a single price on such variety?
Again: expected value. We make educated guesses about how much we
might get paid back, and then we imagine buying millions and millions of such
bonds. Instead of straining to divine the unknowable price of this specific bond,
we compute the long-run average of all such bonds.
 And there you have it. A $50,000 bond.
On Wall Street, pricing is like breathing: routine, constant, and essential for
survival. But for decades, just about the only goods that banks felt comfortable
evaluating were stocks (pieces of companies) and bonds (pieces of debt). That
excluded “derivatives,” which are neither stocks nor bonds but their mutant
descendants. They lived at the fringes of the finance industry, like a shady
alleyway casino flanking the respectable bank.
With the 1970s came a sea change: “quantitative analysis.” With the power
of mathematical modeling, “quants” figured out how to price derivatives—even
those of Dr. Seuss–level weirdness. Among the most complicated of all were
CDOs: collateralized debt obligations.
Although they came in many varieties, a common recipe went something like
this:

1. Gather thousands of mortgages (like the one we priced earlier) into a

single bundle.
2. Slice the bundle into layers (called “tranches”), ranging from “low-risk”
to “high-risk.”
3. When interest payments come pouring in, the owner of the low-risk
tranche gets paid first, and the owner of the high-risk tranche gets paid
last.
 CDOs offered an à la carte menu of risks and rewards, a tranche to match any
taste. Willing to pay extra for a safe bet? Consider this delicious upper tranche.
Looking for a cheaper higher-risk option? Then I’ve got a spicy lower tranche
you’ll just love. You prefer something in between? Well, just let our chefs know,
and I’m sure they can prepare it to order.
Investors were smacking their lips and banging on the table for more… right
up until September 2008, when the bill came.

3. THE PROBLEM OF THE HOUSE

Flash back to 1936, when surrealist painter René Magritte drew a series called
“Le Problème de la Maison.” The sketches show houses in peculiar spots:
nestled in tree branches, hiding in cliffside caves, piled at the bottom of a giant
ditch. In my favorite, a house stands upright in an empty plain, looking ordinary
except for its two neighbors: a pair of gargantuan dice.

Who knows what Magritte meant by it? This is the man who once painted a
bird grabbing a lady’s shoe and titled it God Is No Saint. But I think he’s
challenging our idea of the home as an emblem of security. Instead—and I don’t
mean to spook you—the house is something chancy and precarious, an
existential risk. Your house will likely be the largest investment of your lifetime,
costing several times your annual salary, incurring a debt that will last a
generation. The house is an icon not of stability but of probability.
Seven decades after Magritte’s visual pun, Wall Street faced its own little
problème de la maison: pricing CDOs. The problem was to identify the
relationship between the various mortgages. Last I checked, you and I are
separate people. So if I default, perhaps that has no bearing on you. On the other
hand, we share the same economy. We can escape a big recession no more than
we can escape summer pop songs. So if I default, perhaps that suggests you’re in
danger too. In Wall Street terms, the question was whether defaults pose an
idiosyncratic risk or a systematic one.
Are houses like thousands of separate dice, each roll independent of the
others? Or are they like a single die, multiplied thousands of times by a hall of
mirrors?
Imagine a CDO (tiny by real-world standards) built from 1000 of the
mortgages we explored above. Since we priced one at $50,000, the whole bundle
should be worth $50 million.
Now, if the mortgages are independent, Wall Street can sleep easy. Sure, our
investment might come in $1 million below what we expected, but it would take
absurd misfortune to lose $2 million, and a loss of $5 million is unthinkable, its
probability less than 1 in 1 billion. Independence stabilizes, all but eliminating
the chance of catastrophic loss.
 Meanwhile, if all the houses represent a single roll of the dice, then Wall
Street shall suffer night sweats and scream dreams. Dangers unthinkable a
moment ago are now very, very thinkable. In this deal, there’s a horrifying 1-in-
3 chance we lose nearly half of our investment, and a bone-chilling 1-in-10
chance that we lose everything.
 Of course, neither vision quite matches reality. We are neither hive-minded
drones living in perfect sync, nor rugged individualists unaffected by our
neighbor’s weather. Instead, our lives are somewhere in between, our futures
delicately intertwined. It seems obvious that one default heightens the likelihood
of another—but by how much, and under what conditions? These are the subtlest
challenges that probabilistic models can face.

Wall Street’s solutions included the notorious “Gaussian copula.” This

formula originated in life insurance, where it helped to adjust the probability of
one spouse’s survival after the other spouse’s death. Replace “spouse” with
“house,” and “death” with “default,” and you’ve got a model for calculating the
dependencies between mortgages.
The formula captures the relationship between two mortgages with a single
number: a correlation coefficient between -1 and 1.
 To its credit, the copula is a nifty piece of mathematics, simple and elegant.
But the global economy doesn’t really go in for simplicity, and in hindsight, it’s
easy to see what the copula (and similar approaches) missed.
First, the data. Wall Street’s computers feasted on spreadsheets full of
housing prices. But most of the figures came, in city after city, from the same
recent time period, during which (it just so happened) home prices had enjoyed a
steady rise. The models calibrated their confidence as if they had studied the
entire history of US housing. In fact, they had witnessed a single roll of the dice,
multiplied by a million.
Second, the model was built for couples (hence “copula”). But houses don’t
come in pairs. They exist, all together, as a loose nationwide market. A single
change can affect every mortgage in the country, all at once—which is what
happened when the soaring market fell to earth. It’s silly to worry about one
domino knocking over its neighbor, when there’s a single giant domino looming
above them all.

Finally, if you know your statistical vocabulary, alarm bells start ringing at
the term “Gaussian.” In mathematics, the word arises whenever you add together
lots of independent events. But that’s the whole problem: these events weren’t
independent.
In all these ways, Wall Street overlooked the risk that Magritte saw, leading
to a calamity as surreal as anything the painter ever envisioned. And still, we
haven’t discussed the worst of it. The mispriced CDOs amounted to a few
trillion dollars—enough to damage the economy but not enough to explain the
body blow it suffered in September 2008. What left the economy collapsed on
the mat, dizzy and spitting teeth, wondering in bewilderment where that sucker
punch had come from? It was a probabilistic failure on an even grander scale.

4. THE $60 TRILLION DOUBLE-OR-NOTHING

As you’ll know if you read the previous chapter—a laugh-out-loud joyride for
the whole family—insurance makes sense. I can’t afford to lose my house
(where would the pizza deliveries come to?), so I’m willing to pay a little
premium each month, in exchange for a big payout should my home ever burn
down. I eliminate an existential risk; the insurance company turns a profit.
Everybody wins.
But here’s a weird notion. What if you insure my house?
You can mirror the bargain I made. Small regular payments, leading to a big
windfall should disaster strike. This duplicates the financial structure of
insurance, but without the purpose. It’s just a wager, a win-lose zero-sum game.
If my house burns down, you win; if my house stays safe, the insurance
company wins. Even weirder: what if thousands of people hop on board, each
buying insurance on my house, standing to win if it burns down?
 I may fret as I begin to receive anonymous gifts, such as cheap firecrackers
and working grenades. But I’m not the only one losing sleep in this scenario; the
insurance company is even more freaked-out than I am. If my house burns,
they’re left holding fistfuls of ash, on the hook for thousands of enormous
payouts.
That’s why no insurance company would agree to the situation. The 95%
chance of easy profit cannot offset the 5% chance of utter ruin.
It’s too bad no one pointed this out to Wall Street.
For Wall Street, doom came with a three-letter acronym: CDS. It stood for
“credit default swap,” which was basically an insurance policy on a CDO. You
make modest, regular payments. As long as the CDO keeps paying interest,
nothing happens. But if enough mortgages default, then the CDS kicks in with a
big payout.
So far, so reasonable. But guess what Wall Street did next? They sold dozens
of CDSs for each underlying CDO. It’s like selling dozens of insurance policies
on the same house, except with a lot more zeroes on the end. By the start of
2008, the amount of money on the line had reached $60 trillion. That is roughly
the GDP of the planet Earth.

A quick recap. CDOs were meant to achieve the stability of a million dice;
instead, they embodied the fragility of a single die. The CDSs doubled down
until the gamble loomed large enough to imperil the entire world economy. It all
prompts a natural question:
How could Wall Street have been so stupid?
You harvest the greatest minds from the fanciest universities, buy them
million-dollar supercomputers, pay them cosmic salaries, work them 90 hours a
week… and then walk in to find them screaming and jamming forks in the wall
sockets?
I wish I could write off these “independence vs. dependence” errors as a one-
time aberration, a peculiar circumstance of CDOs and CDSs. But if wishes were
houses, the CDOs might have stood a chance. The grim fact is that this error
runs to the core of financial markets.
 5. ASHES, ASHES, WE ALL FALL DOWN
At the risk of being dubbed a neoliberal shill, I will state my belief that markets
work pretty well. Heck, I’ll go further: they work really well.
For example, this planet happens to harbor delicious fruit spheres called
“apples.” How should we allocate them? If farmers grow more than consumers
want, we’ll find piles of Red Delicious rotting in the streets. If consumers want
more than farmers grow, we’ll witness strangers clawing at one another to snag
the last McIntosh during a shortage. Yet somehow, against all odds, we manage
to grow just about the right number of apples.
The trick? Prices. Although we think of prices as determining our behavior
(“too expensive, so I won’t buy”), the reverse is just as true. Each of our
individual choices exerts a tiny influence on the price. If enough people refuse to
buy, the price will drop. If enough refuse to sell, the price will rise. The price
functions as an aggregate of all our independent judgments and decisions.
And so, like other aggregates of independent events, prices tend to yield
balanced, stable, reasonable outcomes. Aristotle called it “the wisdom of the
crowd.” Adam Smith called it “the invisible hand.” I call it “the independent
dice all over again, except this time, the dice are us.”

In theory, what works for apples ought to work for CDOs. Some folks will
overvalue them. Others will undervalue them. But in the end, a market full of
independent investors will drive the price to a stable equilibrium.
There’s just one problem: all too often, investors behave less like millions of
independent dice, and more like a single die multiplied by millions.
Take the 1987 stock crash. On October 19, prices nose-dived, plummeting
more than 20%. It came without warning: no market-shaking news, no high-
profile bankruptcy, no “Oh jeez, I’m in over my head” speech from the Fed
chair. Markets just collapsed. Only later did the postmortem identify a peculiar
trigger: lots of Wall Street firms had relied on the same basic theory of portfolio
management. Many even employed the same software. As markets slipped, they
sold the same assets in unison, sending prices into a downward spiral.
Portfolio management’s whole purpose is to bring safety through
diversification. But if everyone diversifies in exactly the same way, then the
resulting market ain’t so diverse.
If investors exercise their own judgment, then day-to-day price changes
should follow a bell-shaped normal distribution: up a little on some days, down a
little on others, but almost never leaping too fast or far. Alas, reality disagrees.
Market movements, with their occasional massive drops, follow more of a
power-law distribution. That’s the same mathematical model we use for
earthquakes, terrorist attacks, and other massive disruptions to highly sensitive
systems.
The market isn’t random like lots of dice added together. It’s random like an
avalanche.
 Leading up to the financial meltdown of 2008, many banks relied on the
same limited set of models (such as the Gaussian copula). Rather than bringing
fresh insight, they herded around a single strategy. Even ratings agencies—
whose purpose and duty is to offer independent analysis—just parroted the
banks’ own claims. The referees became cheerleaders.
Why, back in September 2008, did I arrive to find a somber half-full
gymnasium? Why, in other words, did the financial system melt down?
Well, it’s complicated. Incompetence played its part, as it does in most
failures. (Just watch me bake sometime.) So did misaligned incentives, blind
optimism, naked greed, dizzying complexity, governmental dysfunction, and
interest rates. (Again, see my baking.) This brief chapter has told only a sliver of
the story, focused around a specific theme: the dangerous assumption of
independence, when dependence actually reigns.
Mortgages default together. CDSs pay out together. And actors in the
marketplace herd together around similar pricing strategies.
You want to break the economy with a single pair of dice? Honestly, it’s
easy. Just convince yourself that you’re rolling a million pairs, and then bet your
fortune on the roll.
 IV
STATISTICS

THE FINE ART OF HONEST LYING

A survey of medical professionals asked them to contrast clinical evaluation

with a more statistical approach. These are the words they used to describe each
method:
 On behalf of statisticians everywhere: ouch.
I admit that there is something reductive about the whole project of statistics,
of taming the wild, unpredictable world into docile rows of numbers. That’s why
it’s so important to approach all statistics with skepticism and caution. By
nature, they compress reality. They amputate. They omit. They simplify.
And that, of course, is the precise source of their power.
Why do scientific papers have abstracts? Why do news articles have
headlines? Why do action movies have trailers cataloguing all the best
explosions and postexplosion quips? It’s because simplification is vital work.
Nobody’s got time to admire the kaleidoscopic brilliance of reality all day.
We’ve got places to go, articles to skim, YouTube videos upon which to cast our
mindless gaze. If I’m going to a new city in July, I don’t seek a novelistic
portrait of the humidity and heat; I look up the temperature. Such a statistic is
not “living” or “deep” or “configural” (whatever that means). It’s plain and clear
and useful. By condensing the world, statistics give us a chance to grasp it.
And they do more, too. Statistics classify, extrapolate, and predict, allowing
us to build powerful models of reality. Yes, the whole process depends on
simplification. And yes, simplification is lying by omission. But at its best,
statistics are an honest kind of lying. The process calls upon all the virtues of
human thought, from curiosity to compassion.
In that way, statistics are not so different from stick figures. They’re bad
drawings of reality, missing hands and noses—yet speaking a peculiar truth all
their own.
 Chapter 16
WHY NOT TO TRUST STATISTICS

AND WHY TO USE THEM ANYWAY

Okay, let’s get this out of our system. Statistics are lies. Not to be trusted. All of
history’s wittiest people have said so. Haven’t they?

FAMOUS QUOTE I FOUND BY GOOGLING: “There are three kinds of lies: lies,
damned lies, and statistics,” said Mark Twain.
REALITY I FOUND BY DEEPER GOOGLING: People today often misattribute this to
Twain, which is perhaps fair, since Twain himself misattributed it to
Benjamin Disraeli. The correct attribution is unknown.

FAMOUS QUOTE I FOUND BY GOOGLING: “Do not trust any statistics you did not
fake yourself,” said Winston Churchill. (Or perhaps: “I only believe in
statistics that I doctored myself.”)
REALITY I FOUND BY DEEPER GOOGLING: A lie meant to slander Churchill. It may
have originated with Nazi propagandist Joseph Goebbels.

FAMOUS QUOTE I FOUND BY GOOGLING: “87% of statistics are made up on the

spot.”
REALITY I FOUND BY DEEPER GOOGLING: “And 87% of quotes are misattributed on
the spot,” said Oscar Wilde.

FAMOUS QUOTE I FOUND BY GOOGLING: “There are two kinds of statistics, the kind
you look up and the kind you make up,” said Rex Stout.
REALITY I FOUND BY DEEPER GOOGLING: Rex Stout was a novelist. He didn’t say
this; one of his characters did.
FAMOUS QUOTE I FOUND BY GOOGLING: “Politicians use statistics in the same way
that a drunk uses lampposts—for support rather than illumination,” said
Andrew Lang.
REALITY I FOUND BY DEEPER GOOGLING: Okay, this one is real, and it’s great.

FAMOUS QUOTE I FOUND BY GOOGLING: “Statistical thinking will one day be as

necessary for efficient citizenship as the ability to read or write,” said H. G.
Wells.
REALITY I FOUND BY DEEPER GOOGLING: Ack—even my prostats saying is
misquoted. Wells actually said that the day is foreseeable when it will be “as
necessary to be able to compute, to think in averages and maxima and
minima, as it is now to be able to read and write.”

What’s my point? Yes, numbers can deceive. But so can words—not to

mention pictures, hand gestures, hip-hop musicals, and fundraising emails. Our
moral system blames the liar, not the medium chosen for the lie.
To me, the most interesting critiques of stats drive not at the dishonesty of the
statistician but at the mathematics itself. We can boost statistics’ value by
understanding their imperfections, by seeing what each statistic aims to capture
—and what it deliberately omits. Maybe then we can become the citizens that H.
G. Wells envisioned.

1. THE MEAN
 How It Works: Add up all the data you’ve got. Divide the total by the size
of the data set.
When to Use It: The mean (or “average,” as it’s often known) fills a basic
need in statistics: to capture the “central tendency” of a group. How tall is that
basketball team? How many ice-cream cones do you sell per day? How did the
class do on the test? If you’re trying to summarize a whole population with a
single value, then the mean is a sensible first port of call.
Why Not to Trust It: The mean considers only two pieces of information:
the total, and the number of people contributing to that total.
If you’ve ever split a hoard of pirate treasure, then you see the danger: There
are many ways to share a haul. How much does each individual contribute? Is it
balanced or grossly one-sided? If I eat a whole pizza and leave nothing for you,
is it fair to claim that our “average consumption” was half a pizza? You can tell
your dinner guests that the “average human” has one ovary and one testicle, but
will this not bring conversation to an awkward standstill? (I’ve tried; it does.)
Humans care about questions of allocation. The mean shrugs them off.
 There’s a saving grace, which is that this feature makes the mean simple to
calculate. Say your test scores are 87, 88, and 96. (Yeah, you’re crushing this
class.) What’s your average? Don’t overheat your neurons with addition and
division; instead, reallocate. Take six points from your last test; give three to the
first and two to the second. Now your scores are 90, 90, and 90, with a single
point left over. Sharing that lonely point among the three tests, you arrive at a
mean of 90¹/³, with no brain strain required.

2. THE MEDIAN
 How It Works: The median is the middlemost member of your data set. Half
your data lies below it, and half above.
When to Use It: The median, like the mean, captures the central tendency of
a population. The difference lies in its sensitivity to outliers—or rather, its
insensitivity.
Take household income. In the US, a wealthy family may earn dozens (even
hundreds) of times as much as a poor one. The mean, pretending that every
family has an equal share of the total earnings, is seduced by these exceptional
values, and is led away from the bulk of data points. It arrives at the figure of
$75,000.
The median resists the outliers’ gravitation. It names instead the nation’s
absolute middle household income, the perfect midpoint where half of families
are wealthier and half poorer. In the US, it’s near $58,000. Unlike the mean, this
offers a clean picture of the “typical” household.
Why Not to Trust It: Once you’ve found your median, you know that half
of the data lies above and half below. But how far away are those points—a
hair’s breadth from the median or a transcontinental flight? You’re looking only
at the middle slice of pie, never mind how big or small the other slices are. This
can lead you astray.
 When a venture capitalist invests in new businesses, she expects most of
them to fail. The rare 1-in-10 home run compensates for all those little losses.
But the median misses this dynamic. “The typical outcome is negative,” it
squawks. “Abort mission!”
Meanwhile, an insurance company builds a careful portfolio, knowing that
the rare 1-in-1000 disaster can wipe out years of modest profits. But the median
overlooks the potential for calamity. “Hey, the typical outcome is positive,” it
cheers. “Never stop!”
This is why you’ll often find the median displayed alongside the mean. The
median reports on the typical value; the mean, on the total. Like two imperfect
witnesses, they tell a fuller story together than either can alone.

3. THE MODE
 How It Works: It’s the most common value, the hippest, most fashionable
data point.
What if each value is unique, with no repetitions? In that case, you can group
the data into categories, and call the most common one “the modal category.”
When to Use It: The mode shines in conducting polls and in tabulating
nonnumerical data. If you want to summarize people’s favorite colors, you can’t
very well “total the colors up” to compute a mean. Or, if you’re running an
election, you’ll drive citizens mad if you line up the votes from “most liberal” to
“most conservative” and award the office to the recipient of the median ballot.
Why Not to Trust It: The median ignores the total. The mean ignores its
allocation. And the mode? Well, it ignores the total, its allocation, and just about
everything else.
The mode seeks a single most common value. But “common” does not mean
“representative.” The modal salary in the United States is zero—not because
most Americans are broke and jobless, but because wage earners are spread
across a spectrum from $1 to $100 million, whereas all wageless people share
the same number. The statistic doesn’t illuminate anything about the US. It’s
true in virtually every country, an artifact of how money works.
 Turning to “modal categories” can only partway solve the problem. It places
a surprising power in the hands of the person presenting the data, who can
gerrymander the category boundaries to suit their agenda. Depending on how I
draw the lines, I can claim that the modal household in the US earns $10,000 to
$20,000 (going by increments of 10,000), or $20,000 to $40,000 (going by
increments of 20,000), or $38,000 to $92,000 (going by tax brackets).
Same data set, same statistic. And yet the portrait changes completely,
depending on the artist’s choice of frame.

4. THE PERCENTILE
 How It Works: Recall that the median splits your data set right down the
middle. Well, the percentile is a median with a dimmer switch. The 50th
percentile is the median itself (half the data above and half below).
But you can pick other percentiles, too. The 90th lands near the top of your
data: Only 10% sits above while 90% sits below. Meanwhile, the 3rd resides
near the bottom of your data set: only 3% is below, with 97% above.
When to Use It: Percentiles are handy, flexible, and perfect for that favorite
human pastime: ranking things. That’s why standardized tests often deliver their
scores as percentiles. A raw score like “I got 72% of questions right” isn’t very
informative. Were these vicious snakebite questions or slow-pitch softballs?
However, “I’m at the 80th percentile” reveals right where you stand. You did
better than 80% of test takers and worse than 20%.
Why Not to Trust It: Percentiles share the median’s Achilles heel. They tell
you how much data lies above or below a certain point but not how far away it
is.
Take the finance industry, which employs percentiles to measure the
downside of an investment. You first imagine the spread of possible outcomes,
from triumph to cataclysm, and then pick a percentile (usually the 5th), which
you call “value at risk,” or VaR. It aims to capture the worst-case scenario. But
in fact, you’ll do worse exactly 5% of the time. VaR gives no clue of how much
worse, whether it’s pennies of extra losses, or billions.
You can better visualize the landscape of possibilities by checking more VaR
percentiles—for example, 3, 1, and 0.1—but by nature, a percentile cannot
detect the most violent and extreme losses, and so the true worst-case scenario
will always lurk just beyond sight.

5. PERCENTAGE CHANGE
 How It Works: Don’t just report a change; first, divide it by the original
total.
When to Use It: Percentage change is all about putting things in perspective.
It frames gains and losses as a proportion of the whole.
Take a gain of $100. If I only had $200 to begin with, then this windfall
marks 50% growth, and I’m dancing a hallelujah Snoopy dance. But if I already
had $20,000, then my new income represents just 0.5% growth; I’ll content
myself with a single triumphant fist pump.
Such perspective taking is crucial when you’re watching a quantity grow
over time. If the Americans of 70 years ago had heard that our GDP increased by
$500 billion last year, they’d be awed. If they’d heard that it grew by 3%, they’d
feel right at home.
Why Not to Trust It: Hey, I’m all about perspective. But percentage change,
in its effort to provide context, can annihilate it instead.
When I lived in the UK, the tasty £2-per-jar tomato sauce sometimes went on
sale for £1. Those days felt like jackpots: 50% savings! I’d lug home a dozen
jars, enough for months of ravioli. Then, later, I’d find myself buying plane
tickets to attend a wedding in the US. Delaying the purchase a week might incur
a price bump of 5%. “Ah well,” I’d say as I accepted the higher rate. “It’s only a
little more.”

You can see the problem: my instincts run penny-wise and pound-foolish.
The “huge” discount on tomato sauce saved me £12 while the “modest” increase
on plane tickets cost me £30. A dollar is a dollar is a dollar, whether it’s part of a
$20 grocery bill or a $200,000 mortgage. Big price drops on cheap products are
dwarfed by a small bump on a rare extravagance.

6. THE RANGE
 How It Works: It’s the distance between the biggest and the smallest data
point.
When to Use It: The mean, median, and mode deal with “central tendency”:
they aim to collapse a diverse population down to a single representative value.
The “range” takes the opposite goal: not to sweep disagreement under the rug,
but to quantify and display it, to give a measure of the data’s “spread.”
The range’s merit lies in its simplicity. It conceives of a population as a
spectrum running from “smallest” to “largest,” and it tells us the width of that
spectrum. It’s a quick-and-dirty summary of variety.
Why Not to Trust It: The range cares only about the biggest and smallest
slice of cake. That omits an awful lot of crucial information: namely, the sizes of
all those middle slices. Are they close to the maximum? Close to the minimum?
Scattered evenly in between? The range neither knows nor cares to ask.
The larger the data set, the more dubious the range becomes, as it disregards
millions of intermediate values to inquire about the two most extreme outliers. If
you were an alien who learned that there’s a 7-foot range of adult human heights
(from the record low of less than 2 feet to the high of almost 9), you’d be quite
disappointed to visit Earth and meet all of us dull 5-to-6-footers.

7. THE VARIANCE (AND THE STANDARD

DEVIATION)
 How It Works: Standard deviation tells you, in rough terms, how far the
typical data point is from the mean.
If you want to cook up a variance in your home kitchen, then the recipe goes
like this: (1) find the mean of your data set; (2) find how far each data point is
from the mean; (3) square those distances; and (4) average those squared
distances. This gives you the “average squared distance from the mean,” i.e., the
variance.

If you take a square root at the end, you’ll get the “standard deviation.” It’s
the more intuitive measure, since variance has strange squared units. (What’s a
“square dollar”? No one knows.)
Since variance and standard deviation go together, that’s how I’ll discuss
them.
When to Use It: Like the range, variance and standard deviation quantify a
data set’s variety, but—and I say this with the impartiality of a loving parent—
they’re better. The range is a quick, makeshift measure of spread; the variance is
a load-bearing pillar of statistics. By taking a contribution from every member of
the data set, the variance achieves the sophistication of a symphony, in contrast
to the range’s two-note ditty.
The variance’s logic, though convoluted, makes sense on inspection. It’s all
about the data’s distance from the mean. “High variance” means that the data is
scattered far and wide; “low variance” means that it’s huddled close together.
Why Not to Trust It: Sure, the variance takes a contribution from every data
point. But you can’t tell who is contributing what.
In particular, a single outlier can give the variance a huge boost. Thanks to
the squaring step, a single large distance (e.g., 122 = 144) can yield a greater
contribution than a dozen small ones (e.g., 32 = 9; 12 such terms total to just 96).
 Variance has another feature that stumps lots of folks. (It’s not a bad trait,
just a counterintuitive one.) Students tend to consider a data set with many
different values (e.g., 1, 2, 3, 4, 5, 6) as more “spread out” than a data set with
repeated values (e.g., 1, 1, 1, 6, 6, 6). But variance isn’t interested in “variety”; it
cares only about distance from the mean.

In variance’s eyes, the spread of the latter set (with its repeated far-from-the-
mean values) outweighs the spread of the former (with its nonrepeated but
closer-to-the-mean values).
8. THE CORRELATION COEFFICIENT
 How It Works: A correlation measures the relationship between two
variables. A person’s height and weight, for example. Or a car’s price and its
number sold. Or a movie’s budget and box office receipts.
The scale runs from a maximum of 1 (“wow, they totally go together”) to a
midpoint of zero (“huh, there’s no connection here”) to a minimum of -1 (“hmm,
they go opposite directions”).
That’s the quick-hit summary, anyway. To see how the correlation coefficient
actually works, check the endnotes.
When to Use It: Are wealthier nations happier? Does “broken windows”
policing prevent crime? Does drinking red wine prolong life spans, or just dinner
parties? All these questions deal with connections between pairs of variables,
between imagined causes and conjectured effects. Ideally, you’d answer them by
running experiments. Give 100 people red wine, give 100 grape juice, and see
who lives longer. But such research is slow, expensive, and often unethical.
Think of that piteous wine-deprived control group.
Correlation lets us tackle the same question sideways. Find a bunch of
people, measure their wine intakes and life spans, and see if the wine drinkers
live longer. Admittedly, even a strong correlation can’t determine causation.
Maybe wine lengthens life. Maybe long life drives people to drink. Maybe both
are driven by some third variable (e.g., rich people both live longer and can
afford more wine). It’s impossible to know.
Even so, correlational studies offer a great starting point. They’re cheap and
fast, and they allow for large data sets. They can’t pinpoint causes, but they can
offer tantalizing clues.
Why Not to Trust It: The correlation coefficient is one of the most
aggressive statistical summaries of all. It collapses hundreds or thousands of data
points, each with two measured variables, into a single number between -1 and
1. Suffice it to say, that’s going to leave some things out, a fact illustrated by the
mathematical oddity known Anscombe’s quartet.
Let us step into Anscombe Academy for Witchcraft and Wizardry, where
students have spent weeks preparing for exams in four classes: Potions,
Transfiguration, Charms, and Defense Against the Dark Arts. For each exam, we
shall consider two variables: the hours each student spent studying and that
student’s score (out of 13) on the exam.
From the summary stats, you’d think the four tests were identical:

And yet… well, just look. (Each dot represents a student.)

 The first, Potions, fits my stereotype of how tests work. Studying more tends
to improve your result, but it’s no sure thing. Random noise intervenes. Thus,
the correlation is 0.816.
 The Transfiguration scores, meanwhile, follow a perfect linear relationship,
with every extra hour of study yielding an extra 0.35 points on the test—except
for one peculiar outlier child, who lowers the correlation from a perfect 1 down
to 0.816.

The Charms test follows an even more certain pattern: studying improves
your scores, but with diminishing marginal returns. When you reach 10 hours,
extra study time begins to harm your score (perhaps by costing you sleep).
However, correlation is built to detect linear relationships, and so it misses the
nature of this quadratic pattern, resulting in a coefficient of 0.816.
 Finally, in Defense Against the Dark Arts, every pupil studied for eight
hours, meaning that study time can’t help you predict the final score. There’s
just one exception: a hardworking outlier whose 19 hours of study paid off with
the runaway best score. That single data point suffices to strengthen the
correlation from 0 all the way up to…
0.816.
Each test follows its own distinct logic, obeying its own unique pattern. But
the correlation coefficient misses the story.
Then again, that’s the nature of statistics. As I like to say:

A statistic is an imperfect witness. It tells the truth, but never the

whole truth.

Feel free to quote me. Or, if you favor the traditional method, just make up
your own saying and attribute it to me.
 Chapter 17
THE LAST .400 HITTER

THE RISE AND FALL OF BATTING AVERAGE

Ever since its birth, baseball has been a game of numbers. Today, Wikipedia
lists 122 types of baseball stats, from DICE to FIP to VORP, and I suspect these
only scratch the surface. Go ahead: type three random letters; I wager that
someone, somewhere, keeps a fastidious record of that statistic.

This chapter will explore a single statistic, from its humble origin to its
gradual demise. That statistic is BA: Boston accents. Sorry, BA: beer absorbed.
Okay, fine, it’s BA: batting average.
 Once upon a time, batting average reigned supreme. These days, statisticians
treat it like a piece of kitsch, a relic of simpleminded days. Is it time to put the
batting average out to pasture? Or does the old veteran with the aching joints still
carry a spark of magic?

1. BOXED LIGHTNING
In 1856, an Englishman named Henry Chadwick, working as a cricket reporter
for the New York Times, happened to attend his first baseball game. It caught his
fancy. “In baseball all is lightning,” he gushed, as only a cricket fan could. Like
a sloth smitten with the flash and pizzazz of a tortoise, Chadwick soon devoted
his life to this American pastime. He served on rules committees, wrote the
sport’s first book, and edited its first annual guide. But what earned Chadwick
his nickname as “the Father of Baseball” was something more elemental: the
stats.

Chadwick invented the “box score,” a table for tracking the game’s key
events. By scanning columns of numbers—runs, hits, putouts, and so on—you
could almost watch the innings unfold. Box scores weren’t about long-term
predictive power or statistical validity. They were about credit and blame, heroes
and villains, a story told in numbers. They summarized weather conditions and
highlighted key events, aiming (in a time before radio, action photography, or
MLB.com) to help you glimpse the action. It was SportsCenter for the 1870s.
Chadwick’s idea for “batting average” came from cricket, where there are
only two bases and you score a run every time you move from one to the other.
Cricketers keep batting until they make an out, by which time a good player will
have scored dozens of runs. (The all-time record is 400.) Thus, cricket’s batting
average is defined as “runs scored per out made.” A great player can maintain an
average of 50, even 60.
This definition made little sense in baseball, where a single hit ends your at-
bat. So, like any good mathematician, Chadwick played with the rules, trying out
a few definitions before settling on the one used today.
 Batting average aims to capture your rate of success with a simple fraction:
your number of successful hits divided by the number of hits plus the number of
outs you make. Chadwick called it the “one true criterion of skill at the bat.”
In theory, batting averages range from .000 (never hitting) to 1.000 (never
failing to hit). In practice, almost all players cluster between .200 and .350.
That’s not much of a gap. Baseball’s royals (hitting 30% of the time) and its
peasants (hitting 27.5%) differ by just one hit in every 40 attempts. The naked
eye can’t tell the difference. Even over a whole season, a “worse” hitter might
outshine a “better” one by sheer luck.

Hence, the stats. A player’s batting average is like a stop-motion video of a

flower growing from seedling to full bloom. It reveals a truth that otherwise
would lie beyond our senses. Instead of lightning in a bottle, we’ve got lightning
in a box score.
Statistics, like probability, bridges two worlds. First, there’s the messy day-
to-day reality of bad hops and lucky bounces. Second, there’s the long-run
paradise of smooth averages and stable tendencies. Probability begins with the
long-run world, and imagines how things might turn out on a single day.
Statistics does the opposite: it begins with the day-to-day mess, and strives to
infer the invisible long-run distribution from which that data emerged.
Put another way: a probabilist starts with a deck of cards, and describes what
hands might come up. A statistician looks at the hands on the table, and tries to
infer the nature of the deck.

Baseball, perhaps unique among sports, deals you enough cards to have a real
shot. In every 162-game season, a team’s batters face roughly 24,000 pitches.
For soccer to offer the same wealth of discrete data points, you’d have to reset
the ball at center field every five seconds, all season long. Better yet, whereas
other team sports unfold in multiperson melees, each baseball player bats alone,
his data clean and separable from his teammates’.
That’s the glory of batting average. But every statistic leaves something out
—and in this case, what’s missing is something crucial.

2. THE OLD MAN AND THE OBP

In 1952, Life magazine published the first edition of Ernest Hemingway’s The
Old Man and the Sea. The issue sold 5 million copies; the novelist won a Nobel
Prize.
On August 2, 1954, Life chose to steer the national conversation a different
direction: baseball statistics. Under the headline “Goodby to Some Old Baseball
Ideas,” Pittsburgh Pirates general manager Branch Rickey posited an equation
that would take 10 pages to unpack:

The formula itself was scarcely grammatical. The equals sign doesn’t really
mean “equals,” nor is the minus sign actual subtraction. Still, the article gave a
penetrating critique of several “old baseball ideas”—batting average chief
among them. The attack (credited to Rickey but ghostwritten by Canadian
statistician Allan Roth) begins with two letters: BB, for “base on balls.” Or, in
common parlance, a walk.
As baseball came of age in the 1850s, a player batted until he either hit the
ball into play, or swung and missed three times. With patient enough batters, the
game flowed like cold molasses. Thus, in 1858, “called strikes” were born. If a
player let a juicy pitch sail past, then it was treated as equivalent to swinging and
missing. But now the pendulum swung too far; cautious pitchers refused to
throw anything hittable. The solution, introduced in 1863, was to also call
“balls”: pitches deemed too off target for the batter to hit. Enough such “balls”
would grant the batter a free walk to first base.
 Walks stumped Chadwick. Cricket’s closest equivalent is a “wide,” generally
viewed as a mistake by the thrower. So batting average ignored walks, as if the
at-bats had never occurred. Walks weren’t deemed an official statistic until
1910.
Today, the most skilled and patient hitters walk 18% or 19% of the time;
their reckless, swing-happy peers, just 2% or 3%. Hence the first term in
Rickey’s equation: a convoluted expression for what we now call “on-base
percentage,” or OBP. It’s your rate of getting on base, via either hits or walks—
in other words, your rate of not making an out.
Which statistic better predicts the number of runs a team will score: BA or
OBP? Running the correlations for 2017, BA is pretty strong, with a coefficient
of 0.73. But OBP is outstanding, with a coefficient of 0.91.
 Next, Rickey (i.e., Roth) highlighted another shortcoming of batting average.
Hits come in four flavors, from a single (one base) to a home run (four bases).
As with desserts or retweets, more is better—but batting average treated them all
the same. Hence, the second term in Rickey’s equation, which amounts to a
measure of “extra bases” beyond first.
Today, we prefer a related statistic: “slugging percentage,” or SLG. It
calculates the average number of bases achieved per at-bat, from a minimum of
0.000 to a theoretical maximum of 4.000 (a home run every time). In practice, no
hitter has ever topped 1.000 for a whole season.
Like batting average, SLG ignores walks, and like any stat, it collapses
meaningful distinctions. For example, to slug .800 over 15 at-bats, you need to
get 12 total bases (since 12/15 = 0.8). There are a lot of ways to do that, not all
of them equal in value:

Since OBP and SLG each shed light on a different corner of the game,
they’re often used in conjunction. One common shortcut is to add the two,
arriving at a statistic known as “on-base plus slugging,” or OPS. In the 2017
data, its correlation with runs scored is an eye-popping 0.935—better than either
OBP or SLG alone.
On the 50th anniversary of the Life article, the New York Times showed the
formula to New York Yankees general manager Brian Cashman. “Wow!” he
said. “The guy was generations ahead of his time.” His praise belies the damning
truth: Cashman had never heard of the Life piece. Even after its publication,
batting average would reign for decades more, while OBP and SLG huddled in
the shadows. The baseball dialogue in The Old Man and the Sea probably had
more impact on the game than Rickey’s manifesto.
What was baseball waiting for?

3. KNOWLEDGE MOVES THE CURVE

To revolutionize anything, you need two conditions: the knowledge and the
need.
The knowledge came, in large part, from writer Bill James. In 1977, while
working night shifts as a security guard, he self-published the first Bill James
Baseball Abstract, a peculiar 68-page document consisting mostly of statistics,
with rigorous answers to insider questions like “Which pitchers and catchers
allow the most stolen bases?” It sold just 75 copies, but the buzz was good. The
next year’s edition sold 250. Five years later, James landed a major publishing
deal. In 2006, Time magazine profiled James (now on the payroll of the Boston
Red Sox) in its “Time 100” as one of the most influential people on planet Earth.

James’s clear-eyed analytical approach sparked a statistical renaissance in

baseball. He dubbed it “sabermetrics.” Among the movement’s insights was that
batting average is a sideshow, a crude indicator for an actual outcome—like
inferring the quality of a dinner by sampling one ingredient. If you’re serious
about assessing the meal, you need to taste all the ingredients—or, better yet, the
dish itself.
As the archivists at Life magazine can attest, this knowledge was old. What
brought it into the forefront was not just James, but also a need born from the
changing economic conditions of baseball. Until the early 1970s, baseball
players lived under the shadow of the “reserve clause.” It meant that, even after a
contract expired, the team retained rights to the player. You couldn’t sign (or, for
that matter, negotiate) anywhere else, except with your old club’s permission.
Then, in 1975, an arbiter’s ruling redefined the reserve clause, launching the
era of “free agency.” With the floodgates open, salaries soared.
 A decade earlier, owners could buy players like groceries. Now, the groceries
had agents, and those agents had salary demands. The new financial pressure
should have nudged owners away from crude measures like BA toward more
reliable stats like OBP and SLG—but as everyone but Henry Chadwick knows,
baseball is a slow-moving game. It took 20 years for one team, the Oakland A’s,
to gain sentience and begin using OBP to evaluate players.
The fire started under general manager Sandy Alderson in the early 1990s
and burned bright under his successor, Billy Beane. Soon, the A’s built an
outrageous run of success on smart statistical shopping. Then, in 2003, a Bay
Area resident named Michael Lewis wrote a book about Billy Beane. He called
it Moneyball, and along the way to selling 67 quadrillion copies, it did what Life
magazine had failed to do: it bid “goodby” to some old baseball ideas. With
Lewis’s help, OBP and SLG leapt from the fringes of fandom to the front-office
mainstream.

4. THE DRAMA OF THE 4TH DECIMAL

“Baseball is the only field of endeavor,” outfielder Ted Williams once said,
“where a man can succeed three times out of ten and be considered a good
performer.”
 In 1941, Williams was on pace to do one better: to succeed four times out of
10. This is called “batting .400” and it makes you a demigod, a legend. Williams
entered the final week of the season at .406, poised to become the first .400 hitter
in 11 years.
Then he faltered. In the next four games, he got just three hits in 14 chances,
bringing his average down to a heartbreaking .39955.
The number looks fake, like an example you’d cook up to test a student’s
grasp on decimals. Is it .400, or not? The next day, leading newspapers made
clear their view: It was not. “Williams at .3996,” the New York Times said.
“Williams Drops Under .400,” announced the Chicago Tribune. “Williams Falls
to .399,” the Philadelphia Inquirer reported, bending the rules of rounding,
while Williams’s hometown paper, the Boston Globe echoed that stat: “Average
Now Is .399.”
How can you resist a sport that musters such passion for the fourth decimal
place?

The final two games of the season fell back-to-back on September 28. The
evening before, Williams prowled 10 miles across the streets of Philadelphia,
unable to sleep. Prior to the first game, according to one sportswriter, he “sat on
the bench, biting his fingernails. His mammoth hands trembled.” He later
reported “shaking like a leaf when [he] went to bat the first time.”
But the 23-year-old persevered. That afternoon, he managed six hits in eight
chances, boosting his average to .4057. (The headline writers didn’t quibble
about calling this .406.) Nearly 80 years have elapsed, and no one has hit .400
since.
In 1856, Henry Chadwick glimpsed a dusty, bare-handed game, an excuse for
a summer afternoon. The Englishman gave it numbers. Numbers gave it reach. A
century and a half later, baseball is a corporate triumph, with team payrolls
running in the hundreds of millions. The 19th-century batting average struggles
to keep pace in a 21st-century game, like a boy trying to field a scorching line
drive with his naked hands.
Still, against all the odds, .400 retains its magic. In April and May, when the
season is young and the sample sizes are as small as spring buds, you can often
find a player or two flirting with the .400 line. Before long, they’ll fade. But for
a week or so there’s a breath of hope across the land, a sense that mythical
creatures like dragons and .400 hitters have not yet vanished. The .500 OBP and
the .800 SLG will never stir the heart in quite the same way. We love .400
neither for its predictive power nor for its mathematical elegance, but because it
carries an electric charge, because it tells a story in three decimal places—four, if
you’re feeling picky.
Maybe no one will ever hit .400 again. Maybe it will happen next year. For
his part, Williams shrugged it all off. “If I had known hitting .400 was going to
be such a big deal,” he said 50 years later, “I would have done it again.”
 Chapter 18
BARBARIANS AT THE GATE OF SCIENCE

THE P-VALUE CRISIS

Hooray! It’s fun science fact time!

To begin: Did you know that you’re more likely to cheat on a test after
reading a passage that argues there’s no such thing as free will?

Or that after plotting two close-together points on graph paper, you feel
emotionally closer to your family members than after plotting two far-apart
points?
 Or that adopting certain “power poses” can inhibit stress hormones while
boosting testosterone, prompting others to judge you as more confident and
impressive?

I’m not making these up. They’re real studies, performed by real scientists,
wearing real lab coats and/or jeans. They’re grounded in theory, tested by
experiment, and vetted by peers. The researchers obeyed the scientific method
and hid no rabbits up their sleeves.
Yet these three studies—and dozens more, in fields as far-flung as marketing
and medicine—are now in question. They may be wrong.
 Across the sciences, we are living through an epoch of crisis. After decades
of the best science they could muster, many scholars now find their life’s work
hanging in the balance. The culprit isn’t dishonesty, or a lack of integrity, or too
many passages arguing against free will. The illness runs deeper, all the way
down to a single statistic at the heart of the research process. It’s a figure that
made modern science possible—and that now threatens its stability.

1. TO CATCH A FLUKE
Every science experiment asks a question. Are gravitational waves real? Do
millennials hate solvency? Can this new drug cure antivax paranoia? No matter
the question, there are two possible truths (“yes” and “no”) and, given the
inherent unreliability of evidence, two possible outcomes (“you get it right” and
“you get it wrong”). Thus, experimental results can fall into four categories:
 Scientists want true positives. They are known as “discoveries” and can win
you things like Nobel Prizes, smooches from your romantic partner, and
continued funding.
True negatives are less fun. They’re like thinking you’d tidied the house and
done the laundry, only to realize that, nope, that was just in your head. You’d
rather know the truth, but you wish it were otherwise.
By contrast, false negatives are haunting. They’re like looking for your lost
keys in the right place but somehow not seeing them. You’ll never know how
close you were.
Last is the scariest category of all: false positives. They are, in a word,
“flukes,” falsehoods that, on a good hair day, pass for truths. They wreak havoc
on science, sitting undetected in the research literature for years and spawning
waste-of-time follow-ups. In science’s never-ending quest for truth, it’s
impossible to avoid false positives altogether—but it’s crucial to keep them to a
minimum.
That’s where the p-value comes in. Its whole purpose is to filter out flukes.
To illustrate, let’s run an experiment: does eating chocolate make people
happier? At random, we divide our eager subjects into two groups. Half eat
chocolate bars; half eat graham crackers. Then, all report their happiness on a
scale from 1 (agony) to 5 (bliss). We predict that the chocolate group will score
higher.
 But there’s a danger. Even if chocolate makes no difference, one group is
bound to score a little higher than the other. For example, look what happened
when I generated five random samples from the same population:

Thanks to random chance, two theoretically identical groups may deliver

very different results. What if, by sheer coincidence, our chocolate group scores
higher? How will we distinguish a genuine happy boost from a meaningless
fluke?
To weed out flukes, the p-value incorporates three fundamental factors:
1. The size of the difference. A razor-thin margin (say, 3.3 vs. 3.2) is much
likelier to occur by coincidence than a substantial gap is (say, 4.9 vs. 1.2).

2. How big the data set is. A two-person sample inspires little confidence.
Maybe I happened to give the chocolate to an enthusiastic lover of life,
 and the graham cracker to an ungrateful nihilist. But in a randomly
divided sample of two thousand people, individual differences should
wash out. Even a smallish gap (3.08 vs. 3.01) is unlikely to happen by
fluke.

3. The variance within each group. When the scores are wild and high-
variance, it’s easy for two groups to get different results by fluke. But if
the scores are consistent and low-variance, then even a small difference is
hard to achieve by coincidence.

The p-value boils all of this information down into a single number between
zero and 1, a sort of “coincidence craziness score.” The lower the value, the
crazier it would be for these results to happen by coincidence alone. A p-value
near zero signals a coincidence so crazy that perhaps it isn’t a coincidence at all.
(For a slightly more technical discussion, see the endnotes.)
 Some p-values are easy to interpret: 0.000001 marks a one-in-a-million fluke.
Such coincidences are so rare that the effect is almost certainly real: in this case,
that chocolate makes people happier.
Meanwhile, a p-value of 0.5 marks a 1-in-2 event. Such results happen…
well, half the time. They’re as common as weeds. So, in this case, it would seem
that chocolate makes no difference.
Between these clear-cut cases lies a disputed borderland. What about a p-
value of 0.1? What about 0.01? Do these numbers mark out lucky flukes or
results so extreme that they probably aren’t flukes at all? The lower the p-value,
the better; but how low is low enough?

2. CALIBRATING THE FLUKE FILTER

In 1925, a statistician named R. A. Fisher published a book called Statistical
Methods for Research Workers. In it, he proposed a line in the sand: 0.05. In
other words, let’s filter out 19 of every 20 flukes.
Why let through the other one in 20? Well, you can set the threshold lower
than 5% if you like. Fisher himself was happy to consider 2% or 1%. But this
drive to avoid false positives incurs a new risk: false negatives. The more flukes
you weed out, the more true results get caught in the filter as well.
 Suppose you’re studying whether men are taller than women. Hint: they are.
But what if your sample is a little fluky? What if you happen to pick taller-than-
typical women and shorter-than-typical men, yielding an average difference of
just 1 or 2 inches? Then a strict p-value threshold may reject the result as a fluke,
even though it’s quite genuine.
The number 0.05 represents a compromise, a middle ground between
incarcerating the innocent and letting the guilty walk free.
For his part, Fisher never meant 0.05 as an ironclad rule. In his own career,
he showed an impressive flexibility. Once, in a single paper, he smiled on a p-
value of 0.089 (“some reason to suspect that the distribution… is not wholly
fortuitous”) yet waved off one of 0.093 (“such association, if it exists, is not
strong enough to show up significantly”).
To me, this makes sense. A foolish consistency is the hobgoblin of little
statisticians. If you tell me that after-dinner mints cure bad breath (p = 0.04), I’m
inclined to believe you. If you tell me that after-dinner mints cure osteoporosis
(p = 0.04), I’m less persuaded. I admit that 4% is a low probability. But I judge it
even less likely that science has, for decades, overlooked a powerful connection
between skeletal health and Tic Tacs.
All new evidence must be weighed against existing knowledge. Not all 0.04s
are created equal.
 Scientists get this. But in a field that prides itself on standardization and
objectivity, nuanced case-by-case judgments are hard to defend. And so, as the
20th century wore on, in human sciences like psychology and medicine the 5%
line evolved from “suggestion” to “guideline” to “industry standard.” p =
0.0499? Significant. p = 0.0501? Sorry, better luck next time.
Does this mean that 5% of certified results are flukes? Not quite. The reality
is the reverse: 5% of flukes can become certified results. If that sounds
equivalent, it’s not.
It’s much scarier.
Imagine the p-value as a guardian in the castle of science. It wants to
welcome the true positives inside the walls while repelling the barbarian false
positives at the gate. We know that 5% of barbarians will slip through, but on
balance, this seems good enough.

However, what if the attacking barbarians outnumber our own troops 20 to

one? Five percent of the invading forces will equal the entirety of the civilized
ones.
Worse yet, what if there are a hundred barbarians for every honest soldier?
Their 5% will overwhelm the entire civilized army. The castle will teem with
false positives, while the true ones cower in corners.
 The danger, then, lies in scientists running too many studies where the real
answer is “no.” Does lip-syncing make your hair blonder? Can wearing clown
shoes cause acid rain? Run a million junk studies, and 5% will clear the filter.
That’s 50,000. They’ll flood the scientific journals, splash across headlines, and
make Twitter even less readable than usual.

If that’s not discouraging enough, it gets worse. Without meaning to,

scientists have equipped the barbarians with grappling hooks and battering rams.
 3. HOW FLUKES BREED
In 2006, a psychologist named Kristina Olson began to document a peculiar
bias: Children prefer lucky people to unlucky ones. Olson and her collaborators
found the bias across cultures, from age three to adulthood. It applied to those
suffering little mishaps (like falling in the mud) or big calamities (like
hurricanes). The effect was robust and persistent—a true positive.
Then, in 2008, Olson agreed to advise the senior thesis of a feckless 21-year-
old named “me.” With her copious help, I devised a modest follow-up, exploring
whether five-and eight-year-olds would give more toys to the lucky than the
unlucky.
I tested 46 kids. The answer was “no.”
If anything, the reverse held: my subjects seemed to give more to the unlucky
than the lucky. Far from “fun science fact time,” this felt obvious; of course you
give a toy to someone who lost one. Needing to squeeze 30 pages out of the
experience, I looked to my data. Each subject had answered eight questions, and
I had tested a variety of conditions. Thus, I could carve up the numbers in
several ways:
 Here, in the unassuming columns of a spreadsheet, is where the danger
began.
By all appearances, my thesis was a barbarian at the gates. The crucial p-
value was well above 0.05. But with an open mind, other possibilities emerged.
What if I considered just the five-year-olds? Or just the eight-year-olds? Or just
the lucky recipients? Or just the unlucky ones? Did gender make a difference?
What if eight-year-old girls proved more context-sensitive than five-year-old
boys in giving to children whom they rated at least 4 on the six-point “liking”
scale?
What if, what if, what if…
By slicing and reslicing my data, I could transform one experiment into 20. It
didn’t matter if the p-value rejected my barbarian once, twice, or 10 times. I
could equip it with new disguises until it finally snuck through into the castle.
 Thus is born perhaps the greatest methodological crisis of the young century:
p-hacking. Take a bunch of truth-loving scientists, place them in a winner-take-
all competition to achieve positive results, and then watch as, in spite of
themselves, they act like 21-year-old me, rationalizing dodgy decisions. “Well,
maybe I could run the numbers again…” “I know this result is real; I just need to
exclude those outliers…” “Ooh, if I control for the 7th variable, the p-value
drops to 0.03…” Most research is ambiguous, with a mess of variables and a
multitude of defensible ways to interpret the data. Which are you going to pick:
the method that brands your results “insignificant” or the one that nudges them
below 0.05?
It’s not hard to dredge up this kind of nonsense. At the website Spurious
Correlations, Tyler Vigen combs through thousands of variables to find pairs
that share a close yet coincidental alignment. For example, the number of people
who drowned by falling into a pool, from 1999 to 2009, shows a shocking
correspondence with the number of films featuring Nicolas Cage.
Bombard the fluke filter, and a few false positives are bound to sneak
through.
Wanting to verify this for myself, I split 90 people into three groups. Each
subject got a drink: tap water, bottled water, or a mixture. Then I measured four
variables from each subject: their 100-meter running time, their IQ score, their
height, and their love for Beyoncé. Afterward, I looked at all possible
comparisons. Were the tap-water drinkers faster runners than the bottled-water
ones? Were the bottled-water folks bigger Beyoncé fans than the mixture
drinkers? And so on. The research took me eight months.
Except, of course, it didn’t. I simulated it in a spreadsheet, running it 50 times
in a few minutes.
 Every subject was identical in principle: a collection of random numbers
generated by the same process. Any differences that emerged were necessarily
flukes. Still, with three groups to compare, and four variables for comparison, I
achieved 18 “significant” results in 50 trials.
The p-value wasn’t letting through 1-in-20 flukes. It was more than 1 in 3.

There are other ways to hack a p-value. A 2011 survey found large fractions
of psychologists admitting, under the veil of anonymity, to various “questionable
research practices”:

Even the most innocent-looking of these dogs can bite. Like, say, gathering
more data when your initial results prove inconclusive. Seems harmless, right?
To gauge the power of this p-hack, I simulated a study I call “Who’s the
Better Coin Flipper?” It’s simple: two “people” (read: spreadsheet columns)
each flip 10 coins. Then we check if one of them achieved more heads. In 20
simulations, I achieved a significant result once. That’s exactly as you’d expect
for p = 0.05: 20 trials, one fluke. Reassuring.
Then, I allowed myself the freedom to keep going. Flip another coin. And
another. And another. Stop the study if the p-value dips below 0.05 (or if we
reach 1000 flips with no success).
The results transformed. Now, 12 out of 20 studies yielded a significant
result.
 Such tricks aren’t good science, but they’re not quite fraud, either. In their
paper reporting the survey results, the three authors called such methods “the
steroids of scientific competition, artificially enhancing performance and
producing a kind of arms race in which researchers who strictly play by the rules
are at a competitive disadvantage.”
Is there any way to level the playing field?

4. THE WAR ON FLUKES

The replication crisis has rekindled an old rivalry between two gangs of
statisticians: the frequentists and the Bayesians.
Ever since Fisher, frequentists have reigned. Their statistical models aim for
neutrality and minimalism. No judgment calls. No editorializing. The p-value,
for instance, doesn’t care whether it’s testing a surefire hypothesis or a mad-
scientist one. That kind of subjective analysis comes later.
 Bayesians reject this impartiality. Why should statistics feign indifference
between plausible hypotheses and absurd ones, as if all 0.05’s are created equal?
The Bayesian alternative works something like this. You begin with a
“prior”: an estimate of your hypothesis’s probability. “Mints cure bad breath”?
High. “Mints cure bad bones”? Low. You bake this estimate into the
mathematics itself, via a rule called Bayes’s formula. Then, after you run the
experiment, statistics help you update the prior, weighing the new evidence
against your old knowledge.
Bayesians don’t care whether the results clear some arbitrary fluke filter.
They care whether the data suffice to persuade us, whether they move the needle
on our prior beliefs.
 The Bayesians feel that their time has come. The frequentist regime, they
assert, has led to ruin, and it’s time to inaugurate a new era. Frequentists reply
that priors are too arbitrary, too vulnerable to abuse. They propose their own
reforms, such as lowering the p-value threshold from 0.05 (or 1 in 20) to 0.005
(or 1 in 200).
While statistics deliberates, science isn’t waiting around. Researchers in
psychology have begun the slow, difficult process of eliminating p-hacks. It’s a
litany of protransparency reforms: preregistering studies, listing every variable
measured, and specifying in advance one’s rules for when to stop data collection
and whether to exclude outliers. All of this means that if a false positive
manages to finesse its p-value below 0.05, you can glance at the paper and see
the 19 failed attempts that came before. One expert told me that it’s this suite of
reforms, much more than questions of mathematical philosophy, that will make
the difference.
For what it’s worth, my senior thesis adhered to most of these standards. It
listed all variables gathered, excluded no outliers, and made clear the exploratory
nature of the analysis. Still, when I showed Kristina this chapter, she noted, “It’s
really funny in 2018 to read about a thesis from 2009. Today my thesis students
all preregister their hypotheses, sample sizes, etc. How much we have grown and
learned!”
All of this should slow the stream of barbarians through the gates. Still, that
leaves the matter of the barbarians already inside. To identify those, there is only
one solution: replication.
Imagine 1000 people predict 10 coin tosses. Chances are that one of them
will guess all 10 correctly. Before you start snapping selfies with your newfound
psychic friend, you ought to replicate. Ask this person to predict 10 more flips.
(And perhaps 10 more. And maybe 10 after that.) A true psychic should manage
the feat. A lucky guesser will revert to normal.
The same is true of all positive results. If a finding is true, then rerunning the
experiment should generally yield the same outcome. If it’s false, then the result
will vanish like a mirage.

Replication is slow, unglamorous work. It takes time and money while

producing nothing new or innovative. But psychology knows the stakes and is
beginning to face its demons. One high-profile project, published in 2015,
performed careful replications of 100 psychological studies. The findings made
headlines: 61 of the 100 failed to replicate.
In that grim news, I see progress. The research community is taking a
sobering look in the mirror and owning up to the truth, as ugly as it may be. Now
social psychologists hope that other fields, such as medicine, will follow their
lead.
Science has never been defined by infallibility or superhuman perfection. It
has always been about healthy skepticism, about putting every hypothesis to the
test. In this struggle, is the field of statistics an essential ally. Yes, it has played a
part in bringing science to the brink, but just as surely, it will play a part in
bringing science back.
 Chapter 19
THE SCOREBOARD WARS

WHEN IS A MEASURE NOT A MEASURE?

During my twenties, I taught perhaps 5000 classroom lessons. No more than 15

were observed by other adults. Even in our age of “school accountability,” the
classroom remains an obscure, dim-lit, poorly explored place. I understand why
politicians and outsiders are desperate for a glimpse inside. And I find it
fascinating to watch, through their eyes, as they strain to see the inner workings
of schools—and to influence them.

1. THE BEST TEACHER IN AMERICA

In 1982, Jay Mathews was the Los Angeles bureau chief for the Washington
Post. In theory, that meant covering the biggest stories in the western United
States.
In practice, it meant crashing high school calculus classes.
 Day after day, Mathews couldn’t help dropping by the East Los Angeles
classroom of teacher Jaime Escalante. The Bolivian immigrant was a force of
nature. With insult comedy, tough love, Spanish catchphrases, and relentless
expectations, Escalante helped the students of Garfield High School achieve
unparalleled results on the Advanced Placement Calculus exam. These pupils
weren’t rich or advantaged. Of the 109 that Mathews surveyed, only 35 had a
parent with a high school diploma. Yet in one of the nation’s toughest courses,
they thrived. By 1987, this one school accounted for more than a quarter of all
Mexican Americans passing AP Calculus nationwide. By 1988, Escalante would
be the most famous teacher in the country: name-dropped in a presidential
debate by George H. W. Bush; portrayed in Stand and Deliver by Edward James
Olmos (who got an Oscar nomination for the role); and profiled by Mathews in a
book titled Escalante: The Best Teacher in America.
Apart from the quotient rule, Mathews learned a clear lesson from Escalante.
Students excel when pushed. A good classroom is a challenging one. And so,
hoping to compare schools along this axis, Mathews sought a statistic. Which
schools—regardless of socioeconomics and demographics—are doing the most
to challenge their students?
He rejected “average AP score.” This stat, he believed, spotlighted schools
that limit APs to the strongest few, shutting out the bulk of typical students. It
struck Mathews as extraordinary and perverse that schools would try to block
students from an intellectual challenge. He wanted to measure an AP program’s
reach, not its exclusivity.
 Nor did he want to count “average number of AP exams passed.” In his
understanding, that tended to correlate with socioeconomic status. An AP class
was supposed to leave you better prepared for college than when you began, no
matter whether you passed. The experience outweighed the score.
In the end, Mathews chose something even simpler: the number of AP (and
other college-level) exams taken per graduating student. The scores didn’t
matter, just the attempt. He dubbed it the Challenge Index. A list of top-
performing schools appeared in Newsweek in 1998, and again in 2000, and once
again in 2003, this time as the featured cover story.
 From the start, the rankings courted controversy. One Newsweek reader
called them “sheer mockery.” A professor of education called the list “a
disservice to the thousands of schools in which devoted teachers bust their chops
daily to provide a challenging and appropriate education for millions of young
people who, for many valid reasons, will never see the inside of an AP or IB
classroom.”
It’s been 20 years. The list now appears annually in the Washington Post, and
Mathews stands by his methods. “I rank,” he has written, “with the hope that
people will argue about the list and in the process think about the issues it
raises.”
Maybe this makes me a gullible fish, but I’ll take the bait. I think the
Challenge Index brings up profound questions—not just about our educational
priorities but about the struggle to quantify a messy, multifaceted world. Should
we use complex measures or simple ones? What’s the trade-off between
sophistication and transparency? And most of all, are statistics like the Challenge
Index attempts to measure the world as it is or to transform it into something
new?

2. THE HORROR SHOW OF BAD METRICS

There are two kinds of people in life: those who like crude dualities and those
who do not. And now that I’ve unmasked myself as the first, please allow me to
introduce a statistical distinction I find helpful: windows vs. scoreboards.
 A “window” is a number that offers a glimpse of reality. It does not feed into
any incentive scheme. It cannot earn plaudits or incur punishments. It is a rough,
partial, imperfect thing—yet still useful to the curious observer. Think of a
psychologist asking a subject to rate his happiness on a scale from 1 to 10. This
figure is just a crude simplification; only the most hopeless “1” would believe
that the number is happiness.
Or imagine you’re a global health researcher. It’s not possible to quantify the
physical and mental well-being of every human in a country. Instead, you look
to summary statistics: life expectancy, childhood poverty, Pop-Tarts per capita.
They’re not the whole reality, but they’re valuable windows into it.

The second kind of metric is a “scoreboard.” It reports a definite, final

outcome. It is not a detached observation, but a summary judgment, an incentive
scheme, carrying consequences.
Think of the score in a basketball game. Sure, bad teams sometimes beat
good ones. But call the score a “flawed metric for team quality,” and people will
shoot you the side-eye. You don’t score points to prove your team’s quality; you
improve your team’s quality to score more points. The scoreboard isn’t a rough
measure, but the desired result itself.
Or consider a salesperson’s total revenue. The higher this number, the better
you’ve done your job. End of story.
 A single statistic may serve as window or scoreboard, depending on who’s
looking. As a teacher, I consider test scores to be windows. They gesture at the
truth but can never capture the full scope of mathematical skills (flexibility,
ingenuity, affection for “sine” puns, etc.). For students, however, tests are
scoreboards. They’re not a noisy indicator of a nebulous long-term outcome.
They are the outcome.
Many statistics make valuable windows but dysfunctional scoreboards. For
example, heed the tale of British ambulances. In the late 1990s, the UK
government instituted a clear metric: the percentage of “immediately life-
threatening” calls that paramedics reached within eight minutes. The target:
75%.
Nice window. Awful scoreboard.
First, there was data fudging. Records showed loads of calls answered in
seven minutes and 59 seconds; almost none in eight minutes and one second.
Worse, it incentivized bizarre behavior. Some crews abandoned their
ambulances altogether, riding bicycles through city traffic to meet the eight-
minute target. I’d argue that a special-built patient-transporting truck in nine
minutes is more useful than a bicycle in eight, but the scoreboard disagreed.
Allow me to drive this theme home with a series I call “The Horror Show of
Bad Metrics”:

CLICKS
 Turning back to Mathews and Newsweek, I want to confront the natural
question: what kind of metric is the Challenge Index?

3. WINDOW OR SCOREBOARD?
In his 1998 introduction to the very first index, Mathews wrote:

Nearly every professional educator will tell you that ranking schools is
counterproductive, unscientific, hurtful, and wrong. Every likely criteria
you might use in such an evaluation is going to be narrow and distorted…
I accept all those arguments. Yet as a reporter and a parent, I think that
in some circumstances a ranking system, no matter how limited, can be
useful.

The key word here is “limited.” Schools are characterized by an irreducible

complexity, like ecosystems or daytime soap operas. To capture such
complicated workings with a single metric, you’ve got two basic options: (1)
fuse together many variables into a sophisticated composite score; or (2) just
pick a single clear, easy-to-understand variable.
I’m reminded, like the American Neanderthal I am, of football. One simple
way to measure a quarterback’s performance is completion percentage. How
many of his passes are successfully caught? Most seasons, the league leader
winds up near 70%; the league average is closer to 60%.
Like many windows, completion percentage straddles a line between
“simple” and “simplistic.” It treats a conservative 5-yard pass on par with a
game-changing 50-yarder: both are “caught.” It puts the mild disappointment of
a dropped pass on par with the calamity of an intercepted one: both are “not
caught.” Still, if all statistics are flawed, at least these flaws are transparent. You
can’t accuse “completion percentage” of false advertising.
At the other end of the spectrum is “passer rating,” a dizzying Frankenstat
that incorporates passing attempts, completions, yards, touchdowns, and
interceptions. It runs from a minimum of zero to a maximum of 158⅓. It is
tightly correlated with team victory, and I have never met anyone who claimed
to understand how it is computed or what its blind spots are.
 That’s the trade-off: sophistication vs. transparency. Passer rating vs.
completion percentage. It’s clear to me that Mathews is a “completion
percentage” kind of guy. In introducing the 2009 Newsweek list, he wrote:

One of its strengths is the narrowness of the criteria. Everyone can

understand the simple arithmetic that produces a school’s Challenge
Index rating and discuss it intelligently, as opposed to ranked lists like
U.S. News & World Report’s “America’s Best Colleges,” which has too
many factors for me to comprehend.

All of this casts his Challenge Index as a rough measure: better than nothing,
and even better for wearing its imperfections on its sleeve. An honest window.
But when you publish your statistic in a national news magazine under the
headline “America’s Best High Schools”—well, it starts feeling an awful lot like
a scoreboard.
“The list has taken on a life of its own,” wrote a National Research Council
committee in 2002. “It is now so important to be included among the top 100
schools on the list that some competitive high schools not included have posted
disclaimers on their Web sites indicating why this is the case.”
“The parents are making the most noise,” said a teacher in Milwaukee,
Wisconsin. “Your standing in the community will go up because you’re offering
more AP classes and you might end up on the Newsweek top 100.”
One hallmark of bad scoreboards is that they’re easy to game. In the case of
the Challenge Index, you can push students into AP classes. “Because your
index only considers the number of AP tests taken, and not the actual scores,”
wrote Mathews’s Washington Post colleague Valerie Strauss, “schools put as
many kids into the test pipeline as they could.”

Another problem lay in the fraction. For convenience’s sake, Mathews had
used “graduating seniors” as his denominator, rather than “all students.”
Assuming every student graduates in four years, the math is equivalent. But a
high dropout rate yields a perverse reward. If three students each take an AP
exam and two later drop out, then by Mathews’s math, the remaining graduate
has taken three APs.
 Here, then, is one way to tell the tale of the Challenge Index. It began as a
good window. By considering exams taken rather than exams passed, it looked
beyond wealth and privilege to the deeper question of whether students were
being challenged. Flawless? No. Valuable? Yes.
Then it grew. No longer was it a lone journalist identifying the “most
challenging” schools. Now it was a prominent newsmagazine crowning the
“best.” This created perverse incentives and strange outcomes, turning the good
window into a bad scoreboard.
We could tie a bow on our story right there and retire to watch football and/or
study for AP exams. But that would miss the most interesting twist—and the real
nature of the game that Mathews is playing.

4. UNLEASHING THE TRIBAL PRIMATE

Typically, consumer rankings help to inform specific choices: which car to buy,
which university to apply to, which movie to watch. But it’s unclear this logic
holds for a nationwide ranking of high schools. Am I going to move my family
from Florida to Montana in pursuit of a Newsweek-approved education? Are you
going to consult the stats before deciding between Springfield, Illinois, and
Springfield, Massachusetts? Who, exactly, is the index for?
By Mathews’s own admission, it’s simple. He ranks for ranking’s sake.
“We cannot resist looking at ranked lists,” he has said. “It doesn’t matter
what it is—SUVs, ice-cream stores, football teams, fertilizer dispensers. We
want to see who is on top and who is not.” In 2017, he wrote, “We are all tribal
primates endlessly fascinated with pecking orders.” The Challenge Index aims to
exploit that quirk of primate psychology, to weaponize it in the cause of making
schools more challenging.

Critics call the list easy to game, but Mathews doesn’t mind. In fact, that’s
the whole point: the more students taking the exams, the better. Schools that
urge, cajole, and incentivize aren’t cheating; they’re doing right by their
students. He’s even happy with the title “best,” telling the New York Times that
the word is “a very elastic term in our society.”
As support for his view, Mathews likes to cite a 2002 study of more than
300,000 students in Texas. Focusing on those with low SAT scores, researchers
found that students scoring a 2 on an AP exam (a nonpassing grade) later
outperformed comparable peers who hadn’t attempted the AP at all. The effort
itself—even without a passing score—seemed to lay the groundwork for success
in college.
All of this flips the narrative on its head. Mathews seems to envision the
Challenge Index as a flawed window, as well as the scoreboard the country
needs.
For better or worse, the list’s impact is real. Mathews has always drawn a line
at 1.000—one AP exam per graduating student. In 1998, only 1% of schools
nationwide qualified. As of 2017, that number is 12%. In Washington DC, at the
epicenter of Mathews’s influence (after all, he writes for the Washington Post), it
is over 70%.

To Mathews, the Challenge Index is a pointed attack on a lethargic and

wrongheaded status quo: “the view that schools with lots of rich kids are good,
and schools with lots of poor kids are bad.” He points with pride to high-ranked
schools full of low-income families. He waves off objections—what about the
kids at Eastside High in Gainesville, Florida, many of whom read below grade
level, or the kids at Locke High in Los Angeles, who drop out at alarming rates?
—by saying that these schools deserve recognition for their efforts, not censure
for their struggles.
Every statistic encodes a vision of the world it seeks to measure. In the case
of the Challenge Index, that vision is tinged with memories of Jaime Escalante
and hopes of replicating his approach across the nation. How you feel about
Mathews’s statistic boils down, in the end, to how you feel about his vision.
 Chapter 20
THE BOOK SHREDDERS

There is a beast afoot in the library of life, a chimera known as the Digital
Humanities. It has the body of a literary critic, the head of a statistician, and the
wild hair of Steven Pinker. Some hail it as a burst of light into a dark cavern.
Others scorn it like a slobbering dog with its teeth sunk into a first edition of
Madame Bovary. What does the creature do?
Simple: it turns books into data sets.

1. WHAT COULD CONCEIVABLY GO WRONG?

Last year, I read Ben Blatt’s delightful book Nabokov’s Favorite Word Is
Mauve, which analyzes literature’s great authors via statistical techniques. The
first chapter, titled “Use Sparingly,” explores a cliché of writing advice: avoid
adverbs. Stephen King, for example, once compared them to weeds, and warned
“the road to hell is paved with adverbs.”
So Blatt tallied the rate of -ly adverbs (“firmly,” “furiously,” etc.) in works by
various authors. This is what he found:
 The adverb-friendliness of Austen, one of the language’s finest novelists,
seemed an adequate refutation of that advice. But then Blatt pointed to a funny
pattern. Within a given author’s body of work, the greatest novels often used the
fewest adverbs.
(See the endnotes for how “greatness” was measured.)

F. Scott Fitzgerald’s lowest-adverb novel? The Great Gatsby. Toni

Morrison’s? Beloved. Charles Dickens’s? A Tale of Two Cities, followed by
Great Expectations. Sure, there were exceptions—Nabokov’s adverb rate peaks
in Lolita, arguably his best-regarded novel—but the trend was clear. Low adverb
rates yielded clear, powerful writing. High adverb rates suggested flabby second-
tier stuff.
I was reminded of one day in college, when my roommate Nilesh smiled and
told me, “You know what I love? How often you use the word ‘conceivably.’
It’s one of your signature words.”
I froze. I reflected. And in that moment, the word “conceivably” dropped out
of my vocabulary.

Nilesh mourned the loss for months while I wrestled with the guilt of having
betrayed two friends at once, both the word and the roommate. I couldn’t help it.
The ghost in my brain, the one that turns meanings into words, works by instinct
and thrives in shadow. Calling attention to a particular word choice spooked the
ghost. It retreated.
Presented with Blatt’s statistics, it happened again. I became adverb-
paranoid. Ever since, I’ve been writing like a restless fugitive, afraid those -lys
may slip into my prose like spiders crawling into my sleeping mouth. I recognize
this is a stilted, artificial approach to language, not to mention a naïve
correlation-equals-causation approach to statistics. But I can’t help it. It’s the
promise and the danger of the digital humanities all in a nutshell—with the
emphasis, in my case, on “nut.”
 Taken as a collection of words, literature is a data set of extraordinary
richness. Then again, taken as a collection of words, literature is no longer
literature.
Statistics work by eliminating context. Their search for insight begins with
the annihilation of meaning. As a lover of stats, I’m drawn in. As a lover of
books, I flinch. Is there peace to be made between the rich contextuality of
literature and the cold analytical power of stats? Or are they, as I sometimes fear,
born enemies?

2. STATISTICIANS WILL BE GREETED AS

LIBERATORS
In 2010, 14 scientists (led by Jean-Baptiste Michel and Erez Lieberman Aiden)
published a blockbuster research article titled “Quantitative Analysis of Culture
Using Millions of Digitized Books.” Whenever I read its opening line, I can’t
help whispering, Daaaaaamn. It begins, “We constructed a corpus of digitized
texts containing about 4% of all books ever printed.”
Daaaaaamn.
Like all statistical projects, this required aggressive simplification. The
authors’ first step was to decompose the entire data set—5 million books,
totaling 500 billion words—into what they called “1-grams.” They explained:
“A 1-gram is a string of characters uninterrupted by a space; this includes words
(‘banana,’ ‘SCUBA’) but also numbers (‘3.14159’) and typos (‘excesss’).”
Sentences, paragraphs, thesis statements—they all vanish. Only snippets of
text remain.
 To probe the depths of their data, the authors compiled a list of every 1-gram
with a frequency of at least one per billion. Looking at the start, middle, and end
of the 20th century, their corpus revealed a growing language:

Upon inspection, less than half of the 1-grams from the year 1900 turned out
to be actual words (as opposed to numbers, typos, abbreviations, and the like),
while more than 2/3 of the 1-grams from the year 2000 were words.
Extrapolating from hand-counted samples, the authors estimated the total
number of English words in each year:
 Then, cross-referencing these 1-grams with two popular dictionaries, they
found that lexicographers are struggling to keep pace with the language’s
growth. In particular, the dictionaries miss most of the rarer 1-grams:
 In my own reading, I don’t run into many of those rare nondictionary words.
That’s because… well… such words are rare. But the language is populated by
vast, silent ranks of one-in-a-hundred-million obscurities. Overall, the authors
estimated that “52% of the English lexicon—the majority of the words used in
English books—consists of lexical ‘dark matter’ undocumented in standard
references.” The dictionaries only scratch the surface, missing out on gems like
“slenthem.”
For these researchers, spelunking through the lexicon was a mere warm-up.
The authors went on to examine the evolution of grammar, the trajectories of
fame, the fingerprints of censorship, and the shifting patterns of historical
memory. All this they accomplished in just a dozen pages, mainly by tracking
the frequencies of well-chosen 1-grams.
The paper dropped jaws beyond mine. The journal Science, sensing its
magnitude, made it free to nonsubscribers. “A New Window on Culture,”
declared the New York Times.
Literary scholars tend to study an exclusive “canon,” an elite handful of
authors who enjoy deep, focused analysis. Morrison. Joyce. The cat who sat on
Joyce’s keyboard and produced Finnegans Wake. But this paper gestured toward
another model: an inclusive “corpus,” in which a far-flung population of books,
from the celebrated to the obscure, can all share attention. Statistics could serve
to topple the literary oligarchy and to instate democracy.
Now, there’s no reason both approaches can’t live side by side. Close reading
and statistics. Canon and corpus. Still, phrases like “precise measurement” point
to a conflict. Can the meanings of literature be made “precise”? Can they be
rendered “measurable”? Or will these powerful new tools lead us away from the
hard-to-quantify depths of art, in search of nails that our hammer can strike?

3. THIS SENTENCE IS WRITTEN BY A WOMAN

I tend to think of prose as an androgynous thing. Mine is androgynous like a sea
sponge; Virginia Woolf’s, like a galaxy or a divine revelation. But in Woolf’s A
Room of One’s Own, she takes the opposite view. By 1800, she asserts, the
prevailing literary style had evolved to house a man’s thoughts, rather than a
woman’s. There was something gendered in the pace and pattern of the prose
itself.
That idea rattled around my skull for a few months, until I came across an
online project called Apply Magic Sauce. Among other algorithmic feats, it
could read copy-and-pasted excerpts from your writing, and, via mysterious
analysis, predict your gender.
I had to try it.
 In an internet daze, I spent an hour copying and pasting 25 blog posts, written
from 2013 to 2015. The final results looked like this:

Since the Apply Magic Sauce team guards the secret of its technique, I began
snooping around to figure out how the algorithm might operate. Did it diagram
my sentences? Sniff out the latent patriarchy of my sentiments? Did it infiltrate
my thoughts, the way I imagined Virginia Woolf could have, elevating book
reading into a form of soul reading?
Nope. Chances are that it just looked at word frequencies.
In a 2001 paper titled “Automatically Categorizing Written Texts by Author
Gender,” three researchers managed to achieve 80% accuracy in distinguishing
male from female writers, just by counting the occurrences of a handful of
simple words. A later paper, titled “Gender, Genre, and Writing Style in Formal
Written Texts,” laid out these differences in plain terms. First, men use more
noun determiners (“an,” “the,” “some,” “most”…). Second, women use more
pronouns (“me,” “himself,” “our,” “they”…).
 Even the frequency of the single innocuous word “you” gives a clue about
the author’s gender:

The system’s accuracy is made more impressive by its utter simplicity. The
approach disregards all context, all meaning, to focus on a tiny sliver of word
choices. As Blatt points out, it would evaluate the sentence “This sentence is
written by a woman” as most likely written by a man.
 If you expand your sights to include all words, and not just little grammatical
connectors, then the results take a turn toward the stereotypical. When a data
company called CrowdFlower trained an algorithm to infer the gender of the
owners of Twitter accounts, it spat out the following list of gender-predictive
words:

And in Nabokov’s Favorite Word Is Mauve, Ben Blatt finds the most gender-
indicative words in classic literature:
 It seems that Apply Magic Sauce relies on these sorts of clues, too. When
mathematician Cathy O’Neil tried the algorithm on a man’s writing about
fashion, he scored 99% female. When she checked it on a woman’s writing
about mathematics, she scored 99% male. Three of O’Neil’s own pieces scored
99%, 94%, and 99% male. “That’s not an enormous amount of testing,” she
wrote, “but I’m willing to wager that this model represents a stereotype,
assigning the gender of the writer based on the subject they’ve chosen.”
The inaccuracies don’t quiet my creepy feeling. It seems my maleness so
pervades my thinking that an algorithm can detect it in two nonoverlapping
ways: my pronoun usage and my fondness for Euclid.
I know that, on some level, this vindicates Woolf. She saw that men and
women experience different worlds, and believed that the struggle to give
women a voice had to begin all the way down at the level of the sentence. Crude
statistics bear this out: women write about different topics, and in different ways,
than men do.
Still, I find it all a little depressing. If Woolf’s writing reveals her femaleness,
I like to think it’s embedded in her wisdom and her humor, not in her low
density of noun determiners. To hear Woolf distinguish masculine from
feminine prose feels like going to a trusted doctor. To have an algorithm do the
same feels like getting a pat-down at the airport.

4. THE BUILDING, THE BRICKS, AND THE

MORTAR
The Federalist Papers, written in 1787, helped to define American governance.
They’re full of political wisdom, shrewd argumentation, and timeless
catchphrases (“spectacles of turbulence and contention,” anyone?). They’d make
a killer line on the CV, except for one problem.
The authors didn’t sign their names.
Of the original 77, historians could attribute 43 to Alexander Hamilton, 14 to
James Madison, five to John Jay, and three to joint authorship. But 12 remained
a mystery. Were they Hamilton’s or Madison’s? Almost two centuries later, the
case had long since gone cold.
Come the 1960s, and enter two statisticians: Frederick Mosteller and David
Wallace. Fred and Dave recognized the subtlety of the problem. Hamilton’s
sentences averaged 34.55 words; Madison’s, 34.59. “For some measures,” they
wrote, “the authors are practically twins.” And so they did what good
statisticians do when confronting a vexing problem.
They tore The Federalist Papers to shreds.
Context? Gone. Meaning? Annihilated. As long as The Federalist Papers
remained a collection of foundational texts, they were useless. They had to
become strips of paper, a pile of tendencies—in other words, a data set.
Even then, most words proved useless. Their frequency depended not on the
author, but on the topic. Take “war.” “In discussions of the armed forces the rate
is expected to be high,” Fred and Dave wrote. “In a discussion of voting, low.”
They labeled such words “contextual” and sought at all costs to avoid them.
They were too meaningful.
 Their search for meaninglessness struck gold with “upon,” a word that
Madison almost never used but that Hamilton treated like an all-purpose
seasoning:

Armed with their data, Fred and Dave could reduce each author to something
like a deck of cards, dealing out words at predictable rates. Then, by tallying
word frequencies in the disputed papers, they could infer from which deck the
text had been dealt.
It worked. “The odds are enormously high,” they concluded, “that Madison
wrote the 12 disputed papers.”
In the half century since, their techniques have become standard. They have
helped explore the authorship of ancient Greek prose, Elizabethan sonnets, and
speeches by Ronald Reagan. Ben Blatt applied the algorithm almost 30,000
times, using just 250 common words, to determine which of two authors had
written a given book. His success rate was 99.4%.
My intellect knows there’s nothing wrong with this. Still, my emotions rebel.
How can you understand a book by shredding it to bits?
In 2011, a team of authors at the Stanford Literary Lab attempted a tricky
leap: from fingerprinting authors to fingerprinting genres. They used two
methods: word-frequency analysis and a more sophisticated sentence-level tool
(called Docuscope). To their surprise, both methods converged on accurate genre
judgments.
Take this excerpt, from a passage the computer identified as the most
“Gothic” page in their entire 250-novel corpus:

He passed over loose stones through a sort of court, till he came to the archway;
here he stopped, for fear returned upon him. Resuming his courage, however, he
went on, still endeavouring to follow the way the figure had passed, and
suddenly found himself in an enclosed part of the ruin, whose appearance was
more wild and desolate than any he had yet seen. Seized with unconquerable
apprehension, he was retiring, when the low voice of a distressed person struck
his ear. His heart sunk at the sound, his limbs trembled, and he was utterly
unable to move. The sound which appeared to be the last groan of a dying
person, was repeated…

I find this to be two kinds of creepy. First, there’s the Gothic creepiness of
ruined archways and deathly groans. And second, there’s the creepiness of a
computer detecting the Gothic air without even glancing at the words “archway,”
“ruin,” or “last groan of a dying person.” It had tagged the passage on the basis
of pronouns (“he,” “him,” “his”), auxiliaries (“had,” “was”), and verb
constructions (“struck the,” “heard the”).
I’m unnerved. What does the algorithm know that I don’t?
To my relief, the authors offered a tentative answer. There is no single
element that distinguishes a writer or genre, no unique feature from which all
others follow. Rather, writing has many distinctive features, running from the
galactic structure of the novel down to the molecular structure of the syllable.
Statistical tendencies and larger meanings can coexist, living side by side in the
same sequence of words.
Most of the time, I read for architecture. Plot, theme, character. It’s the high-
level structure: the aspects visible to any passerby but impenetrable to statistics.

If I look closer, I can see the brickwork. Clauses, sentence constructions, the
design of a paragraph. It’s the microlevel structure that my high school English
teachers taught me to scrutinize. Computers can learn to do the same.

And beneath this, hidden from my view, is the mortar. Pronouns,

prepositions, indefinite articles. It’s the nanolevel structure, the limestone
cement that holds it all together, too subtle for my eyes, but ideal for the
chemical analysis of a statistician.

I know it’s just a metaphor, but metaphor is the language that the ghost in my
brain speaks. Heartened, I grabbed the first section of this book (“How to Think
Like a Mathematician”) and sampled the rate of -ly adverbs. It landed at 11 per
1000 words—around the same as Virginia Woolf, which I took as a friendly
omen. Then, unable to help myself, I purged the document of extraneous -lys
until the rate dipped below eight per 1000. That territory belongs to Ernest
Hemingway and Toni Morrison. I was cheating, and it felt great.
Can the new statistical techniques find harmony with the older, richer, more
human ways of understanding language? Yes; conceivably.
 V
ON THE CUSP

THE POWER OF A STEP

The Fitbit-clad among us have a pretty good idea of how many steps they take:
3000 on a sedentary day; 12,000 on an active one; 40,000 if they’re fleeing all
day from a very slow bear. (Maybe just four or five if they’re fleeing from a fast
one.)
This kind of counting masks a truth we all know: Not all steps are created
equal.
 Mathematicians distinguish between two kinds of variables. Continuous
variables can change by any increment, no matter how small. I can drink 1 liter
of diet soda, or 2 liters, or any tooth-dissolving quantity in between. A
skyscraper can loom 300 meters, or 300.1, or 300.0298517. Between any two
values, no matter how close together, you can always split the difference.
Discrete variables, by contrast, move by jumps. You can have one sibling, or
two, but not 1¼. When you buy a pencil, the store can charge 50 cents, or 51
cents, but not 50.43871 cents. With discrete variables, some differences cannot
be split.
 Life is a funny mixture of continuous and discrete variables. Ice cream is a
continuous quantity (giving continual pleasure), and yet it is served in discrete
sizes. The quality of job interviews varies continuously, and yet any application
yields a discrete number of job offers (zero or one). The speed of traffic is
continuous; the speed limit is a single discrete barrier.
This conversion process can amplify tiny increments into huge changes. A
tiny burst of acceleration earns you a speeding ticket. A misplaced hiccup during
a job interview costs you the offer. A desire for just a little more ice cream
forces you, through no fault of your own, to order an 18-scoop bucket called “the
BellyBuster.”
That’s what happens in a world that turns “continuous” into “discrete.” For
every cusp in life, there is a critical turning point—an infinitesimal step with the
power to change everything.
 Chapter 21
THE FINAL SPECK OF DIAMOND DUST

About 250 years ago, the economist Adam Smith asked a question you might
expect from a toddler: why do diamonds cost so much more than water?

He devoted 100 words to posing the puzzle, followed by 13,000 to

developing a theory of prices, by the end of which he still didn’t have a solution.
Put yourself in the shoes of an alien visitor, and you too may find yourself
stumped. What foolish planet would place a lower value on droplets of life-
giving H2O than on hard bits of decorative carbon? Do usefulness and price have
no connection whatsoever? Are humans just irredeemably illogical?
The pursuit of these questions transformed economics. Scholars began the
journey in the holistic spirit of moral philosophers and ended it with the ruthless
rigor of mathematicians. But they did, at last, find an answer. You want to know
why we put a higher price on shiny rocks than on the juice that makes our
kidneys function?
Easy. Just think on the margin.

1. THE TIME HAS COME, LÉON WALRAS SAID, TO

TALK OF MANY THINGS
Classical economics lasted a century, from the 1770s to the 1870s. During that
century, brilliant minds helped to push forward the boundaries of human
understanding. But I do not come to eulogize classical economics; I come to
poke fun at it. Classical economists bit down hard on an idea that, to the modern
palate, tastes like dust and mistakes: the “labor theory of value.”
The theory states that the price of a good is determined by the labor required
to make it.
To tease out this premise, let’s travel to a hunter-gatherer society. Here, it
takes six hours to hunt a deer, and three hours to gather a basket of berries.
According to the labor theory of value, the costs will depend on a single factor:
not scarcity, tastiness, or the current diet fads, but the ratio of required labor. A
deer takes twice as long to obtain as a berry basket, so it will cost twice as much.
 Of course, labor itself is not the only input. What if deer hunting requires
fancy deer-spears (which take four hours to make) while berry gathering requires
only a generic basket (one hour to make)? Then we total it up: in all, a deer
involves 6 + 4 = 10 hours of labor, and a berry basket only 3 + 1 = 4 hours of
labor, so a deer should trade for 2.5 times as much as the berries. Similar
adjustments can account for all inputs, even worker training.
Under this theory, everything is labor, and labor is everything.
In this vision of the economy, the “supply side” sets the price, and the
“demand side” determines the quantity sold. The logic feels quite natural. When
I go to the market to buy deer meat and iPads, I don’t get to pick the prices. The
seller does that. My only choice is whether or not to buy.
Appealing. Intuitive. And, according to today’s best experts, quite wrong.
In the 1870s, economics experienced an intellectual growth spurt. A loose
karass of thinkers across Europe, mulling over the successes and struggles of
their predecessors, came to feel that vague philosophizing was not enough. They
sought to put economics on sturdier foundations. Individual psychology. Careful
empiricism. And most of all, rigorous mathematics.
These new economists wrestled with the question of discrete vs. continuous.
In reality, I can buy one diamond, or two, but nothing in between. We purchase
in discrete chunks.
That’s unfortunate, because it’s much harder, mathematically, to manipulate
quantities that change by leaps and spasms than ones that vary by smooth,
continuous growth. Thus, to ease the path, these new economists assumed that
we could buy any quantity of a product, down to an infinitesimal increment. Not
a whole diamond, but a speck of diamond dust. It’s a simplification, of course:
not true, but powerfully useful.

This unlocked a new kind of analysis: Economists began to think on the

margin. The question became not “What’s the average berry basket worth?” but
“What is one extra berry basket worth?” or better yet “What is the value of a
single extra berry?” It was the dawn of modern economics, a moment now
known as the “marginal revolution.”
Its leaders included William Stanley Jevons, Carl Menger, and an unlikely
hero named Léon Walras, later dubbed by a commentator “greatest of all
economists.” Before achieving that title, he built an eclectic résumé: engineering
student, journalist, railway clerk, bank manager, and Romantic novelist. Then,
one summer night in 1858, he went for a fateful stroll with his father. Old Papa
Walras must have been a persuasive man, because by the end of the evening,
Léon had overcome his wanderlust and set his heart on economics.
Classical economists had tackled big, bold questions about the nature of
markets and society. Marginalists narrowed the focus to individuals making tiny
decisions on the margin. Walras aimed to unify the two levels of analysis, to
build a sweeping vision of the entire economy on a foundation of tiny
mathematical steps.
 2. OF MUFFINS, FARMS, AND COFFEE SHOPS; OF
GENTLY LOADED SPRINGS
It’s role-play time. Congratulations: you’re a farmer!
But I’m afraid you’ll find, as farmers do, that the more land you cultivate, the
less fertile each additional acre becomes.
Land is heterogeneous. When you begin to farm, you first pick the richest,
ripest patches of soil. As you exhaust the best opportunities, each new acre
offers slightly lower returns than the acre before. Eventually, only stretches of
barren rock remain.
 This idea long predates marginalism. One preclassical economist gave a
mechanical analogy:

The earth’s fertility resembles a spring that is being pressed downwards

by the addition of successive weights… After yielding a certain amount…
weights that formerly would have caused a depression of an inch or more
will now scarcely move it by a hair’s breadth.

The marginalist breakthrough came in expanding the same notion from

agriculture to human psychology. For example, just watch me eating cornbread
muffins:
 Perhaps all muffins are created equal, but that’s not what it feels like as I eat
them. The more bites I’ve consumed, the less delightful each additional bite
becomes. Before long, they stop satiating and begin nauseating.
It’s not just food. The same principle holds for all consumer goods: scarves,
SUVs, even Kurt Vonnegut novels. In each case, it’s hard to draw the same
marginal pleasure from your 10th item as you did from your first; the benefit of
a step depends on how many steps you’ve already taken. Economists today dub
this “the law of diminishing marginal utility,” although I prefer to call it “the
reason I didn’t really enjoy Breakfast of Champions.”
How fast is the decline? Well, it varies. As leading marginalist William
Stanley Jevons wrote:

This function of utility is peculiar to each kind of object, and… to each

individual. Thus, the appetite for dry bread is much more rapidly satisfied
than that for wine, for clothes, for handsome furniture, for works of art,
or, finally, for money. And every one has his own peculiar tastes in which
 he is nearly insatiable.

In addition to revealing a lot about William Stanley Jevons, this passage

points toward a new theory of economic demand. People make decisions on the
basis of marginal utility.
Imagine a glorious economy with only two goods: muffins and coffee. How
should I allocate my spending? Well, with each dollar, I ask the same question:
will I draw more joy from adding this to the muffin budget, or the coffee budget?
Even a muffin-loving fool like me will eventually prefer a first coffee to an
eleventh pastry, and even a caffeine-addicted economist will eventually prefer a
first muffin to an eleventh coffee.
This logic boils down to a simple criterion: In a perfect budget, the last
dollar spent on each good will yield exactly the same benefit.

This Walrasian insight marked a new, more psychological foundation for

economics. As Jevons wrote, “A true theory of economy can only be attained by
going back to the great springs of human action—the feelings of pleasure and
pain.” For the first time, scholars understood “the economy” to include not just
the visible ledger of transactions, but the invisible psychology of preference and
desire.
Going further, marginalists judged that sellers function in a similar way to
consumers. For example, say you’re now a coffee shop owner. How many
workers should you hire?
 As your business grows, you draw less and less benefit from each new
employee. Eventually, you hit a point where adding another worker will cost
more in wages than it will bring in coffee sales. At that point, stop hiring.
This logic helps balance the various inputs of a coffee shop. It’s pointless to
buy a dozen espresso machines without hiring enough workers to operate them,
or to pour money into advertising without stocking enough coffee beans to serve
the new customers. How can our budget achieve a sensible harmony among
these inputs? Simple: Keep buying until your last $1 spent on each input yields
$1 of extra productivity. Spend any less, and you’d miss out on possible benefit.
Spend any more, and you’d eat into profit.

In school, I learned that economics has mirror symmetry between consumers

and producers. Consumers buy goods so as to maximize utility. Producers buy
production inputs so as to maximize profits. Each keeps buying until the
marginal benefit of the next item would no longer justify the cost. For this tidy
parallel framework, we can thank (or, depending how you feel about capitalism,
blame) the marginalists.

3. AND WHY A DIAMOND COSTS SO MUCH BUT

WATER, NOT A THING
The old labor theory of prices held half a truth. After all, it does take more labor
to extract diamonds from mines than water from wells. Still, it leaves a key
question unanswered: why would anyone exchange huge amounts of money for
tiny amounts of carbon?
Well, imagine you’re rich. (Perhaps you’ve played this game before.) You’ve
already got all the water you need. Enough to drink, shower, water the petunias,
and maintain the backyard water park. Another $1000 of water? Worthless.
But $1000 of diamond? Novel! Shiny! It’s like your first cup of coffee after
11 muffins.
What’s true for one rich person is even truer in the aggregate. The first
diamond available would find an eager buyer at an absurd price (say, $100,000).
The second diamond would find a slightly less eager buyer (for $99,500). As
you exhaust the most gung-ho buyers, you’ve got to lower the price. The more
diamonds on the market, the lower the utility of the final diamond.
 On the supply side, a parallel principle operates. If the first diamond can be
mined for $500, the next will cost $502, and so on, each additional diamond
costing a bit more to obtain.
 As the market grows, the numbers converge. Inch by inch, the cost of supply
increases. Inch by inch, the utility of consumption drops. Eventually, they meet
at a single price known as “market equilibrium.”
 Yes, the economy’s first cup of water is worth far more than its first shard of
diamond. But prices don’t depend on the first increment, or even the average
increment. They depend on the last increment, on the final speck of diamond
dust.
It’s a beautiful theory. And, like a too-good-to-be-true factoid or a perfect-
skinned model on a magazine cover, it raises a question: Is it real? Do producers
and consumers actually think on the margin?
Well, no. To quote Yoram Bauman, an economics PhD and self-described
stand-up economist: “Nobody goes to a grocery store and says, ‘I’m going to
buy an orange. I’m going to buy another orange. I’m going to buy another
orange…’”
Marginalism doesn’t capture people’s conscious thought processes any more
than the theory of evolution describes the motivations of early mammals. It is an
economic abstraction, throwing out the details of reality to provide a useful,
simplified map. The measure of a theory is its predictive power—and here,
marginalism triumphs. It’s not always conscious, but on some abstract level, we
are all creatures of the margin.
 4. THE MARGINAL REVOLUTION
The marginal revolution marked a point of no return for economics, a historical
cusp. It’s the moment that econ went mathy.
John Stuart Mill, the last great classical economist, had gestured in this
direction. “The proper mathematical analogy is that of an equation,” he wrote.
“Demand and supply… will be made equal.” Take the diamond market. If
demand exceeds supply, then buyers will compete, driving the price up. If
supply exceeds demand, then sellers will compete, driving the price down. In
Mill’s succinct phrasing: “competition equalizes.”
Compelling as Mill’s analysis is, the marginalists dissed it as insufficiently
mathematical. “If economics is to be a real science at all,” Jevons wrote, “it must
not deal merely with analogies; it must reason by real equations.” Or as Walras
says: “Why should we persist in using everyday language to explain things in the
most cumbrous and incorrect way… as John Stuart Mill does repeatedly… when
these same things can be stated far more succinctly, precisely, and clearly in the
language of mathematics?”
 Walras walked the walk. His signature work, Elements of Pure Economics, is
a mathematical tour de force. It builds from clear assumptions to a sweeping
theory of equilibrium. As if to prove his detachment from reality, he devotes
more than 60 pages to the analysis of the simplest economy imaginable: two
people exchanging fixed quantities of two goods. This work, with its
unprecedented rigor and profound abstraction, later earned praise as “the only
work by an economist that will stand comparison with the achievements of
theoretical physics.” Walras would have relished the compliment. His driving
goal, ever since the stroll with his father that one evening, was to elevate
economics to the level of a rigorous science.
More than a century later, it’s worth checking in. How has the mathematical
pivot been working out for Walras and his clan?
For better or worse, the marginal revolution science-ified economics in one
way. It made it less accessible. Adam Smith and the classicists wrote for a
general (if educated) audience. Walras wrote for mathematically skilled
specialists. His vision has pretty much won out: PhD programs in economics
today prefer a student with a degree in math and a little knowledge in economics
to a student with a degree in econ but little mathematical training.
The marginalist revolution can claim another scientific legacy in economics:
renewed empiricism. Researchers today agree that economic ideas can’t just
appeal to intuition or logic; they’ve got to match real-world observations.
Economics still isn’t physics, of course. Human-made systems like markets
don’t obey tidy mathematical laws. At best, they resemble the swirling
complexity of phenomena like the weather and fluid turbulence—precisely the
systems that math still struggles to grasp.
 As marginalism edged out competing systems of thought, it brought yet
another change: Economics, like its brethren in the natural sciences, became
ahistorical. Before marginalism, ideas rose, fell, and recirculated across the
decades. You’d read past thinkers in order to soak in their holistic vision of the
economy, absorbing their worldview through their language. Today, economists
pretty much agree on the foundations and don’t care to read ideas in their
original, archaic formulations. Give them the streamlined modern version, thank
you very much. The marginalists helped economists to stop caring as much
about historical thinkers—including, in a final irony, the marginalists.
 Chapter 22
BRACKETOLOGY

Sometimes you can reveal a misconception just by uttering a phrase. “Soar like
a penguin.” Nope, they don’t fly. “Famous historical Belgians.” Sorry,
Belgium’s most celebrated native is the waffle. “Too full for dessert.” C’mon,
nothing justifies refusing cupcakes. However, my favorite error-in-a-single-
breath is one you’ve probably heard countless times:
“Bumped into the next tax bracket.”
Those six words encapsulate a genuine, widespread, and wholly misguided
fear: What if earning a little extra lifts me into a higher tax bracket? Could I
wind up losing money overall?
 In this chapter, I’ll explain the elementary mathematics behind the income
tax, narrate a brief history of its role in American civic life, and identify which
Disney characters are its most zealous advocates. But first, I want to reassure
you that tax brackets do not go bump in the night. The dreaded “tax bracket
bump” is as fictional as a Belgian celebrity.
Our tale begins in 1861. As the American Civil War looms, the federal
government begins cooking up get-rich-quick schemes, but none of them hold
much promise. The longtime tariff on foreign goods has stopped bringing
adequate funds. A tax on consumer purchases might alienate voters by hitting
poor Americans harder. And a tax targeting rich people’s wealth (including
property, investments, and savings) would violate the Constitution’s ban on
“direct” taxes. What’s a cash-strapped Congress to do?
The only thing they could. In August, the federal government introduced a
temporary emergency tax on income. On any earnings above $800, you now
owed 3%.
The rationale? Money has a diminishing marginal utility. The more dollars
you’ve got, the less another dollar means. Thus, it causes less pain to tax
somebody’s thousandth dollar than to tax somebody else’s first. The resulting
systems, wherein higher earnings face a higher marginal tax rate, are called
“progressive.” (A “regressive” tax draws a larger percentage from those with
lower incomes; a “flat” tax takes the same percentage from everybody.)
You can imagine our American forefathers, scratching their extravagant
beards, fretting whether a small raise might “bump” them from the freebie
bracket to the taxpayer bracket. After all, an income of $799.99 incurred no tax,
but an income of $800.01 did. Could a $0.02 raise really cost you $24 in tax?
 Thankfully, no. Like its modern equivalent, the tax didn’t apply to all income
—just on the margin, to the last dollar earned. Whether you were a poor
Appalachian farmer or Cornelius Vanderbilt (the Bill Gates of the railroad age),
your first $800 was tax-free.
If you made $801, then you’d pay the 3% on the last dollar alone, for a total
liability of just $0.03.
If you made $900, then you’d pay the 3% on the last $100, for a total liability
of just $3.
And so on.
 To visualize this, I like to imagine the government divides your money into
buckets. The initial bucket fits $800 and is labeled “Survival Money.” You fill
this first, and pay no tax on it.
The second bucket is labeled “Leisure Money.” Once the first bucket hits
capacity, the rest of your money goes here. Of this, the government takes 3%.
By 1865, the government had raised the rates and brought in a third bucket:
 Then the war ended. The income tax lay dormant for a few decades, until a
flurry of activity at the end of the century. In 1893, a financial panic wracked the
country. In 1894, the income tax came riding in to save the day. Then in 1895,
the Supreme Court ruled that—ha ha, oops—that ban on “direct” taxes applies to
you too, income tax. You’re unconstitutional. How embarrassing.
It took two decades for a constitutional amendment to revive the tax. Even
then, it returned without flags and fanfare, preferring to sneak back in on tiptoes.
In 1913, the first year of the permanent income tax, only 2% of households paid
it. The marginal rates maxed out at a measly 7%, which didn’t even kick in until
$500,000 (the inflation-adjusted equivalent of $11 million today).
 The rates have a pleasing ABC elegance, climbing from 1% to 7% in a tidy
linear progression, with cutoffs at nice round numbers.
But why these in particular? No reason, honestly.
“Individual judgments will naturally differ,” President Woodrow Wilson
wrote, “with regard to the burden it is fair to lay upon incomes which run above
the usual levels.” No airtight mathematical formula can dictate the “right” rates.
Rate setting is a subjective process, full of guesswork, politicking, and value
judgments.
Of course, arbitrary legal lines are not unique to taxation. In the US, the day
before your 21st birthday, you can’t buy alcohol; the day after, you can. In
theory, the law could ease you in gradually—allow you to buy beer at 19, wine
at 20, and hard liquor at 21—but society has chosen a bright, simple line over a
hard-to-enforce gradient.
The architects of that 1913 income tax went the other way. I find it funny that
they broke such a modest tax into such tiny increments. Their seven brackets
match the number we have today, except ours cover a range of 27%, and theirs
just 6%. It seems that the government didn’t trust the system’s marginal nature,
and thus sought to avoid big “jumps” between rates. Perhaps they were thinking
in terms of this graph:

When you focus on the rates themselves, every bracket transition looks sharp
and jagged. So lawmakers endeavored to smooth the system, to make it appear
safe and gradual.
Wasted energy, if you ask me. Savvy taxpayers don’t care about the abstract
properties of the tax scheme. They care what they’re paying—both overall and
on the margin—and by design, the income tax ensures that earning an extra
dollar will never alter the taxes already paid. The relevant graph is this one:
 Whereas the first graph shows stark jumps, the second shows a single
continuous line. It reflects the shifting rates via its changing slope: nice and
shallow for a low tax rate, harsh and steep for a high one. This better captures
how people actually experience bracket transitions: not as abrupt leaps
(discontinuities, to a mathematician) but as moments where one slope gives way
to another (nondifferentiable points, in calculus-speak).
Evidently, nobody told Congress, because they took their cliff wariness to a
comical extreme in the 1918 tax scheme:
 Looking at this mess, you’ll first notice that the rates have hulked out. In just
five years, the marginal rate on America’s wealthiest has grown elevenfold. I
guess that’s what happens when your country becomes embroiled in something
called “the war to end all wars” (although the change would long outlast the
conflict; the top rate hasn’t dipped below 24% since).
Even more striking—to me, at least—is the sheer number of tax brackets
instituted at that time. The US had more tax brackets than states, 56 to 48.
It reminds me of a favorite project from when I taught precalculus in
California. Every year, I’d ask 11th graders to design their own income tax
system. One industrious fellow named J. J. decided to take the “gradual
transition” idea to its natural conclusion, designing a system where the marginal
rates varied continuously, with no jumps whatsoever.
Imagine you want your marginal rates to start at zero and eventually reach
50% (on income above $1 million). You could do this with two brackets:

Or you could take the World War I–era approach, and break this into 50
brackets:
But why stop there? Make it 1000 brackets.
Or better yet, a million brackets.
 Heck, at the extreme, you could just connect your endpoints with a straight
line.
 This final graph embodies a tax system where every fraction of a penny is
taxed at an infinitesimally higher rate than the penny before. There is no such
thing as “brackets”; each atom of income faces its own unique rate. Everywhere
is a transition, and so, in some paradoxical sense, nowhere is.
Under such a system, the graph of total tax paid will look something like this:
 The slope grows gradually. No nondifferentiable points here. It’s as smooth
as a German car accelerating on the Autobahn.
Look at the income tax through US history, and you’ll witness a similar
acceleration. Over the first half of the 20th century, the tax grew from
“unconstitutional proposal” to “tentative experiment” to “wartime necessity” to
“the government’s primary funding mechanism.” In 1939, at the dawn of World
War II, fewer than 4 million Americans paid income tax. By 1945, more than 40
million did. Revenue showed a similar growth, from $2.2 billion when the war
began to $25.1 billion by its end. Today, its various state and federal forms
amount to $4 trillion per year—almost a quarter of the country’s economy.
As the income tax surged in 1942, the government commissioned Walt
Disney to create a short film to inspire Americans to pay their taxes. The
secretary of treasury demanded a fresh-faced custom-animated protagonist, but
Walt Disney insisted on using the company’s biggest star at the time. Thus, more
than 60 million Americans got to see on their movie screens a display of
patriotic, taxpaying fervor (“Oh boy! Taxes to beat the Axis!”) from none other
than Donald Duck.
 The cartoon must have worked, because two years later, the top marginal tax
rate achieved its historical high-water mark: 94%. Other countries pushed even
higher. In the 1960s, the UK taxed top earners at a marginal rate of 96%,
prompting the Beatles to begin their greatest album with the song “Taxman” and
these scathing lyrics:

Let me tell you how it will be

There’s one for you, nineteen for me

The most extraordinary tale of high marginal tax rates comes from—guess
where?—Sweden. Rather than tell the tale myself, I shall defer to children’s
book author Astrid Lindgren, who in 1976 published a satire of her experiences,
in the evening tabloid Expressen. It tells of a woman named Pomperipossa,
living in a land called Monismania. Some folks in this land complained about the
“oppressive taxes” that funded the welfare state. But Pomperipossa, despite her
83% marginal tax rate, did not. Instead, she contented herself with keeping 17%,
and, “filled with joy,” she “kept skipping down the road of life.”
Now, Pomperipossa (a stand-in for author Lindgren) wrote children’s books.
In the eyes of the government, this made her a “small business owner,”
subjecting her to “social employer fees.” But Pomperipossa did not understand
the implications of this until a friend pointed them out:

“Are you aware of the fact that your marginal tax rate this year is
102%?”
“You are talking nonsense,” Pomperipossa said. “That many percent
does not exist.”
For she was not particularly familiar with Higher Mathematics.

This created an extraordinary scenario. For every dollar that Pomperipossa

earned, she owed the government the entire dollar… plus an additional two
cents. It was like the old nightmare of the “bracket bump” had come to life:
higher earnings, lower wealth. Except instead of a singular cliff, after which
normality returns, Pomperipossa had stepped through the looking glass, and now
faced a never-ending downward slide. The more books she sold, the more
impoverished she became.
 “These terrible little children that are sitting in all the nooks and crannies
of the world… how much money will their disastrous eagerness to read
bring me this year?”… Big fat checks could mercilessly attack her when
she least suspected it.

Of the first $150,000 she earned, she would get to keep $42,000. But above
that lay only heartbreak. Additional earnings would do worse than just vanish;
they would take some of her earlier income along with them. Each additional
$100,000 pretax would reduce her income by $2000 posttax.

At worst, Pomperipossa figured, she’d earn $2 million. If that happened,

she’d be left with a measly $5000. Pomperipossa couldn’t believe it.

She said to herself: “My dear old woman… There are decimal points and
all those things, surely you have counted wrong, there must be 50,000 left
for you.” She started over, but the result did not change one bit… She
now understood that it was something dirty and shameful to write books,
 since it was punished so severely.

The story ends with Pomperipossa, her earnings all sapped by taxes, going on
welfare. The final line: “And she never, ever wrote any books again.”
The figures in the story come directly from Lindgren’s life, and make
Sweden’s top rate today—a mere 67%—look as quaint as a butter churn.
“Pomperipossa in Monismania” caught fire in Sweden, igniting a fierce
debate that helped deal the Social Democratic Party its first electoral defeat in 40
years. Lindgren, a longtime Social Democratic supporter, kept voting for the
party in spite of her grievances.
Back in the US, although the arguments surrounding the income tax are now
a century old, they’re as vibrant (and acrimonious) as ever.
My students’ projects captured pretty much every side of the debate. Some
tweaked the rates upward in the name of redistribution, or downward in the
name of promoting economic growth. One student, who is as clever as he is
cantankerous, designed a regressive system, whereby higher income levels faced
a lower marginal tax rate. He argued that the poor needed an “incentive” to earn
money. Maybe he was sincere, maybe he was satirizing Republican politics, or
maybe he was just trolling me. (He got an A.) Still others sought a radical Robin
Hood system of economic justice and chose top rates close to 100%. A few
students chose rates above 100%, wishing to inflict Pomperipossa’s fate on the
ultrarich. Still other students wanted to step outside the income tax framework
altogether, to envision a whole new regime.
In short, I saw inventiveness everywhere and consensus nowhere. I guess
that’s what they call “America.”
 Chapter 23
ONE STATE, TWO STATE, RED STATE, BLUE STATE

In America, we believe in government of the people, by the people, for the

people. So it’s a shame that the people never seem to agree on anything.

I find it hard enough to build consensus among a family choosing pizza

toppings. And yet somehow, our noisy continent-spanning democracy must
make collective yes-or-no decisions on even graver matters. We go to war, or we
don’t. We elect this president, or that one. We incarcerate Lakers fans, or let
them walk the streets with impunity. How can we blend 300 million voices into
a single national chorus?
Well, it demands some technical work. Taking a census. Totaling votes.
Tabulating. Apportioning. This quantitative labor, if a little dry, is nevertheless
vital. That’s because, at its heart, representative democracy is an act of
mathematics.

The simplest democratic system is called “majority rules.” This creates a

single tipping point, a threshold at 50%. There, one vote can flip you from
“loser” to winner” (or vice versa).
But forget “simple.” This is America! For electing our president, we’ve
crafted a doozy of a mechanism known as the Electoral College, a mathematical
and political oddity that introduces dozens of tipping points into the mix. On top
of fueling our endless talk of “red states” and “blue states,” it endows our
elections with fascinating mathematical properties. In this chapter, I aim to
follow these twists and turns in the story of the Electoral College:

1. It begins as a layer of human beings.

2. It becomes a layer of mathematics.
3. It transforms into a state-by-state “winner-take-all” system.
4. After all that, it behaves pretty much like a popular vote—but with a few
fun surprises.

In the end, understanding the Electoral College comes down to a matter of

marginal analysis. In the 300-million-legged race called US democracy, what is
the meaning of a single step?

1. DEMOCRACY AS A GAME OF TELEPHONE

In the summer of 1787, 55 dudes in wigs met daily in Philadelphia to hammer
out a new plan for national government. Today, we call their design “the
Constitution,” and it sits alongside the cheesesteak at the foundation of our
national character.
The Constitution outlined an elaborate system for determining the
“President” (as in “one who presides”). He would be appointed every four years
by special electors: “men chosen by the people for the special purpose, and at the
particular conjuncture.” I think of the electors as a one-time CEO-search
committee. They convene for a single purpose, select the nation’s new leader,
and then disband.
Where would the electors come from? Well, the states. Each got one elector
per congressman, and could select them by whatever method it pleased.
The logic was that local voters knew only local politicians. In a benighted
time, decades before hashtags, how could citizens judge the oratory and haircuts
of distant, unknown men? That’s why the Electoral College filtered people’s
preferences through three steps: First, you vote for your local state
representative; second, they choose electors (or create a system for their
selection); and third, the electors determine the president.

But by 1832, every state except South Carolina had tossed its electors into the
democratic fray, choosing to let the people decide.
What changed?
 The Philly 55 had envisioned an enlightened democracy of conscience-driven
statesmen, unencumbered by those nasty, rivalrous things called “political
parties.” Then they went out and formed things that looked an awful lot like
political parties.
Hypocrisy aside, this had advantages. Now, you didn’t need to study the
conscience of every single candidate. You could just familiarize yourself with
the party platforms and vote for the team you prefer. No need for an
intermediary.
And so the country moved from human “electors” to abstract “electoral
votes,” the layer of statesmen replaced by a layer of mathematics.
Of course, as with many elements of the Constitution, this system encoded an
advantage for slavery. How did it work? Well, recall that a state got one elector
for each congressman. That included senators (two per state) and House
members (varying in proportion to population; in 1804, they ranged from one to
22 per state).
 The Senate half of the equation gave a boost to small states, by treating little
Rhode Island (fewer than 100,000 people back in the day) on par with big
Massachusetts (more than 400,000). But the real plot twist unfolded over in the
House. The Philly 55 had debated whether, when apportioning House members,
the tally of people represented ought to include slaves. If yes, then the slave-
owning South would get extra representatives; if not, the North stood to benefit.
They compromised on counting them partway. Every five slaves would
register as three people. Hence the most infamous fraction in legal history: 3/5.
The Electoral College imported this proslavery compromise into presidential
elections, like a virus downloaded from an email attachment. Here’s the elector
count from 1800, compared to a hypothetical system where slaves went
uncounted:
 Perhaps a 10-elector swing doesn’t sound like much. It was. Of the country’s
first 36 years, it spent 32 led by slave-owning Virginians. The only interruption
came from John Adams of Massachusetts, who lost a close reelection bid in
1800… So close, in fact, that 10 electors would have swung it.

2. WHY “WINNER-TAKE-ALL” WON AND TOOK

ALL
In his inaugural address after that bitter election of 1800, Thomas Jefferson
struck a conciliatory note. He said that the two parties shared common
principles, common dreams. “We are all Federalists,” he said. “We are all
Republicans.”
But look at the Electoral College today, and you will not see a kumbaya
country of yin-yang balance. You’ll see a map filled with red and blue.
California is all Democrats. Texas is all Republicans. There is no “we,” only
“us” and “them,” like the highest-stakes two-player board game imaginable.
How did the Electoral College get here?
This is where a mathematician’s nose starts twitching. Sure, we’re letting the
people vote, but how will those votes be aggregated and tabulated? What
mathematical process will convert raw preferences into a final choice of
electors?
Say you’re modern-day Minnesota. You’ve got to boil 3 million votes down
to 10 electors. How will you do it?
One option: assign electors in proportion to the actual votes. A candidate
with 60% of the vote gets six electors. A candidate with 20% gets two. And so
on.
 Though logical, this system has gained no traction. In the early days, states
tried some weird stuff—in Tennessee, for example, voters chose county
delegates who chose electors who chose the president—but none saw fit to go
proportional.
An alternative: assign electors by geography. You’ve got 10 electors; why
not break your state into 10 districts, and give one elector to the winner of each?
 This system enjoyed a heyday in the 1790s and early 1800s. It has since gone
extinct.
Its closest living relative is a version where you pick electors by House
district. Then, because each state has two more electors than it has House seats,
you give your final pair of electors to the winner of the statewide vote.
 Today, this funky system is practiced in just two states: Nebraska and Maine.
Nowhere else.
What the heck are the other 48 states doing? They follow a radical scheme:
winner-take-all. Under this approach, the statewide winner gets every elector.
 “Winner-take-all” carries a powerful implication: margin of victory doesn’t
matter. In 2000, George W. Bush won Florida by less than 600 votes. In his
reelection, he won it by 400,000. But the larger margin wasn’t any “better” for
him than the razor-thin one. Winner-take-all collapses the continuous spread of
possible percentages into two discrete outcomes, in a way the Philly 55 neither
specified nor envisioned.
Why, then, is it practiced in 96% of states?
This is a problem straight out of game theory, the mathematics of strategy
selection. To understand the outcome, we’ll need to climb inside the mind of a
state-level politician.
Let us begin in California. Democrats reign in the state legislature. And—
surprise, surprise—Democrats tend to get more presidential votes, too. Imagine
you’ve got two choices: go proportional, or go winner-take-all.
Well, if you’re a Democrat, winner-take-all boosts your party. To do
anything else is to gift a handful of electors to the Republicans. Why even
consider it?
In Texas, the same logic holds, with the color scheme reversed. Winner-take-
all locks up every elector for Republicans. Why relinquish precious electors to
your opponent?

In theory, if all states went proportional, neither party would benefit. It would
be like both sides in a duel laying down their guns. That’s the upper-left box.
But it’s not a stable equilibrium. Once your foe has dropped her weapon, you
might as well grab yours back. If every state does the same, then we soon land in
the lower-right box—which is, of course, where 96% of the country sits today.
Listening to a description of the Electoral College, you’d think that state
borders matter. That to share a driver’s license is to share a special kinship. But
that’s not how state legislators act. By choosing the all-or-nothing system, they
do what’s best for the party, even if it means marginalizing the state.
For example, here’s how the last 10 elections have turned out in Texas:
 Going winner-take-all amounts to posting a big sign at the border that reads,
“Don’t Worry What My Voters Think; We Always Go Republican!” Under
winner-take-all, 55% is as good as 85%, and 45% is no better than 15%. Going
winner-take-all means the election is over before it begins, and so neither party
has any reason to tailor its policies to win over your citizens. Why waste
resources where they will make no difference?
On the other hand, here’s how those elections look if you allocate electors
proportionally:
 Now, modest shifts in the vote leave a real impact on the elector count. If
Texas wants its voters to have a stake in the game, then it ought to go
proportional, so that an extra Texan vote actually boosts a candidate’s chances of
winning. It would give campaigns reason to campaign.
Why doesn’t this happen? Because “maximizing the impact of a marginal
vote in my state” is not what motivates legislators. First and foremost, they are
not Texans, or Californians, or Kansans, or Floridians, or Vermonsters.
They are all Democrats. They are all Republicans.

3. THE PARTISAN TIDES

After more than a century in which the Electoral College always aligned with
the nationwide popular vote, we’ve seen two divergences in the last five
elections. In 2000, the Democrats won the popular vote by 0.5%, while the
Republicans won the Electoral College by five votes. Paper-thin margins on both
fronts. In 2016, the disparity widened: the Democrats won the nationwide vote
by 2.1%, while the Republicans won the Electoral College by 74.
Is the Electoral College now tilted to favor Republicans?
Nate Silver, the modern-day statistical prophet of the Electoral College, has a
beautiful method for answering this question. It allows us to determine the
Electoral College advantage, not just in exceptional years like 2000 and 2016,
but in any election.
The procedure (which I’ll apply to 2012) goes something like this:

1. Line up the states from “reddest” to “bluest.” In 2012, that begins

with Utah (a 48-point Republican victory), then Wyoming, then
Oklahoma, then Idaho… all the way until Vermont, Hawaii, and finally
DC (which went Democratic by 84 points).
2. Go to the middle and identify the “tipping point” state. That’s the
state that pushes the winner past the 270-elector threshold required to
win. In 2012, it was Colorado, which went Democrat by 5.4 points.
3. Gradually lower the winner’s vote share until the election is a virtual
tie. In reality, the Democrat won a comfortable nationwide victory. But in
theory, he could have done 5.4% worse nationwide and still won the
election, carrying Colorado by a single vote. So let’s subtract 5.4% from
his total in every state, to simulate a superclose election.

4. This new, adjusted popular vote tells us the Electoral College

advantage.
 Sometimes, the winning party has enjoyed an advantage that it didn’t need
(e.g., in 2008). Other times, it has managed to overcome an advantage of its
opponent’s (e.g., in 2004). What’s most striking, though, is how the water level
has risen and fallen across the years.
 The last 10 elections have favored five Republicans and five Democrats.
They average out to a Democratic advantage of less than 0.1%. Silver notes that
“there’s almost no correlation between which party has the Electoral College
advantage in one election and which has it four years later. It can bounce back
and forth based on relatively subtle changes in the electorate.”
In this light, it seems that 2000 and 2016 really were flukes. In a nearby
parallel universe, Republicans are fuming that Barack Obama won two elections
in which he lost the popular vote, while Democrats gloat about their electoral
invincibility.
What, then, to make of the Electoral College?
For all its mathematical complexity, it is more or less a randomizer. It echoes
the popular vote, except when (for idiosyncratic reasons) it doesn’t. There’s no
way to predict which direction the partisan advantage will tilt until November is
almost upon us. That being the case, should we ditch the system?
I’m first and foremost a mathematician. That means I love elegance and
simplicity—which the Electoral College does not have—but also wacky
statistical scenarios, which it does.
I’m amused by a proposal called the National Popular Vote Interstate
Compact, or, if you’d rather save nine syllables, the Amar Plan. The idea, crafted
by two law professor brothers, is simple. States commit their electors to the
winner of the nationwide popular vote—regardless of what happens within the
state’s borders. Thus far, the proposal has become law in 10 states and DC,
totaling 165 electors. If enough states join to cross the 270-elector threshold,
they’ll gain control of the Electoral College. (In the meantime, the status quo
reigns, since the laws are written not to take effect until a critical mass signs on.)
The Constitution lets states allocate electors as they wish. If they want a
popular vote, it’s their prerogative. In the peculiar history of the Electoral
College, it would be another marginal step, a strange new cusp of history.
 Chapter 24
THE CHAOS OF HISTORY

You eye the title of this chapter with justified skepticism. “History?” you ask.
“What do you know of history, mathematician?”
I mumble a few incoherencies about walruses, tax law, and some wigs in
Philadelphia; your pity grows.
“Historians find causal patterns in the sprawl of the past,” you explain. “Your
tidy formulas and quaint quantitative models have no place in this messy human
realm.”
I hunch my shoulders and begin to sketch a graph, but you shush me.
“Run home, mathematician!” you say. “Go, before you embarrass yourself!”
Alas, I lost my chance to avoid embarrassment the day I blogged my first
stick figure, and so, in halting tones, I begin to tell my tale.
 1. A NOR’EASTER FROM A ROUNDING ERROR
In the winter of 1961, two surprises hit the East Coast, more or less
simultaneously.
First, in Washington, DC, 8 inches of snow fell on the eve of President
Kennedy’s inauguration. Anxious Southern drivers, possibly interpreting the
snow as a signal of Armageddon, abandoned their cars by the thousands.
Apocalyptic traffic jams ensued. The US Army Corps of Engineers managed to
clear a path for the inaugural parade only by recruiting hundreds of dump trucks
and flamethrowers to the cause.
It was, in short, chaos.
 Second, back in Kennedy’s home state of Massachusetts, a researcher named
Edward Lorenz found something funny. Since the year before, he’d been
developing a computer model of the weather. To begin, you entered some initial
conditions. Then, the computer ran them through a set of equations. In the end, it
delivered a printout of the resulting weather. It could use these results as the
initial conditions for the next day, iterating the process to create months’ worth
of weather from a single starting point.
One day, Lorenz wanted to re-create an earlier weather sequence. One of his
technicians retyped the inputs, rounding them a little (e.g., 0.506127 to 0.506).
The tiny errors—smaller than what weather instruments could even detect—
should have faded into the background. And yet, within weeks of simulated
weather, the new sequence had diverged completely. A tiny adjustment had
created a whole new chain of events.
It was, in short, chaos.
 This moment marked the birth of a new experimental style of mathematics,
an interdisciplinary insurgency that soon became known as “chaos theory.” This
field explored a variety of dynamic systems (gathering storms, turbulent fluids,
populations in flux) with a strange set of shared features. They tended to follow
simple and rigid laws. They were deterministic, with no room for chance or
probability. And yet, because of the subtle interdependence of their parts, they
defied prediction. These systems could amplify small changes into tremendous
cascades, a tiny upstream ripple into an enormous downstream wave.
Lorenz and the nation’s capital were both stunned by the unpredictability of
the weather. But the link between the events runs deeper than this. Forget the
chaos of the snowstorm, and consider instead the fact that John F. Kennedy was
being inaugurated at all.
Three months earlier, he had defeated Richard Nixon in one of the closest
elections in US history. He won the nationwide vote by just 0.17% and carried
the Electoral College thanks to narrow margins in Illinois (9,000 votes) and
Texas (46,000 votes). Half a century later, historians still debate whether voter
fraud might have tipped Kennedy over the edge. (Verdict: Probably not, but who
knows?) It is not hard to imagine a nearby parallel universe where Nixon eked
out a win.
But it is very hard to imagine what would have happened next.
The Bay of Pigs invasion, the Cuban missile crisis, Kennedy’s assassination,
LBJ’s ascension to the presidency, the Civil Rights Act, the Great Society, the
Vietnam War, the Watergate break-in, Billy Joel’s timeless hit “We Didn’t Start
the Fire”… All this and more hinged upon decisions made in the White House.
A 0.2% swing in November of 1960 could have altered the path of world history,
like a rounding error that births a nor’easter.
Ever since I was old enough to notice the world changing, I’ve been
wondering how to conceptualize those changes. Civilization charts a path that’s
unknowable, unforeseeable, and unimaginable except after the fact. How can we
make sense of a system where enormous and untold consequences can issue
from a single indiscernible step?

2. TWO KINDS OF CLOCKWORK

“Ah, silly mathematician,” you say. “You are being as skittish and alarmist as a
DC driver in a slow flurry.”
I gaze at you with wide Disney-sidekick eyes. Hey, I crave a predictable
world as much as the next human.
“Human history is not chaos,” you say. “It exhibits patterns and trends.
Nations come and go. Political systems emerge and fade. Tyrants rise, accrue
Instagram followers, and one day come tumbling down. All of this has happened
before, and all of it will happen again.”
I scratch my head for a moment and then reply with the tale of the pendulum.
 Back at the dawn of the 17th century, when science first turned its
bespectacled eye to the pendulum, it found a mechanism more reliable than any
existing timepiece. The pendulum obeyed a simple equation: measure its length
in meters, take the square root, double this number, and you’ll arrive at the
length of each cycle in seconds. There’s a connection, then, between physical
length and temporal length. A unity between space and time. Pretty cool.
The pendulum is what a mathematician calls “periodic,” which means
“repeating after a regular interval.” It’s like an undulating wave, or the rise and
fall of the tides.
 Sure, the pendulum suffers imperfections—friction, air resistance, fraying
ropes—but these meaningless nudges and background noises don’t spoil its
reliability any more than a breeze unsettles a mountain. By the early 20th
century, the finest pendulum-based clocks kept time accurate to within a second
per year. That’s why, to this day, the pendulum is a lovely mental token of a
well-ordered universe.
But now for a plot twist: the double pendulum.
 It’s just a pendulum attached to another, still governed by the laws of
physics, still described by a set of equations—so it ought to behave like its
cousin, right? And yet, watch. Its swings are wild and erratic. It kicks left,
gestures right, spins like a windmill, pauses to rest, and does it all again, except
different.
What’s going on here? It’s not “random” in the mathematical sense. There
are no cosmic dice, no quantum roulette wheels. It’s a rule-bound, gravity-
governed system. So why is its behavior so wacky? Why can’t we predict its
movements?
Well, in a word: sensitivity.
 Release the double pendulum from one position and record its trajectory.
Then release it from a nearly identical position, just a millimeter away, and
watch as it follows a completely different path. In technical terms, the double
pendulum is “sensitive to initial conditions.” As with the weather, a tiny
perturbation at the start can yield a dramatic transformation by the end. To make
sound predictions, you’d need to measure its initial state with virtually infinite
precision.

Now, does history strike you more as a single pendulum or a double one?
History kicks right; a tyrant topples. It swings left; a war begins. It pauses as
if resting, and a startup scheme hatched in a California garage remakes the world
in its emoji-drenched image. Human civilization is an interconnected system,
hypersensitive to change, with brief intervals of stability and flurries of wild
activity, deterministic yet wholly unpredictable.
In mathematics, an “aperiodic” system may repeat itself, but it does so
without consistency. We’re told that those who do not learn their history are
doomed to repeat it, but perhaps the situation is worse than that. Maybe, no
matter how many books we read and Ken Burns documentaries we watch, we
are all doomed to repeat history—and still, in spite of ourselves, to be caught off
guard, recognizing the repetitions only in hindsight.
 3. THE GAME OF LIFE
“Come now, mathematician,” you cajole. “People just aren’t that complicated.”
I frown.
“Don’t take it personally,” you say. “But you’re not that hard to predict.
Economists can model your financial choices. Psychologists can describe your
cognitive shortcuts. Sociologists can characterize your sense of identity and pick
apart your choice of Tinder photo. Sure, their colleagues in physics and
chemistry may roll their eyes and make vomiting gestures, but the social
scientists can achieve a surprising degree of accuracy. Human behavior is
knowable, and history is the sum of human actions. So shouldn’t it be knowable,
too?”

That’s when I grab a computer and show you the Game of Life.
Like all fun things, it involves a grid. Each square—known as a “cell”—can
assume two states: “alive” or “dead.” In fairness, the word “game” may overstate
things, because Life unfolds from step to step according to four automatic,
immutable rules:
 That’s it. To start the game, just bring a few cells to life. Then, watch as the
board changes, step by step, according to these rules. It’s like a snowball that
needs only a nudge to begin rolling downhill. No further input is required.

In this world, even more so than ours, earning a psychology degree is

supereasy. Just memorize the four rules above, and you can now predict any
cell’s moment-to-moment behavior with flawless accuracy. Good work, doctor.
And yet, the board’s long-term future remains opaque. Clusters of cells
interact in subtle and hard-to-foresee ways. A pattern like the one below can give
rise to never-ending growth; delete just one living cell from the original grid, and
the growth will peter out.
Simplicity on a small scale gives way to strange emergent behavior on a large
one.

I’m reminded of Amos Tversky’s reply to the question of why he became a

psychologist:

The big choices we make [in life] are practically random. The small
 choices probably tell us more about who we are. Which field we go into
may depend on which high school teacher we happen to meet. Who we
marry may depend on who happens to be around at the right time of life.
On the other hand, the small decisions are very systematic. That I became
a psychologist is probably not very revealing. What kind of psychologist I
am may reflect deep traits.

In Tversky’s view, small choices obey predictable causes. But large-scale

events are the products of a devilishly complicated and interconnected system, in
which every motion depends on context.
The same, I contend, is true of human history. A person is predictable; people
are not. Their delicate interrelationships amplify some patterns and extinguish
others, without any clear rhyme or reason.
It’s no accident that the Game of Life, like Lorenz’s weather simulation,
comes from the computer age. By nature, chaos is not the kind of thing you can
hold in your thoughts. Our minds are too inclined toward smoothing things over,
too prone to round off the truth at a convenient decimal place. We needed brains
bigger and faster than ours to make chaos tractable, to reveal its patterns—or its
lack thereof.

4. NOT A BRANCH, BUT A THICKET

“Okay, mathematician,” you sigh. “I see what you’re getting at. What you’re
talking about is alternate history. Envisioning how civilization might have
unfolded, if the past differed in some tiny yet consequential way.”
I shrug.
“Well, you should have said so!” you exclaim, relieved. “Of course history’s
path sometimes forks—there are decisive battles, key elections, pivotal
moments. That doesn’t make history chaotic, just contingent. It’s still subject to
logic and causality. You needn’t treat historical analysis as some kind of lost
cause.”
 Now it’s my turn to sigh and explain that, no, the problem runs deeper.
Start with a bomb—or rather, a pair of them. In 1945, the United States
dropped two atom bombs on Japanese cities: Hiroshima (on August 6) and
Nagasaki (on August 9). Alternate-history stories have wrestled with this
moment. Take Kim Stanley Robinson’s novella The Lucky Strike, in which the
Enola Gay, rather than dropping the bomb on Hiroshima, is destroyed in a freak
accident the day before, changing the course of the nuclear era.
Still, Robinson can tell only a single story, out of the trillions possible. What
if the historic Japanese capital of Kyoto had remained atop the target list, as it
did until late July? What if the August 9 weather over the city of Kokura had
been clear instead of rainy, and the bomb intended for that city were never
rerouted to Nagasaki? What if US president Harry Truman had realized before
the bombing that Hiroshima was a civilian target, not (as he’d somehow
believed) a strictly military one? There are more possibilities than any story can
reckon with. Alternate history gives us branches; chaos warns that these have
been pruned from a sprawling thicket.
 Even at their best, alternate histories can never reckon with the real nature of
chaos. When pivotal importance is latent in every moment, linear storytelling
becomes impossible. Take another of alternate history’s favorite questions: What
if the slaveholding Confederacy had won the American Civil War? It’s a
common topic for speculation; even Winston Churchill, not primarily known for
his science fiction, weighed in. So, when television’s HBO announced its plans
for a show on this premise called Confederate, the writer Ta-Nehisi Coates—an
expert both in African American history and in speculative fiction—issued an
eloquent groan:

Confederate is a shockingly unoriginal idea, especially for the allegedly

avant garde HBO. “What if the South had won?” may well be the most
trod-upon terrain in the field of American alternative history…
[Consider] all the questions that are not being chosen: What if John
Brown had succeeded? What if the Haitian Revolution had spread to the
rest of the Americas? What if black soldiers had been enlisted at the onset
of the Civil War? What if Native Americans had halted the advance of
whites at the Mississippi?

Alternate histories tend to dwell on the “great men” and “important battles”
of conventional history, missing quieter possibilities that cut against the grain of
the dominant culture. The rules of plausibility (what stories we find credible and
compelling) don’t always mirror those of probability (what things actually came
close to happening).
 True chaos is a narrative-destroying idea, a thought as anarchic as a bomb.

5. THE COASTLINE OF WHAT WE KNOW

“Okay then, mathematician,” you say, holding on to your patience by your
fingernails. “You’re saying that historical understanding is a sham. That the
trends of history are aperiodic mirages. That our attempts to infer causality are
doomed, because in a hyperconnected system like human civilization, where
microscopic changes yield macroscopic effects, everything causes everything.
That we can never, ever predict what will happen next.”
“Well, when you put it that way, I sound like kind of a jerk,” I say.
You glare. Point taken.

Maybe human history is what chaos theorists call an “almost-intransitive”

system. It looks pretty stable for long stretches and then, suddenly, shifts.
Colonialism gives way to postcolonialism. Divine monarchy gives way to liberal
democracy. Liberal democracy gives way to corporate-run anarcho-capitalism.
That kind of thing. Even if historians can’t show us what lies beyond the next
cusp, they can at least characterize the shifts that came before and shed light on
the current state of human affairs.
Chaos counsels humility. It teaches us, again and again, the limits of what we
can know.
In 1967, legendary rascal and chaos mathematician Benoit Mandelbrot
published a brief bombshell of a paper titled “How Long Is the Coast of
Britain?” The problem is harder than it sounds, because—weird as it seems—the
length of Britain’s coastline depends on how you measure it.
Start with a ruler 10 kilometers long, and you’ll measure a certain length.
 Then zoom in and switch to a 1-kilometer ruler. Stretches that previously
appeared straight now reveal themselves to be quite jagged. Your shorter ruler
can fit into these nooks and crannies, and so you arrive at a longer total length.

We’re not done. Switch to a 100-meter ruler, and the same process will
repeat. Curves and crinkles overlooked by the longer rulers now become
apparent, increasing the total length.
 We can keep repeating this process. The closer you look, the longer the
coastline becomes—in theory, forever.
Suffice it to say, this is kind of weird. With most inquiries, looking closer
helps clarify the answer. Here, in a frightening inversion, looking closer causes
the problem to unravel. It never simplifies, never resolves.
This trend reaches its extreme with the Koch snowflake, a mathematical
object consisting of bumps upon bumps upon bumps. Although it occupies only
a small patch of the page, the length of its boundary is theoretically infinite.
 In the graphic novel From Hell—a work of speculative history, about the
series of murders in London’s Whitechapel neighborhood in 1888—writer Alan
Moore brings in Koch’s snowflake as a metaphor for the nature of historical
scholarship. “Each new book,” he says in the afterword, “provides fresh details,
finer crenellations of the subject’s edge. Its area, however, can’t extend past the
initial circle: Autumn 1888, Whitechapel.”
In Moore’s telling, there is something bottomless about the study of history.
The further we zoom, the more we see. A finite stretch of time and space can
enclose infinite layers of detail, allow for endless chains of analysis. Chaos is
complexity all the way down, never pixelating, never terminating, never
resolving.
Is history chaotic like the Game of Life—simple on a small scale yet
unpredictable on a large one? Or is its unpredictability like the weather’s—with
wild swings on the small day-to-day scale yet averaging out in the long run to a
stable climate? Or is it perhaps like the Koch snowflake—with chaos on every
level, complexity at every scale? These metaphors compete in my mind, like
three PowerPoint presentations projected onto a single screen. I sometimes feel
I’m on the cusp of figuring it out—and then I check the news, and the world has
changed again, into yet another strange and unknowable shape.
 ACKNOWLEDGMENTS

The elements of this book are a bit like the atoms in my body: only nominally
and temporarily “mine.” They’ve circulated for years, coming from too many
sources to credit or trace. The best I can do is gesture toward the whole
ecosystem that made this book possible.
For the book’s style, I thank all of the witty, kindhearted folks at the Yale
Record, with special hat tips to Davids Klumpp and Litt, and Michael Gerber
and Thornton.
For the book’s perspective, I thank my extraordinary colleagues at King
Edward’s School, with special affection for Tom, Ed, James, Caz, Richard, Nei
—ah, heck, for everybody. Teachers are a playful, critical, inclusive, curious,
and slightly batty tribe that I’m proud to call my people.
For the book’s sense of purpose, I thank my students and my teachers, who
have shaped in uncountable ( 1) ways my thinking about math and about the
world.
For the book’s errors (especially any omitted thank-yous), I apologize in
advance.
For the book’s existence, I thank: the dozens of gracious people who gave
feedback and advice (see the endnotes); Chank Diesel, for elegantly AutoTuning
my handwriting into something radio-ready; Mike Olivo, for educating me about
wookiees; Paul Kepple, for assembling so many bad drawings into a beautiful
whole; Elizabeth Johnson, for bringing peace to the ugly war between myself
and hyphens, and for knowing how many a’s are in “Daaaaaamn”; Betsy
Hulsebosch, Kara Thornton, and the rest of the Black Dog & Leventhal team;
Dado Derviskadic and Steve Troha, for being the first to envision this book—
eons before I could—and for helping me to get there; and Becky Koh, for her
wonderful editing, a job that seems to combine the trickiest elements of
Executive Producing and parenting.
Love and gratitude to my family: Jim, Jenna, Caroline, Lark, Farid, Justin,
Diane, Carl, my happy triangle Soraya, my potato wizard Skander, Peggy, Paul,
Kaya, and the whole Orlin, Hogan, and Williams clans. With fond memories of
Alden, Ros, Pauline, and, of course, Donna.
Finally: thanks, Taryn. You picked math and I’m glad I came along for the
ride. I love you even more than I troll you.
 ENDNOTES
ASIDES, SOURCES, THANK-YOUS, MATHEMATICAL
DETAILS, AND JOKES THAT WERE TOO WEIRD AND
DISRUPTIVE FOR THE TEXT
 Section I: How to Think Like a Mathematician
Chapter 1: Ultimate Tic-Tac-Toe

• Their board looked like this: This game’s origins are murky. It may have first
appeared in Games magazine in the late 1990s or early 2000s (although the folks
at Games hadn’t heard of it when I asked them). In 2009, a wooden board-game
version called Tic-Tac-Ku won an award from Mensa. Perhaps the game had
multiple independent discoveries, like dancing or calculus.
• with my students: When I first showed the game to my students at Oakland
Charter High School in 2012, they took to calling it Ultimate Tic-Tac-Toe. My
2013 blog post with this title seems to have marked an inflection point for the
game’s popularity: Wikipedia, a handful of academic papers, and a larger
handful of phone apps all refer to the game by that name. The obvious
conclusion: take pride, Matadors! You named this thing.
• “what does any of this have to do with math?”: I thank Mike Thornton, who
read an early draft of this chapter and asked exactly this question. Mike’s edits
are like Leonard Cohen’s songwriting or Hemingway’s prose: I always knew
they were good, but the older I get, the more I appreciate them.
• deeper insight into the nature of rectangles: The key idea is that thin
rectangles have disproportionately large perimeters, whereas squarish rectangles
have disproportionately large areas. So just pick a long thin rectangle (e.g., 10 by
1) and a squarish one (e.g., 3 by 4).
• Again, see the endnotes: If you require the solutions to use only whole
numbers, the problem is a lot of fun. Here’s my derivation of a formula that
generates an infinite family of solutions:
 That gives you infinite solutions, but not all of them, because other choices
for d may still yield integer values for c. For example, this formula misses my
favorite solution: 1 by 33 and 11 by 6. My colleague Tim Cross, a dexterous
hand with Diophantine equations, showed me a nifty way to characterize all
possible integer solutions. In the misanthropic tradition of my field, I’ll leave
that as an “exercise for the reader.”
• a guaranteed winning strategy: The actual strategy is a little too complicated
to describe here, but you can see it implemented by some folks from Khan
Academy at https://www.khanacademy.org/computer-programming/in-tic-tac-
toe-ception-perfect/1681243068.
• a clever Brit locked himself in an attic: I recommend the full story: Simon
Singh, Fermat’s Last Theorem (London: Fourth Estate Limited, 1997).
• the great Italian mathematician Cardano: This quote comes from the only
dictionary I’ve ever read for pleasure: David Wells, The Penguin Dictionary of
Curious and Interesting Numbers (London: Penguin Books, 1997).
Chapter 3: What Does Math Look Like to Mathematicians?

• I zip through five chapters of a Twilight novel: I’m more of a Hunger Games
guy, to be honest.
• mathematicians employ strategies: Michael Pershan, the Most Curious and
Analytical Man in the World, articulated this “strategy” idea before I had gotten
there myself. My thanks for his help with this chapter.
• peeling the circle like an onion: There’s a nice animation at
www.geogebra.org/m/WFbyhq9d.
Chapter 4: How Science and Math See Each Other

• forgetting their spouse’s names on occasion: I’ve been married to a

mathematician for five years. So far, she seems to remember me.
• the twisted history of knot theory: For more, see: Matt Parker, Things to
Make and Do in the Fourth Dimension (London: Penguin Random House,
2014).
• a shifting landscape of strange curvatures: Thanks to Matthew Francis and
Andrew Stacey for their help on this point. I wanted to assert that the universe is
“hyperbolic” or “elliptic” as opposed to “Euclidean,” but they informed me that
the real picture is a hard-to-fathom patchwork of these simpler geometries.
Stacey said: “Riemannian geometry generalises [sic; he’s not American, but
don’t hold it against him] Euclidean geometry in many ways and is far richer,
but a few things get lost on the way, primarily anything that relies on saying how
things relate at different points in space.” That includes the concept of “parallel.”
Francis adds this interesting historical wrinkle: “William Kingdon Clifford
proposed using non-Euclidean geometry to replace forces in the 19th century,
but he never got much farther than ‘this would be neat.’ It wouldn’t surprise me
if others were thinking those thoughts too.” Einstein, of course, collaborated
closely with mathematicians; no breakthrough stands alone.
• folks like Bertrand Russell: Experience this story in graphic novel form:
Apostolos Doxiadis et al., Logicomix: An Epic Search for Truth (New York:
Bloomsbury, 2009).
• One logician’s mother: James Gleick, The Information: A History, a Theory,
a Flood (New York: Knopf Doubleday, 2011). Stellar book. This delightful fact
comes from here.
• “unreasonable effectiveness”: Eugene Wigner, “The Unreasonable
Effectiveness of Mathematics in the Natural Sciences,” Richard Courant lecture
in mathematical sciences delivered at New York University, May 11, 1959,
Communications on Pure and Applied Mathematics 13 (1960): 1–14. It’s a
cracking essay.
Chapter 5: Good Mathematician vs. Great Mathematician

• teachers like me: My own teacher for this chapter was David Klumpp, whose
feedback combines the erudition of someone like Carl Sagan with the gentle
humanity of someone like Carl Sagan. (David Klumpp is Carl Sagan.)
• algebra was perhaps the dullest: Israel Kleiner, “Emmy Noether and the
Advent of Abstract Algebra,” A History of Abstract Algebra (Boston:
Birkhäuser, 2007), 91–102, https://link.springer.comchapter10.1007/978-0-
8176-4685-1_6#page-2. I’ve done some real violence to Kleiner’s argument; the
key point is that the 19th century brought big advances in analysis and geometry,
while algebra remained in a more concrete and primordial state.
• a mathematician named Emmy Noether: Joaquin Navarro, Women in Maths:
From Hypatia to Emmy Noether. Everything is Mathematical. (Spain: R.B.A.
Coleccionables, S.A., 2013).
• “She often lapsed”: Professor Grace Shover Quinn, as quoted in: Marlow
Anderson, Victor Katz, and Robin Wilson, Who Gave You the Epsilon? And
Other Tales of Mathematical History (Washington, DC: Mathematical
Association of America, 2009).
• Sylvia Serfaty: All the Serfaty quotes come from her interview with: Siobhan
Roberts, “In Mathematics, ‘You Cannot Be Lied To,’” Quanta Magazine,
February 21, 2017, https://www.quantamagazine.org/sylvia-serfaty-on-
mathematical-truth-and-frustration-20170221. I recommend Roberts’s writing on
math the way I recommend R.E.M.’s best albums: zealously.
• Grothendieck: Colin McLarty, “The Rising Sea: Grothendieck on Simplicity
and Generality,” May 24, 2003,
http://www.landsburg.com/grothendieck/mclarty1.pdf.
• Gaussian correlation inequality: Natalie Wolchover, “A Long-Sought Proof,
Found and Almost Lost,” Quanta Magazine, March 28, 2017,
https://www.quantamagazine.org/statistician-proves-gaussian-correlation-
inequality-20170328. The story is a great read, even after you’ve encountered
my spoilers of it.
• a beloved physics teacher: Farhad Riahi, 1939–2011.
• A student named Vianney: Whereas Corey is a pseudonym, Vianney is not. I
figure she deserves the victory lap. I’ve fudged the dialogue to fill gaps of
memory, but this is pretty much how it went down.
• Ngô B u Châu: The quotes come from a press conference Châu gave at the
2016 Heidelberg Laureate Forum. My enormous thanks to A+ human being
Wylder Green and the HLF team for making my attendance there possible.

Section II: Design

• this will never pan out: The interior angle of a regular pentagon is 108°. If
you try to arrange three around each vertex, you’ll have 36° left over, which isn’t
enough room for a fourth. The only regular polygons that can tile a plane are
triangles (60°), squares (90°), and hexagons (120°), because you need an angle
that goes evenly into 360°.
• Geometry’s laws are not like that: There is some flexibility. In my examples,
I’ve taken Euclid’s assumption about parallels; you could take another. But once
you do, every other rule will follow by logical necessity.
Why relegate this important caveat to an obscure endnote? Well, I figure
anyone finicky enough to question whether we’re in Euclidean space is finicky
enough to read endnotes, too.
Chapter 6: We Built This City on Triangles

• the Great Pyramid of Giza: I should credit here the essential source for this
chapter: Mario Salvadori, Why Buildings Stand Up (New York: W. W. Norton,
1980). It’s a stately, upstanding book, without which this chapter would have
toppled like an erroneous proof. My thanks also to Will Wong, architect of
thought and of intramural sports victories, for his help with this chapter.
• a peculiar loop of rope: I learned about the Egyptian rope pullers from: Kitty
Ferguson, Pythagoras: His Lives and the Legacy of a Rational Universe
(London: Icon Books, 2010).
• frustums: Truncated pyramids with trapezoids for lateral faces. A word worth
knowing.
 • less than 0.1% of its internal volume: Wikipedia lists dimensions for
three passageways (descending, ascending, horizontal) and three rooms (king’s
chamber, queen’s chamber, grand gallery). These volumes totaled to 1340 m3,
which in a structure of 2,600,000 m3 amounts to roughly 0.05%. I rounded this
up to 0.1%, then (with more thanks to Wikipedia) calculated the volume of the
Empire State Building (2,800,000 m3). Taking 0.1% of this, you get 2800 m3.
Dividing by the area of a single floor (roughly 7400 m2), you get a height of 38
centimeters, which is about 15 inches. I rounded up to 2 feet. Then, just as I
completed this manuscript, a hidden chamber was discovered! Still, my rounding
should more than cover the possible error.
• crucial supporting character: Pun 100% intended.
• the beam’s effect: This discussion is pilfered from Salvadori’s Why Buildings
Stand Up, and no doubt loses much in the retelling.
• Architects are no fools: Ted Mosby notwithstanding.
• hence, the term “I-beam”: I borrow again from Salvadori. As Will Wong
points out, a more traditional presentation would foreground the desirable
properties (stress distribution, torque prevention, etc.) created by the cross-
section’s I shape.
• The Pratt truss: My knowledge of trusses derives from that greatest of human
creations, Wikipedia. For more, check out the “Truss” and “Truss Bridge” pages.
Chapter 7: Irrational Paper

• When I moved to England: Thanks to Caroline Gillow and James Butler, two
people with hearts so wide they make the Atlantic look piddling, for their help
and encouragement on this chapter, and for making that “I moved to England”
experience as wonderful as it was.
• My 12-year-old student Aadam: The same day Aadam coined “disintegers,”
an 11-year-old student named Harry responded to my greeting of “Hello,
algebraists!” with this reply: “Why not call us alge-zebras?” I tell ya, teaching is
a good gig.
• all exactly the same proportion: Another delight is that each sheet of A1 has
an area of precisely 1 m2, and each step down the series cuts the area in half.
Thus, eight sheets of A4 will give you precisely a square meter (except not
precisely, because… irrationals).
Chapter 8: The Square-Cube Fables

• Size matters: While I was working on this chapter, my colleague-slash-role-

model Richard Bridges pointed me to a delightful presentation of the same ideas
written 80 years earlier. I went on to borrow liberally from that essay: J. B. S.
Haldane, “On Being the Right Size,” March 1926,
https://irl.cs.ucla.edu/papers/right-size.pdf.
• the only available pan has double the dimensions: I’m ignoring height here,
since when baking brownies, you don’t fill the pan up to the brim.
• local sculptor Chares: I learned this story from Kitty Ferguson’s Pythagoras.
Like all good fables, it is most likely apocryphal.
• It was three-dimensional: John Cowan—whom I thank for fact-checking this
chapter, and for wielding his omniscience in the friendliest, least scary way
possible—adds this wrinkle: “In fact, the Colossus of Rhodes, like the Statue of
Liberty, was hollow. It was made of bronze plates with iron reinforcing rods.
Consequently, as the height grew by n, the cost grew only by n2.” Still too much
for poor Chares.
• There’s no such thing as giants: I first learned this in college in Laurie
Santos’s psychology course Sex, Evolution, and Human Nature. I mean, I
already knew there weren’t giants, but Professor Santos’s explanation of why
(which I now see was perhaps inspired by Haldane) helped form the genesis of
this chapter.
• the overwhelming weight of his torso: Please, call your senators and urge
them to invest in our critical Dwayne Johnson infrastructure before it is too late.
• why bricks plummet and paper flutters: The mathematics of air resistance
underlies another quick fable: “Why Big Sailboats Need Huge Sails.” When you
double your boat’s dimensions, the sails’ area (2D) quadruples, but the ship’s
weight (3D) octuples. You’ll be catching proportionally less wind. Thus, a
double-length ship needs roughly triple-height sails.
• the way I fear them: John Cowan adds: “Another point about ants is that they
don’t bleed, because being so small they are ‘all surface’ and don’t need internal
fluid to distribute oxygen to the interior.” In surface-heavy creatures, diffusion
suffices.
• babies are small: See www.thechalkface.net/resources/baby_surface_area.pdf.
This vignette draws inspiration from a math teacher who goes by the nickname
“The Chalk Face.”
• square-cube fables are everywhere: Here are a few extras that didn’t fit in the
chapter.

#1: Why are big hot-air balloons more cost-effective? Because the canvas
required depends on the surface area (2D), while the lift provided depends
on the volume of helium (3D).
#2: Why do huge dino-birds like turkeys take so long to cook? Because heat
absorption grows with surface area (2D), but heat required grows with
volume (3D).
#3: Why is dry wheat perfectly safe while wheat dust is an explosive hazard?
Because exothermic chemical reactions occur on the surface of a
substance, and tiny dust particles are far more surface-heavy than intact
wheat stalks.

• “paradox of the dark night sky”: I first learned about Olber’s paradox from
Pieter van Dokkum’s, astronomy course Galaxies and Cosmology. If you want
someone to credit (or blame) for this chapter’s eclecticism, look no further than
my liberal arts education.
• brighten you to death: I give John Cowan the final word here: “What if there
are Dark Things (planets, dust, etc.) between us and the stars? Wouldn’t that
make some stars invisible and eliminate the paradox? No, because the stars
behind the Dark Things would over time heat them up to the same temperature
as the stars, eliminating all darkness.”
• Edgar Allan Poe prose poem: E. A. Poe, Eureka: A Prose Poem (New York:
G. P. Putnam, 1848).
 Chapter 9: The Game of Dice

• these Roman emperors: The historical facts in this chapter come from three
sources, listed here in rough order of how much I plundered from them:
• Deborah J. Bennett, Randomness (Cambridge, MA: Harvard University
Press, 1998).
• “Rollin’ Bones: The History of Dice,” Neatorama, August 18, 2014.
Reprinted from the book Uncle John’s Unsinkable Bathroom Reader,
http://www.neatorama.com/2014/08/18/Rollin-Bones-The-History-of-
Dice/.
• Martin Gardner, “Dice,” in Mathematical Magic Show (Washington, DC:
Mathematical Association of America, 1989), 251–62.
• does not make a fair die: The snub disphenoid has two cousins, where every
face is an equilateral triangle yet the die is unfair: (1) the triaugmented triangular
prism (shown to me by Laurence Rackham), and (2) the gyroelongated square
bipyramid (shown to me by Tim Cross and Peter Ollis). If you prefer
quadrilateral faces, there’s (3) the pseudo-deltoidal icositetrahedron (shown to
me by Alexandre Muñiz).

• why aren’t they more popular?: To be fair, there are examples. Ancients in
the Indus Valley threw triangular long dice molded from clay. Their
contemporaries in India rolled rectangular ones, carved from ivory.
• demands a red carpet: This perhaps explains why all the modern long dice
I’ve seen are either (a) tiny, or (b) twisted so as to keep the long faces equivalent
while generating a more satisfying and contained tumble. The Dice Lab, whose
work helped to inspire this chapter, sells some nice instances of the latter.
• mathematical principle: continuity: For more on this line of argument, see:
Persi Diaconis and Joseph B. Keller, “Fair Dice,” American Mathematical
Monthly 96, no. 4 (April 1989): 337–39,
http://statweb.stanford.edu/~cgates/PERSI/papers/fairdice.pdf.
• this approach bugs people: Nevertheless, several civilizations have done
similar. Ancient Inuits rolled chair-shaped ivory dice, on which they counted
only three of the six sides. The Papago Indians threw bison bones, counting only
two of the four sides.
• people try to cheat: For ingenious and evil details, see: John Scarne, Scarne
on Dice, 8th ed. (Chatsworth, CA: Wilshire Book Company, 1992).
• Renumber the faces: A related approach, unsubtle yet elegant, comes from the
gangster Big Jule in the musical Guys and Dolls. His lucky dice are blank, with
no spots on them at all. But don’t worry: Big Jule remembers where all the spots
used to be.
• nothing appears to be wrong: Well… almost. A true expert can recognize a
tap, because from certain angles the faces do not follow the usual configuration,
but rather a mirror image. For this reason and others, taps con artists always
move their cheat dice in and out of play quickly, to avoid detection.
• to roll that d100: As Ralph Morrison, knower of all things trivial and
nontrivial, pointed out to me, the Dice Lab sells a gorgeous 120-sided die. For
more: Siobhan Roberts, “The Dice You Never Knew You Needed,” Elements,
New Yorker, April 26, 2016, https://www.newyorker.com/tech elementsthe-dice-
you-never-knew-you-needed.
Chapter 10: An Oral History of the Death Star

• the team responsible: In reality, I drew heavily from: Ryder Windham, Chris
Reiff, and Chris Trevas, Death Star Owner’s Technical Manual (London:
Haynes, 2013). I wrote this chapter in part to bring a smile to the face of Neil
Shepherd, but don’t tell him I said that.
• architects, engineers, and grand moffs: If you enjoy this kind of thing, I
recommend: Trent Moore, “Death Star Architect Defends Exhaust Ports in
Hilarious Star Wars ‘Open Letter,’” SyFy Wire, February 14, 2014,
http://www.syfy.com/syfywire/death-star-architect-defends-exhaust-ports-
hilarious-star-wars-open-letter%E2%80%99.
• face of a dianoga: My thanks go to Gregor Nazarian, who makes my heart
soar like a John Williams theme, and who helped greatly to improve this chapter.
(I’ve tagged him here because he suggested I use the dianoga, not because his
face resembles one.)
• at a diameter of about 400 kilometers: The rock and ice figures come from
Megan Whewell of the National Space Centre, as quoted by: Jonathan
O’Callaghan, “What Is the Minimum Size a Celestial Body Can Become a
Sphere?” Space Answers, October 20, 2012,
https://www.spaceanswers.com/deep-space/what-is-the-minimum-size-a-
celestial-body-can-become-a-sphere/. The figure for “imperial steel” is my own
blind extrapolation.
• only 140 kilometers across: My research turned up figures ranging from 120
kilometers to 150 kilometers.
• 2.1 million people on the Death Star: I searched and searched to find a more
realistic figure, but the sources all seem to agree. Star Wars fans have debated
this question (https://scifi.stackexchange.com/questions/36238/what-is-the-total-
number-of-people-killed-on-the-2-death-stars-when-they-explode) and tend to
arrive at figures between 1 and 2 million. Windham et al. give roughly 1.2
million people and 400,000 droids. Wikipedia gives 1.7 million people and
400,000 droids. To prove the robustness of my point, I’ve taken the highest
numbers I could find.
• exploiting a thermal vent: I’m not sure that I’ve successfully shoehorned this
narrative into the new continuity established by Rogue One. No doubt certain
stubborn and admirable purists will be upset by the liberties here taken. On the
other hand, I’ve given these characters from “a galaxy far, far away” access to
2010 census data on West Virginia, so maybe redistributing the architectural
credit isn’t my worst crime against canon.

Section III: Probability

Chapter 11: The 10 People You Meet in Line for the Lottery
• People earning at least $90,000 are likelier: Zac Auter, “About Half of
Americans Play State Lotteries,” Gallup News, July 22, 2016,
http://news.gallup.com/poll/193874/half-americans-play-state-lotteries.aspx.
Even so, it’s still true that the lottery is a “regressive tax,” because when a poor
and a rich person spend the same amount of money, the poor person is spending
a larger percentage of their income.
• my home state of Massachusetts: Paige DiFiore, “15 States Where People
Spend the Most on Lotto Tickets,” Credit.com, July 13, 2017,
http://blog.credit.com/2017/07/15-states-where-people-spend-the-most-on-lotto-
tickets-176857/. Though the rankings fluctuate from year to year, Massachusetts
has been at or near the top since my birth in 1987.
• $10,000 Bonus Cash: I drew these odds from MassLottery.com:
http://www.masslottery.com/games/instant/1-dollar/10k-bonus-cash-142-
2017.html.
• Put “10,000” and “bonus” in front of any word: Try it. 10,000 bonus
tortillas; 10,000 bonus fist bumps; 10,000 bonus puppy snuggles. It’s foolproof.
• 20% of our tickets are winners: Almost half of these merely cover the $1
cost of the ticket, so “winner” is perhaps less apt than “nonloser.”
• Poll Americans about why they buy lottery tickets: Charles Clotfelter and
Philip Cook, “On the Economics of State Lotteries,” Journal of Economic
Perspectives 4, no. 4 (Autumn 1990): 105–19,
http://www.walkerd.people.cofc.edu/360/AcademicArticles/ClotfelterCookLottery.pdf.
• when states introduce new lottery games, overall sales rise: Kent Grote and
Victor Matheson, “The Economics of Lotteries: A Survey of the Literature,”
College of the Holy Cross, Department of Economics Faculty Research Series,
Paper No. 11-09, August 2011, http://college.holycross.edu/RePEc/hcx/Grote-
Matheson_LiteratureReview.pdf. Also, I offer my enormous gratitude to Victor
Matheson for taking the time to look this chapter over.
• the UK National Lottery draw offered: Alex Bellos, “There’s Never Been a
Better Day to Play the Lottery, Mathematically Speaking,” Guardian, January 9,
2016, https://www.theguardian.com/science2016jan/09/national-lottery-lotto-
drawing-odds-of-winning-maths.
• 11% of lottery drawings fit the bill: Victor Matheson and Kent Grote, “In
Search of a Fair Bet in the Lottery,” College of the Holy Cross Economics
Department Working Papers. Paper 105, 2004,
http://crossworks.holycross.edu/econ_working_papers/105/.
• from losing too much: For example, a syndicate led by Stefan Klincewicz
bought up 80% of the possible tickets for an Irish National Lottery in 1990. His
team wound up splitting the top prize but turned a profit thanks to the generous
secondary prizes. Klincewicz told reporters that he avoids the UK lottery
because it lacks those smaller prizes. Source: Rebecca Fowler, “How to Make a
Killing on the Lottery,” Independent, January 4, 1996,
http://www.independent.co.uk/news/how-to-make-a-killing-on-the-lottery-
1322272.html.
• the 1992 Virginia State Lottery: I first learned about this story in one of the
best recent pop math books: Jordan Ellenberg, How Not to Be Wrong (New
York: Penguin Books, 2014). I then tracked its trajectory through a trio of
delicious old news stories: (1) “Group Invests $5 Million to Hedge Bets in
Lottery,” New York Times, February 25, 1992,
http://www.nytimes.com/1992/02/25/us/group-invests-5-million-to-hedge-bets-
in-lottery.html; (2) “Group’s Lottery Payout is Postponed in Virginia,” New York
Times, March 7, 1992, http://www.nytimes.com/1992/03/07/us/group-s-lottery-
payout-is-postponed-in-virginia.html; and (3) John F. Harris, “Australians Luck
Out in Va. Lottery,” Washington Post, March 10, 1992,
https://www.washingtonpost.com/archive/politics/1992/03/10 australians-luck-
out-in-va-lotterycbbfbd0c-0c7d-4faa-bf55-95bd6590dc70/?
utm_term=.9d8bd00915e8.
• the late 1600s: Anne L. Murphy, “Lotteries in the 1690s: Investment or
Gamble?” University of Leicester, dissertation research,
http://uhra.herts.ac.uk/bitstream/handle/2299/6283/905632.pdf?sequence=1. I
love the names of 17th-century English lotteries like “the Honest Proposal” and
“the Honourable Undertaking.” They might as well be called “Yes, We Know
We Could Swindle You, but We Promise We Won’t.”
• a fun would-you-rather: If you’re the kind of well-read person who skims the
endnotes, then you surely know it already, but I shall nevertheless cite: Daniel
Kahneman, Thinking Fast and Slow (New York: Farrar, Straus and Giroux,
2011).
• prospect theory: Daniel Kahneman and Amos Tversky, “Prospect Theory: An
Analysis of Decision Under Risk,” Econometrica 47, no. 2 (1979): 263,
http://www.princeton.edu/~kahneman/docs/Publications/prospect_theory.pdf.
• far likelier to be poor: Clotfelter and Cook, “On the Economics of State
Lotteries.”
• a profit of $30 billion per year: Derek Thompson, “Lotteries: America’s $70
Billion Shame,” Atlantic, May 11, 2015,
https://www.theatlantic.com/business/archive/2015/05/lotteries-americas-70-
billion-shame/392870/. See also: Mona Chalabi, “What Percentage of State
Lottery Money Goes to the State?” FiveThirtyEight, November 10, 2014,
https://fivethirtyeight.com/features/what-percentage-of-state-lottery-money-
goes-to-the-state/.
• I suspect it’s hypocrisy: Back in the 1790s, French revolutionaries considered
the lottery an evil excess of the monarchic state. But they hesitated to abolish it
after seizing power, for the simple reason that they needed the money. How else
can you turn a taxphobic citizenry into dutiful taxpayers? Source: Gerald
Willmann, “The History of Lotteries,” Department of Economics, Stanford
University, August 3, 1999, http://willmann.com/~gerald/history.pdf.
• rip-off of a popular illegal game: Clotfelter and Cook.
• post-USSR Russians: Willmann, “The History of Lotteries.”
• so little in winnings: Bingo averages a payout of $0.74 in winnings for every
dollar spent. Horseracing pays $0.81. Slot machines pay $0.89. State lotteries
average around $0.50. Source: Clotfelter and Cook.
• This tradition goes way back: Willmann.
• offset, dollar for dollar: Grote and Matheson, “The Economics of Lotteries.”
• John Oliver quips: “The Lottery,” Last Week Tonight with John Oliver, HBO,
published to YouTube on November 9, 2014, https://www.youtube.com/watch?
v=9PK-netuhHA.
• a chance to fantasize about winning money: Why do middle-aged people
buy lottery tickets at higher rates? I’m just spitballing, but perhaps it’s because
middle age is the best time to be a Dreamer. Young adults can imagine other
paths to fortune. Older adults are beyond such longings. Only the middle-aged
are old enough to realize that no magic financial transformation is forthcoming,
but young enough to crave it.
• winning a lottery jackpot tends to make people less happy: For a full and
up-to-date discussion, see: Bourree Lam, “What Becomes of Lottery Winners?”
Atlantic, January 12, 2016, https://www.theatlantic.com/business/archive
201601/lottery-winners-research/423543/.
• next socioeconomic stratum: See, for example: Milton Friedman and L. J.
Savage, “The Utility Analysis of Choices Involving Risk,” Journal of Political
Economy 56, no. 4 (August 1948): 279–304,
http://www2.econ.iastate.edu/classes/econ642/babcock/friedman%20and%20savage.pdf.
• instant games, with their modest top prizes: Grote and Matheson, “The
Economics of Lotteries.”
• odd economy of scale: My example is adapted from Clotfelter and Cook.
Chapter 12: Children of the Coin

• I was pressed into teaching 10th-grade biology: Witness this typical scene
from April 2010:
KISA (with bright-eyed curiosity): What exactly happens in the endoplasmic
reticulum?
ME: There’s no way of knowing. It is an unsolvable enigma, beyond human
imagination.
TIM (in a bored monotone): The textbook says that’s where proteins are
folded.
ME: Well, obviously, Tim. I meant aside from that.
• The riddles of resemblance strike at the heart of biology: For a take that’s
more sophisticated than mine but still plenty readable, see: Razib Khan, “Why
Siblings Differ Differently,” Gene Expression, Discover, February 3, 2011,
http://blogs.discovermagazine.com/gnxp/2011/02/why-siblings-differ-
differently/#.Wk7hKGinHOi.
• unfolding in very few ways: This same logic underlies the concept of entropy,
the universe’s tendency toward disorder.
 Take a collection of bricks. There are precious few ways to form a building,
and many dull, indistinguishable ways to form a pile of rubble. Over time,
random changes will accumulate, almost all of them making your arrangement
more rubble-y, and almost none making it more building-y. Thus, time favors
rubble over bricks.
Similarly, there are very few ways for the particles of food dye to gather on
one side of a water glass; it’d be like the molecules all flipping heads. But there
are many, many ways for those particles to scatter more or less evenly through
the liquid; each such scattering is like a different combination of heads and tails.
That’s why random processes lead in erratic but inexorable ways toward greater
entropy, an even mixing of the universe’s ingredients. The cosmic preference for
disorder is, at its heart, combinatorial.

• between 16 and 30 heads: This has a probability around 96%, so for one in 25
readers, my prophecy will prove false. That said, if as many as 25 readers
actually go flip 46 coins, then the math-book-reading audience is even more
hard-core than I suspected.
• Blaine Bettinger: The graph here comes straight from him. See more at:
https://thegeneticgenealogist.com/wp-content/uploads201606/Shared-cM-
Project-Version-2-UPDATED-1.pdf.
The x-axis is labeled “centimorgans,” which are (to me, anyway) science’s
most confusing unit. A centimorgan is a length of chromosome that has a 1%
chance of being broken up by a chromosomal crossover in any given generation.
With close relatives, you’ll share lots; with distant relatives, very few. Thus,
“centimorgans shared” is a measure of genetic proximity.
So far, so good. But because crossing over happens with different
probabilities all over the genome, a centimorgan is not a constant length. Where
crossing over is common, centimorgans are short; where it’s rare, they’re long.
On top of that, different DNA-sequencing companies divide the human genome
into different numbers of centimorgans. And did I mention that 100
centimorgans do not make a morgan?
A further puzzle: when I converted the centimorgans in this graph to
percentages, I found the distribution centered not around 50%, but 75%. Why?
My wife, Taryn, explained: It’s because commercial DNA kits can’t distinguish
between having both chromosomes in common and having just one. Each counts
as “matching.” By the logic of coin flips, two siblings will tend to have 50% of
their DNA single-matching, 25% double-matching, and 25% not matching at all.
Thus, 75% will match singly or doubly. Hence, the graph centered on 75%.
• roughly twice per chromosome: The number I found was 1.6. It’s higher for
women than for men. In any case, my tripling is a gross underestimate of
possible genomes, because crossing over can happen (in theory) at any point
along the DNA sequence, which allows for myriad additional possibilities. For
more, see Ron Milo and Rob Phillips, “What Is the Rate of Recombination?”
Cell Biology by the Numbers, http://book.bionumbers.org/what-is-the-rate-of-
recombination/.
Chapter 13: What Does Probability Mean in Your Profession?

• Actual Probability vs. How People Treat It: Adapted from Kahneman,
Thinking Fast and Slow, 315.
• Donald Trump defeated Hillary Clinton: For more on this outcome, see
99.997% of all content published on social media since.
• your job was pretty boring: Michael Lewis, Liar’s Poker: Rising Through the
Wreckage on Wall Street (New York: W. W. Norton, 1989).
• You’re increasing the odds: Nate Silver, The Signal and the Noise: Why So
Many Predictions Fail—but Some Don’t (New York: Penguin Books, 2012),
135–37.
Chapter 14: Weird Insurance

• insurance is a strange and silly thing: For this chapter, I drew tremendous
help from Shizhou Chen, my former 10th-grade student and current intellectual
superior. The original draft had a more presumptuous title and a less self-aware
intro; Shizhou cut me down to size. “Quirky and funny examples,” she wrote,
“but not exactly the full picture of what insurance is about.” Touché, Shizhou.
• China five millennia ago: Emmett J. Vaughan, Risk Management (Hoboken,
NJ: John Wiley & Sons, 1996), 5.
• “Iran, 400 BCE”: Mohammad Sadegh Nazmi Afshar, “Insurance in Ancient
Iran,” Gardeshgary, Quarterly Magazine 4, no. 12 (Spring 2002): 14–16,
https://web.archive.org/web/20080404093756/; http://www.iran-
law.com/article.php3?id_article=61.
• National Lottery Contingency Scheme: “Lottery Insurance,” This Is Money,
July 17, 1999, http://www.thisismoney.co.uk/money/mortgageshome/article-
1580017/Lottery-insurance.html.
• What might surprise you is the margin: Shizhou makes a good point here: In
niche insurance markets like this, there’s less competition, and thus higher
markups than in big markets like dental or home insurance.
• if you can’t insure ’em, join ’em: Shizhou makes this astute addendum:
“Small business owner? Yup. Large firm? Totally illegal. The accounting
department would rather put $1 million under ‘insurance’ than $5 under
‘lottery.’”
• “multiple-birth insurance”: Laura Harding and Julian Knight, “A Comfort
Blanket to Cling to in Case You’re Carrying Twins,” Independent, April 12,
2008, http://www.independent.co.uk/money/insurance/a-comfort-blanket-to-
cling-to-in-case-youre-carrying-twins-808328.html. The quote from David Kuo
also comes from this piece.
• financial and psychological: For similar analysis, see: “Insurance: A Tax on
People Who Are Bad at Math?” Mr. Money Mustache, June 2, 2011,
https://www.mrmoneymustache.com/2011/06/02/insurance-a-tax-on-people-
who-are-bad-at-math/. As the Mustache man writes: “Insurance of all types—
car, house, jewelry, health, life—is a crazy field swayed by lots of marketing,
fear, and doubt.”
• alien-abduction policy: Check it out at http://www.ufo2001.com. My source:
Vicki Haddock, “Don’t Sweat Alien Threat,” San Francisco Examiner, October
18, 1998, http://www.sfgate.com/news/article/Don-t-sweat-alien-threat-
3063424.php.
• selling abduction coverage in earnest: Teresa Hunter, “Do You Really Need
Alien Insurance?” Telegraph, June 28, 2000,
http://www.telegraph.co.uk/finance/4456101/Do-you-really-need-alien-
insurance.html.
• the following criteria: These are my own invention. Shizhou lent me her notes
from an undergraduate module, and the list of “characteristics of an insurable
risk” differed slightly:
1. “Potential loss is significant enough that people would exchange
premium for coverage.”
2. “Loss and its economic value is well-defined and out of the
policyholder’s control.”
3. “Covered losses reasonably independent among policyholders.”
• Hole-in-One Prize Insurance: Two such companies are Insurevents
(http://www.insurevents.com/prize.htm) and National Hole-in-One
(http://holeinoneinsurance.co.uk).
• Jordan’s Furniture ran a glitzy promotion: Scott Mayerowitz, “After Sox
Win, Sofas Are Free,” ABC News, October 29, 2007,
http://abcnews.go.com/Business/PersonalFinance/story?id=3771803&page=1.
• the company Wedsure: See its policies at http://www.wedsure.com.
• reimbursing for weddings canceled: Haddock, “Don’t Sweat Alien Threat.”
• began offering to reimburse third-party: Amy Sohn, “You’ve Canceled the
Wedding, Now the Aftermath,” New York Times, May 19, 2016,
https://www.nytimes.com201605/22/fashion/weddings /canceled-weddings-
what-to-do.html.
• Become so expert in the risk: This helps broaden your business, too, by
allowing you to serve as a sort of risk consultant. “Special expertise is one of the
reasons firms buy insurance,” Shizhou told me. “When filming movies, there’s
always an insurance inspector to make sure actors are safe. Without them,
movies would have more mad and pointless explosions than what we see today.”
• “reinsurance”: Olufemi Ayankoya, “The Relevance of Mathematics in
Insurance Industry,” paper presented February 2015.
• insure against career-ending injury: “Loss-of-Value White Paper: Insurance
Programs to Protect Future Earnings,” NCAA.org,
http://www.ncaa.org/about/resources/insurance/loss-value-white-paper.
• purchasing an additional plan: Andy Staples, “Man Coverage: How Loss-of-
Value Policies Work and Why They’re Becoming More Common,”
SportsIllustrated.com, January 18, 2016, https: //www.si.com/college-
football201601/18/why-loss-value-insurance-policies-becoming-more-common.
• Jake Butt: A great name, right? Will Brinson, “2017 NFL Draft: Jake Butt
Goes to Broncos, Reportedly Gets $500K Insurance Payday,” CBS Sports, April
29, 2017, https://www.cbssports.com/nfl/news/2017-nfl-draft-jake-butt-goes-to-
broncos-reportedly-gets-500k-insurance-payday/.
• health insurance: You could fill a bookshelf with this stuff. I recommend the
work of Vox journalist Sarah Kliff, https://www.vox.com/authors/sarah-kliff.
Shizhou recommends the two-part This American Life episode: “More Is Less”
(#391), and “Someone Else’s Money” (#392),
http://hw3.thisamericanlife.org/archive/favorites/topical.
Chapter 15: How to Break the Economy with a Single Pair of Dice

• What is the basic activity of Wall Street banks?: Far and away the most
essential source for this chapter was: David Orrell and Paul Wilmott, The Money
Formula: Dodgy Finance, Pseudo Science, and How Mathematicians Took Over
the Markets (Hoboken, NJ: John Wiley & Sons, 2017).
• make educated guesses: Okay, in my case, it’s more like “make up random
probabilities.” Wall Street employs two more serious approaches. First, they
anchor on historical data. And second, they look at the market price for similar
bonds and use this to infer what the probability of default must be. This latter
method can create spooky dependencies and feedback loops: Instead of
exercising your own judgment, you echo the perceived wisdom of the market.
Put a pin in that idea.
• collateralized debt obligations: Michael Lewis, The Big Short: Inside the
Doomsday Machine (New York: W. W. Norton, 2010). Or, if you prefer
listening to reading: “The Giant Pool of Money,” This American Life, episode
#355, May 9, 2008A. This episode launched the essential Planet Money podcast.
• surrealist painter René Magritte: I saw these sketches in Belgium’s lovely
Musée Magritte on June 4, 2017. I recommend a visit if you’re in Brussels and
would like some surrealism to go with your frites.
• an idiosyncratic risk or a systematic one: These are the standard finance
terms, brought to my attention by Jessica Jeffers, to whom I owe a huge thank-
you for her help with this chapter. I am only 37% joking when I say that Jess is
my choice for Fed chair.
• the notorious “Gaussian copula”: Another essential source for this chapter,
and this discussion in particular, is: Felix Salmon, “Recipe for Disaster: The
Formula That Killed Wall Street.” Wired, February 23, 2009,
https://www.wired.com/2009/02 wp-quant.
• CDS. It stood for “credit default swap”: Also, “complete damn stupidity.”
• How could Wall Street be so stupid?: See another essential source for this
chapter: Keith Hennessey, Douglas Holtz-Eakin, and Bill Thomas, “Dissenting
Statement,” Financial Crisis Inquiry Commission, January 2011, https://fcic-
static.law.stanford.edu/cdn_media/fcic-reports/fcic_final reporthennessey_holtz-
eakin_thomas_dissent.pdf.
• “the wisdom of the crowd”: James Surowiecki, The Wisdom of Crowds: Why
the Many Are Smarter than the Few and How Collective Wisdom Shapes
Business, Economies, Societies, and Nations (New York: Anchor Books, 2004).
• 1987 stock crash: Orrell and Wilmott, 54.

Section IV: Statistics

• A survey of medical professionals: Kahneman, Thinking Fast and Slow, 228.
Chapter 16: Why Not to Trust Statistics

• citizens that H. G. Wells envisioned: Thanks to Richard Bridges for (1) his
help with this chapter; and (2) being a Platonist, a pragmatist, a teacher, and a
brilliant mind, and for proving by construction that all those things can coexist.
• Take household income: All data from Wikipedia. For you, dear reader, only
the best.
• the tasty £2-per-jar tomato sauce: Loyd Grossman’s. They make a solid jar
of tikka masala, too.
• take a square root at the end: My students tend to find it peculiar (and
needlessly complicated) to square the distances, average them, and then take the
square root. Why not just average the distances and leave it at that? Well, you’re
welcome to do this; the result is called the “mean absolute deviation,” and it
serves much the same role as standard deviation. However, it lacks some nice
theoretical properties. Variances can be added and multiplied with ease; this
makes them vital for building statistical models.
• how the correlation coefficient actually works: Okay, buckle up! Things are
about to get hectical, which is my portmanteau of “hectic” and “technical.” To
begin, take a scatter plot—say, height against weight. Represent each person as a
dot.

Now, find the average height and average weight in the population.
 Next, take an individual person. How far is that person from the average
height? How far from the average weight? Count “above average” as positive
and “below average” as negative.
 Then—and this is the crucial step—multiply these two values.
If the person is above average in both, you’ll get a positive result. Same if
they’re below average in both (because the product of two negatives is positive).
But if they’re above average in one and below average in the other, then you’ll
get a negative result (because the product of a negative and a positive is
negative).
 You do this for each and every person, then average together all the products.
The result is something called “covariance” (a cousin of the regular variance).
You’re almost done! As a last step, divide this number, to yield a final result
between -1 and 1.
(Divide by what? Well, consider the shortcoming of covariance: If people’s
weights and heights are both very spread out, then “distance from the average”
will typically be a large number. In other words, covariance is larger for volatile
variables and lower for stable ones, regardless of the relationship between them.
How do we fix this problem? Just divide by the variances themselves.)
Whew! Now for the easy part: interpreting the values.
A positive value (e.g., 0.8) suggests that people above average on one
variable (e.g., height) are usually above average on the other (i.e., weight). A
negative value (e.g., -0.8) suggests the opposite: it’s a population dominated by
tall, light people and short, heavy ones. Finally, a value close to zero suggests
there’s no meaningful relationship at all.
Chapter 17: The Last .400 Hitter
• an Englishman named Henry Chadwick: As quoted in: “Henry Chadwick,”
National Baseball Hall of Fame, http://baseballhall.org/hof/chadwick-henry.
“Every action,” he said of baseball, “is as swift as a seabird’s flight.” As my
friend Ben Miller points out, this raises a question: African swallow, or
European?
• all-time record is 400: This record belongs to West Indies batsman Brian
Lara, who in 2004 scored exactly 400 runs against England without making an
out. I’m grateful for the strange statistical rhyme with this chapter’s title.
• “one true criterion”: As quoted in: Michael Lewis, Moneyball: The Art of
Winning an Unfair Game (New York: W. W. Norton, 2003), 70. It will surprise
no one when I confess that this chapter owes an unpayable debt to Moneyball,
and if you find the story of baseball statistics tolerable (never mind interesting),
then you’ll enjoy the book.
• you’d have to reset the ball at center field every five seconds: I took the
English Premier League’s 38-game season. Calling each game 90 minutes, plus
a generous 10 minutes for stoppage time, that’s 3800 minutes. Twelve data
points per minute (i.e., one per 5 seconds) would give 45,600 data points—still
less than a baseball season’s 48,000, but close enough.
• the first edition of Ernest Hemingway’s: Ernest Hemingway, The Old Man
and the Sea, Life, September 1, 1952. Above the title, it read: “The editors of
Life proudly present for the first time and in full a great new book by a great
American writer.”
• “Goodby to Some Old Baseball Ideas”: Branch Rickey, “Goodby to Some
Old Baseball Ideas,” Life, August 2, 1954. The subhead reads, “‘The Brain’ of
the Game Unveils Formula That Statistically Disproves Cherished Myths and
Demonstrates What Really Wins.”
• in 1858, “called strikes” were born: E. Miklich, “Evolution of 19th Century
Baseball Rules,” 19cBase ball.com, http://www.19cbaseball.com/rules.html.
• Enough such “balls”: It took a long time to settle on how many balls
constituted a walk. At first it took three, then nine, then eight, then six, then
seven, then five, until in 1889 the number finally settled at its current value of
four.
• Walks weren’t deemed an official statistic until 1910: Even then, writers
didn’t exactly embrace them. Just listen to sportswriter Francis Richter:

The figures [for walks] are of no special value or importance… bases on

balls are solely charged to the pitcher, beyond the control of the batsman,
and therefore not properly to be considered in connection with his
individual work, except as they may indicate in a vague way his ability to
‘wait out’ or ‘work’ the pitcher.

• Source: Bill James, The New Bill James Historical Baseball Abstract (New
York: Free Press, 2001), 104.
• hitters walk 18% or 19% of the time: In 2017, for example, the league leader
was Joey Votto, who walked 134 times in 707 plate appearances. That’s 19% of
the time.
• just 2% or 3%: In 2017, Alcides Escobar walked 15 times in 629 plate
appearances, for a rate of 2.4%. Tim Anderson outdid him by walking 13 times
in 606 plate appearances, or 2.1%.
• OPS: As a math teacher, I’ve always hated this junkyard stat, because I lose
life force every time someone adds two fractions with different denominators. I
always wished they’d calculate a new statistic, “bases per plate appearance”—
just like SLG, except you count walks on par with singles. Researching this
chapter, I realized my folly: Although cleaner in concept, this new statistic
would actually be less predictive in practice. For the 2017 numbers, its
correlation with a team’s run scoring is 0.873. That’s worse than OBP.
• On the 50th anniversary of the Life article: Alan Schwarz, “Looking Beyond
Batting Average,” New York Times, August 1, 2004,
http://www.nytimes.com/2004/08/01/sports/keeping-score-looking-beyond-
batting-average.html.
• The baseball dialogue in The Old Man and the Sea: For example, “‘I would
like to take the great DiMaggio fishing,’ the old man said. ‘They say his father
was a fisherman.’”
• writer Bill James: Scott Gray, The Mind of Bill James: How a Complete
Outsider Changed Baseball (New York: Three Rivers Press, 2006). The book
has some choice Bill James aphorisms, including: “There will always be people
who are ahead of the curve, and people who are behind the curve. But
knowledge moves the curve.” Also: “When you add hard, solid facts to a
discussion, it changes that discussion in far-reaching ways.”
• batting average is a sideshow: And yet, through the 1970s, official MLB
statistics listed team offenses in order of batting average, not runs scored. “It
should be obvious,” quipped James, “that the purpose of an offense is not to
compile a high batting average.”
• salaries soared: Michael Haupert, “MLB’s Annual Salary Leaders Since
1874,” Outside the Lines (Fall 2012), Society for American Baseball Research,
http://sabr.org/research/mlbs-annual-salary-leaders-1874-2012. I cheat big-time
in the graph by not adjusting for inflation; for example, Joe DiMaggio’s $90,000
salary in 1951 would be more like $800,000 in 2017 dollars. Even so, it’s hard to
argue with the impact of free agency.
• “Baseball is the only field of endeavor”: Pete Palmer, The 2006 ESPN
Baseball Encyclopedia, (New York: Sterling, 2006), 5.
• a heartbreaking .39955: Bill Nowlin, “The Day Ted Williams Became the
Last .400 Hitter in Baseball,” The National Pastime (2013), Society for
American Baseball Research, https://sabr.org/research /day-ted-williams-
became-last-400-hitter-baseball.
• no one has hit .400 since: Ben Miller (a gentleman, a culinary hero, and an
incurable Sox fan) points out that Williams’s success went far beyond that
number. In 1941, he “had an OBP of .553, which stood as the single-season
record for more than 60 years… Also, Ted Williams’s career OBP of .482 is the
highest ever. Oh, also, Ted Williams’s career batting average of .344 is 6th-
highest all-time, and you have to go all the way to Tony Gwynn at #17 to find
the next all-time BA leader who played after 1940.” No matter how you slice it,
the Splendid Splinter was good at his job.
• “If I had known hitting .400”: Bill Pennington, “Ted Williams’s .406 Is More
Than a Number,” New York Times, September 17, 2011,
http://www.nytimes.com/2011/09/18/sports/baseball/ted-williamss-406-average-
is-more-than-a-number.html.
 Chapter 18: Barbarians at the Gate of Science

• we are living through an epoch of crisis: A widely cited 2011 paper

demonstrated the dangers of the standard statistical techniques by reaching an
absurd conclusion: that listening to the Beatles song “When I’m Sixty Four”
made students younger. Not that they felt younger. They were, chronologically,
younger—or so the statistics said. The paper is cheeky, brilliant, and well worth
reading: Joseph Simmons, Leif D. Nelson, and Uri Simonsohn, “False-Positive
Psychology: Undisclosed Flexibility in Data Collection and Analysis Allows
Presenting Anything as Significant,” Psychological Science 22, no. 11 (2011):
1359–66, http://journals.sagepub.com/doi/abs/10.1177/0956797611417632.
On the science journalism side, I recommend: Daniel Engber, “Daryl Bem
Proved ESP Is Real: Which Means Science Is Broken,” Slate, May 17, 2017,
https://slate.com/health-and-science/2017/06 /daryl-bem-proved-esp-is-real-
showed-science-is-broken.html.
I also owe huge thanks to Kristina Olson (once a mentor, always a mentor!),
Simine Vazir (who delivered feedback on a 17-second deadline), and Sanjay
Srivastava (whom I cite toward the end of the chapter) for their help and
encouragement with this chapter.
• For a slightly more technical discussion, see the endnotes: The p-value is
the probability of achieving a result at least this extreme, given that the
experiment’s hypothesis is false.
Or, to unpack it a little more:
1. Assume that we’re chasing a fluke, that chocolate doesn’t really make
people happier.
2. Envision the distribution of all possible results that our experiment could
have obtained. Most of these results would be middling, unremarkable,
and unlikely to fool us. But a few rare flukes will make it look like
chocolate boosts happiness.
 3. Give our actual results a percentile rank in this distribution.
A low score (e.g., 0.03, or the 97th percentile) marks a serious fluke. Only
3% of false results would be this spectacular and deceptive. That extremity
suggests that—perhaps—it’s not a fluke at all. Maybe the effect we’re seeking is
real.
Crucially, such evidence is indirect. That 3% isn’t the probability of a fluke.
It’s the probability, assuming the result is false, that you’d get a fluke this
persuasive.
• a statistician named R. A. Fisher: Gerard E. Dallal, “Why P=0.05?” May 22,
2012, http://www.jerrydallal.com/lhsp/p05.htm.
• Children prefer lucky people: Kristina Olson et al., “Children’s Biased
Evaluations of Lucky Versus Unlucky People and Their Social Groups,”
Psychological Science 17, no. 10 (2006): 845–46,
http://journals.sagepub.com/doi/abs/10.1111/j.1467-
9280.2006.01792.x#articleCitationDownloadContainer.
• found the bias across cultures: Kristina Olson et al., “Judgments of the Lucky
Across Development and Culture,” Journal of Personality and Social
Psychology 94, no. 5 (2008): 757–76.
• the senior thesis of a feckless 21-year-old: Ben Orlin, “Haves and Have Nots:
Do Children Give More to the Lucky Than the Unlucky?” Yale University,
senior thesis in psychology, 2009. Adviser: Kristina Olson, who deserves credit
for all of the paper’s merits and blame for none of its shortcomings.
• The crucial p-value was well above 0.05: I predicted that eight-year-olds
would be sensitive to context, giving toys to the unlucky when they had lost a
toy, but not when they had experienced an irrelevant form of bad luck (being
assigned to play a game with a disliked classmate), whereas the five-year-olds
would be insensitive to such contextual issues. The p-value for this test was 0.15.
By comparison, when it came to liking rather than giving, my data replicated
Kristina’s result, with subjects preferring lucky kids (p = 0.029).
• the website Spurious Correlations: Highly recommended:
http://www.tylervigen.com/.
• An anonymous 2011 survey: Leslie John, George Loewenstein, and Drazen
Prelec, “Measuring the Prevalence of Questionable Research Practices with
Incentives for Truth Telling,” Psychological Science 23, no. 5 (2012): 524–32,
http://citeseerx.ist.psu.edu/viewdoc/download?
doi=10.1.1.727.5139&rep=rep1&type=pdf.
• Final P-Value vs. Minimum P-Value: This is a somewhat unfair comparison,
because even a p-hacking researcher wouldn’t retroactively go back and “stop”
the study at the minimum p-value. Indeed, the idea of checking the data after
every subject is dubious, too.
For a more naturalistic test, I went back to my “types of water” spreadsheet
and simulated another 50 trials. This time, I started with 30 subjects (10 per
group), then added more subjects as needed in chunks of 15 (five per group), up
to a maximum of 90 total. With this extra degree of freedom, a horrifying 76%
of trials achieved p-values below 0.05. One trial managed a p-value of 0.00001,
which is a 1-in-100,000 absurdity.
• lowering the p-value threshold: On first encounter, I thought this sounded
stupid. Sort of like: “Our roller coaster is meant for people at least 4 feet tall, but
since children are donning trench coats and standing on each other’s shoulders to
sneak in, let’s raise the bar to 5 feet.” Then I read the relevant paper: Daniel J.
Benjamin et al., “Redefine Statistical Significance,” PsyArXiv, July 22, 2017,
https://psyarxiv.com/mky9j/. I’m now persuaded. It seems that dropping from
0.05 to 0.005 would correspond better to intuitive Bayesian thresholds and
would reduce false positives while requiring only a moderate increase in sample
sizes.
It’s also worth noting that Bayesianism is savvier and more sophisticated
than I portray it in the chapter text. The question of “What should the prior be?”
can be sidestepped by performing numerous analyses, on a variety of priors, and
showing a graph of the overall trends. For what it’s worth, Sanjay Srivastava
persuaded me that a switch to Bayesianism wouldn’t really address the causes of
the replication crisis, although it might be a good idea for other reasons.
• careful replications of 100 psychological studies: Open Science
Collaboration, “Estimating the Reproducibility of Psychological Science,”
Science 349, no. 6251 (2015):
http://science.sciencemag.org/content/349/6251/aac4716.
Chapter 19: The Scoreboard Wars

• In 1982, Jay Mathews was the Los Angeles bureau chief: Jay Mathews,
“Jaime Escalante Didn’t Just Stand and Deliver. He Changed U.S. Schools
Forever,” Washington Post, April 4, 2010, http://www.washingtonpost.com/wp-
dyn/content/article/2010/04/02/AR2010040201518.html.
• “sheer mockery”: “Mail Call,” Newsweek, June 15, 2003,
http://www.newsweek.com/mail-call-137691.
• “with the hope that people will argue”: Jay Mathews, “Behind the Rankings:
How We Build the List,” Newsweek, June 7, 2009,
http://www.newsweek.com/behind-rankings-how-we-build-list-80725.
• heed the tale of British ambulances: Tim Harford, Messy: How to Be
Creative and Resilient in a Tidy-Minded World (London: Little, Brown, 2016),
171–73.
• Teacher’s Value Added: This one is real: Cathy O’Neil, Weapons of Math
Destruction: How Big Data Increases Inequality (New York: Broadway Books,
2016), 135–40.
• “Nearly every professional educator”: Jay Mathews, “The Challenge Index:
Why We Rank America’s High Schools,” Washington Post, May 18, 2008,
http://www.washingtonpost.com/wp-
dyn/content/article/2008/05/15/AR2008051502741.html.
• “passer rating”: I’ve been watching football since I was seven, and have
never understood passer rating. I figured it was time to give it a shot.
The formula took a few minutes to unpack, but once I did, I saw that it isn’t
all that complicated. First, it assigns points (1 per yard, 20 per completion, 80
per touchdown, and -100 per interception). Second, it computes the number of
points per passing attempt. And third, it tacks on a meaningless
addition/multiplication flourish. Or, in equation form:

We could leave it there… except this formula allows for negative scores (if
you throw too many interceptions) and has an unattainable maximum (whereas
the actual maximum of 158¹/³ has been achieved in games by more than 60
quarterbacks). To fix this, you need to cap the points contributed by each stat,
and for that, it’s easier to present the formula as it is in the main text.
Why does passer rating feel so confusing? As with most scary formulas, two
factors come into play: (1) it is confusing, being a strangely weighted average of
four variables, each subject to arbitrary cutoffs; and (2) on top of that, most
sources present it in a needlessly counterintuitive and opaque way. Look on
Wikipedia if you want to see what I mean.
• “One of its strengths is the narrowness”: Mathews, “Behind the Rankings.”
• “The list has taken on a life of its own”: J. P. Gollub et al., eds., “Uses,
Misuses, and Unintended Consequences of AP and IB,” Learning and
Understanding: Improving Advanced Study of Mathematics and Science in U.S.
High Schools, National Research Council (Washington, DC: National Academy
Press, 2002), 187, https://www.nap.edu/read/10129 chapter12#187.
• “The parents are making the most noise”: Steve Farkas and Ann Duffett,
“Growing Pains in the Advanced Placement Program: Do Tough Trade-Offs Lie
Ahead?” Thomas B. Fordham Institute (2009),
http://www.edexcellencemedia.net/publications/2009/200904_growingpainsintheadvancedpla
According to the survey, more than one in five AP teachers nationwide
believed that the Challenge Index had had “some impact” on their school’s AP
offerings. In suburbs and cities, the number was closer to one in three.
Meanwhile, just 17% viewed the list as “a good idea.”
• “Because your index only considers the number”: Valerie Strauss,
“Challenging Jay’s Challenge Index,” Washington Post, February 1, 2010,
http://voices.washingtonpost.com/answer-sheet/high-school/challenged-by-jays-
challenge-i.html.
In 2006, the Challenge Index ranked Eastside High in Gainesville, Florida, as
the #6 school in the US. But only 13% of Eastside’s nearly 600 African
American students were reading at grade level. Schools #21 and #38 showed
similar incongruities. Critics saw this as a sign that unprepared students were
being shoveled into college-level courses in an effort to catch Newsweek’s eye.
Source: Michael Winerip, “Odd Math for ‘Best High Schools’ List,” New York
Times, May 17, 2006,
http://www.nytimes.com/2006/05/17/education/17education.html.
• Who, exactly, is the index for?: John Tierney, “Why High-School Rankings
Are Meaningless—and Harmful,” Atlantic, May 28, 2013,
https://www.theatlantic.com/national/archive/2013/05/why-high-school-
rankings-are-meaningless-and-harmful 276122.
• “We cannot resist looking at ranked lists”: Mathews, “Behind the
Rankings.”
• “we are all tribal primates”: Jay Mathews, “I Goofed. But as Usual, a Smart
Educator Saved Me,” Washington Post, June 25, 2017,
https://www.washingtonpost.com/local/education/i-goofed-but-as-usual-a-smart-
educator-saved-me/2017/06/25 /7c6a05d6-582e-11e7-a204-
ad706461fa4f_story.html.
• “a very elastic term in our society”: Winerip, “Odd Math for ‘Best High
Schools’ List.”
• Mathews likes to cite a 2002 study: Jay Mathews, “America’s Most
Challenging High Schools: A 30-Year Project That Keeps Growing,”
Washington Post, May 3, 2017, https://www.wash
ingtonpost.com/local/education/jays-americas-mo st-challenging-high-schools-
main-column/2017/05/03 /eebf0288-2617-11e7-a1b3-faff0034e2de_story.html.
• seemed to lay the groundwork for success in college: Not all scholars agree.
The 2010 book AP: A Critical Examination of the Advanced Placement Program
found a consensus building among researchers that the AP expansion had begun
to see diminishing returns. Coeditor Philip Sadler said, “AP course work does
not magically bestow advantages on underprepared students who might be better
served by a course not aimed at garnering college credit.” Source: Rebecca R.
Hersher, “The Problematic Growth of AP Testing,” Harvard Gazette, September
3, 2010, https://news.harvard.edu/gazette/story/2010/09 philip-sadler.
• As of 2017, that number is 12%: Mathews, “America’s Most Challenging
High Schools.”
• “the view that schools with lots of rich kids are good”: Mathews, “The
Challenge Index.”
• how you feel about his vision: I have a humble suggestion for Mathews:
Instead of counting “exams taken,” count “scores of at least 2.” In my
experience (and in the Texas study that Mathews likes to cite) a 2 shows some
intellectual spark, a signal of growth. But I’m not convinced the typical 1-scorer
has gained from the class. Requiring students to score above the minimum
would eliminate the incentive to throw wholly unprepared kids into the test.
Chapter 20: The Book Shredders

• Ben Blatt’s delightful book: Ben Blatt, Nabokov’s Favorite Word Is Mauve
(New York: Simon & Schuster, 2017).
• the greatest novels: I computed the “greatness score” from data on Goodreads,
where readers give ratings from one to five stars. First, I calculated the total
number of stars awarded to each book. Faulkner’s ranged from under 1500 (for
Pylon) to over 500,000 (for The Sound and the Fury). Then I took a logarithm,
which collapses this exponential scale down to a linear one. The correlation
between adverbs and “greatness” was -0.825. Equivalent analyses for
Hemingway and Steinbeck yielded coefficients of -0.325 and -0.433: substantial,
but hard to discern on a graph. The adverb data comes from Blatt, and the
method is a variation on his. (He uses number of ratings instead of number of
stars; it yields pretty much the same result.)
• a blockbuster research article: Jean-Baptiste Michel et al., “Quantitative
Analysis of Culture Using Millions of Digitized Books,” Science 331, no. 6014
(2011): 176–82,
http://science.sciencemag.org/content/early/2010/12/15/science.1199644. They
wrote:
 The corpus cannot be read by a human. If you tried to read only the
entries from the year 2000 alone, at the reasonable pace of 200
words/minute, without interruptions for food or sleep, it would take eighty
years. The sequence of letters is one thousand times longer than the
human genome; if you wrote it out in a straight line, it would reach to the
moon and back 10 times over.

• “New Window on Culture”: Patricia Cohen, “In 500 Billion Words, New
Window on Culture,” New York Times, December 16, 2010,
http://www.nytimes.com/2010/12/17/books/17words.html.
• “precise measurement”: Michel and Aiden wrote (emphasis mine):

Reading small collections of carefully chosen works enables scholars to

make powerful inferences about trends in human thought. However, this
approach rarely enables precise measurement of the underlying
phenomena.

• Virginia Woolf’s, like a galaxy: Virginia Woolf, A Room of One’s Own

(1929).
• Apply Magic Sauce: Try it for yourself at https://applymagicsauce.com/.
• 25 blog posts: Taken from http://mathwithbaddrawings.com/.
• In a 2001 paper: Moshe Koppel, Shlomo Argamon, and Anat Rachel Shimoni,
“Automatically Categorizing Written Texts by Author Gender,” Literary and
Linguistic Computing 17, no. 4 (2001): 401-12,
http://u.cs.biu.ac.il/~koppel/papers/male-female-llc-final.pdf.
• A later paper: Shlomo Argamon et al., “Gender, Genre, and Writing Style in
Formal Written Texts,” Text 23, no. 3 (2003): 321–46,
https://www.degruyter.com/view/j/text.1.2003.23.issue-3/text.2003.014
/text.2003.014.xml.
• As Blatt points out: Blatt, Nabokov’s Favorite Word Is Mauve, 37.
• a data company called CrowdFlower: Justin Tenuto, “Using Machine
Learning to Predict Gender,” CrowdFlower, November 6, 2015,
https://www.crowdflower.com/using-machine-learn ing-to-predict-gender/.
• gender-indicative words in classic literature: Blatt, 36.
• mathematician Cathy O’Neil: Cathy O’Neil, “Algorithms Can Be Pretty
Crude Toward Women,” Bloomberg, March 24, 2017,
https://www.bloomberg.com/view/articles/2017-03-24 /algorithms-can-be-
pretty-crude-toward-women.
• this vindicates Woolf: In A Room of One’s Own, Woolf wrote:

The weight, the pace, the stride of a man’s mind are too unlike [a
woman’s] for her to lift anything substantial from him… Perhaps the first
thing she would find, setting pen to paper, was that there was no common
sentence ready for her to use.

Though she enjoys that masculine style (“swift but not slovenly, expressive
but not precious”), she added, “It was a sentence that was unsuited for a
woman’s use.”

Charlotte Brontë, with all her splendid gift for prose, stumbled and fell
with that clumsy weapon in her hands… Jane Austen looked at it and
laughed at it and devised a perfectly natural, shapely sentence proper for
her own use and never departed from it. Thus, with less genius for writing
than Charlotte Brontë, she got infinitely more said.

• Enter two statisticians: Frederick Mosteller and David Wallace, “Inference in

an Authorship Problem,” Journal of the American Statistical Association 58, no.
302 (1963): 275–309.
• tore The Federalist Papers to shreds: I mean that literally. Blatt writes:

They took copies of each essay and dissected them, cutting the words
apart and arranging them (by hand) in alphabetical order. At one point
Mosteller and Wallace wrote, “during this operation a deep breath
created a storm of confetti and a permanent enemy.”

I came super close to titling this chapter “A Storm of Confetti and a

Permanent Enemy.”
• Stanford Literary Lab: Sarah Allison et al., “Quantitative Formalism: An
Experiment,” Stanford Literary Lab, pamphlet 1, January 15, 2011,
https://litlab.stanford.edu/LiteraryLabPamphlet1.pdf. I love this paper. In fact, I
recommend every Stanford Literary Lab product I’ve ever read. They’re like
vintage Pixar—no bad apples.

Section V: On the Cusp

• Attempted Light Bulb Designs: “I have constructed three thousand different
theories in connection with the electric light, each one of them reasonable and
apparently to be true,” Thomas Edison said once. “Yet only in two cases did my
experiments prove the truth of my theory.” Of course, “theories” aren’t the same
thing as “design attempts,” and no one knows how many designs he tried before
success. Source: “Myth Buster: Edison’s 10,000 Attempts,” Edisonian 9 (Fall
2012), http://edison.rutgers.edu/newsletter9.html#4.
• 50 cents, or 51 cents, but not 50.43871 cents: You can bend this rule if you’re
doing many transactions at once. Charging $50,438.71 for 100,000 pencils is
equivalent to charging $0.5043871 for each. Financial institutions, carrying out
tremendous numbers of transactions per day, often work in tiny fractions of
pennies.
Chapter 21: The Final Speck of Diamond Dust

• Adam Smith asked a question: He didn’t really phrase it as a question:

Nothing is more useful than water: but it will purchase scarce any thing;
scarce any thing can be had in exchange for it. A diamond, on the
contrary, has scarce any value in use; but a very great quantity of other
goods may frequently be had in exchange for it.

From: Adam Smith, An Inquiry into the Nature and Causes of the Wealth of
Nations (1776), book I, chapter IV, paragraph 13, accessed through the online
Library of Economics and Liberty:
http://www.econlib.org/library/Smith/smWN.html.
• Classical economics lasted a century: My go-to source for this chapter was
this wonderful book: Agnar Sandmo, Economics Evolving: A History of
Economic Thought (Princeton, NJ: Princeton University Press, 2011).
• This idea long predates marginalism: Campbell McConnell, Stanley Brue,
and Sean Flynn, Economics: Principles, Problems, and Policies, 19th ed. (New
York: McGraw-Hill Irwin, 2011). The “earth’s fertility” quote comes from
section 7.1, Law of Diminishing Returns,
http://highered.mheducation.com/sites/0073511447/student_view0/chapter7/origin_of_the_ide
• The earth’s fertility resembles a spring: As Mike Thornton points out—and I
humbly thank him for his help with this chapter—this analogy elides some
subtleties. Heterogeneous land can be perfect for growing various crops that
prefer different conditions, and farmers can take steps (such as rotating crops) in
order to improve the quality of the soil.
• This function of utility is peculiar: William Stanley Jevons, “Brief Account
of a General Mathematical Theory of Political Economy,” Journal of the Royal
Statistical Society, London 29 (June 1866): 282–87,
https://socialsciences.mcmaster.ca/econ ugcm3ll3/jevons/mathem.txt.
• “A true theory of economy”: Jevons again. Quotable guy. Writing this
chapter made me a fan.
• In school, I learned that economics has mirror symmetry: For a refresher, I
drew from the lecture notes of Michal Brzezinski, an assistant professor of
economics at the University of Warsaw,
http://coin.wne.uw.edu.pl/mbrzezinski/teaching/HEeng/Slides/marginalism.pdf.
• self-described “standup economist”: Yoram Bauman. I quote from his video
Mankiw’s Ten Principles of Economics, Translated, which cracked me up when
I first saw it in college. Check out his site at http://standupeconomist.com/.
• “The proper mathematical analogy”: Sandmo, Economics Evolving, 96.
• “persist in using everyday language”: Sandmo, 194.
• “the only work by an economist”: The quote is from Wikipedia. Shhh, don’t
tell anyone.
• renewed empiricism: Jevons embodied this trend. He predicted that Britain
would soon exhaust its coal resources, and argued that the ups and downs of the
business cycle stemmed from cooler temperatures caused by sunspots. Okay, so
he was wrong about both, but we know that thanks to methods that he himself
ushered in.
Walras, by contrast, was a bit of an anti-empiricist. In his view, you must first
filter out the messy details of reality to arrive at pure quantitative concepts.
Second, you must operate on and reason about these mathematical abstractions.
Third, almost as an afterthought, you look back to practical applications. “The
return to reality,” Walras wrote, “should not take place until the science is
completed.” To Walras, “science” was something that happens far away from
reality.
If you meet an economist who still thinks this way today, just follow these
simple steps: (1) raise your arms; (2) begin bellowing; (3) if the economist
continues to approach, box him on the ears. Remember, these economists are as
frightened of us as we are of them.

• Economics, like its brethren in the natural sciences, became ahistorical:

This insight comes straight from Sandmo.
Chapter 22: Bracketology

• “Famous historical Belgians”: I’m not kidding. Belgium’s two greatest

celebrities are both fictional: Hercule Poirot and Tintin. Personally, I count this
as a point in Belgium’s favor. Famous historical people are nothing but trouble.
• Our tale begins in 1861: My source for the history of US taxation was: W.
Elliot Brownlee, Federal Taxation in America: A History, 3rd ed. (Cambridge:
Cambridge University Press, 2016).
• you now owed 3%: The cutoff point of $800 landed quite close to the nation’s
average family income of $900 (per Brownlee). Pretty impressive, given that the
government threw that particular dart blindfolded. I should also note that the
taxes didn’t really have a 0% bracket; rather, they had a universal exemption of
$800. The math is equivalent.
• In 1913, the first year of the permanent income tax: Brownlee, Federal
Taxation in America, 92.
• President Woodrow Wilson wrote: Brownlee, 90.
• the 1918 tax scheme: Numbers drawn from the Wikipedia page “Revenue Act
of 1918.”
• One industrious fellow named J. J.: J. J.’s system requires calculus to pull
off, which goes to show that J. J. was not your typical precalculus student. His
system (unlike the simpler one I illustrate) had several discrete rate jumps
leading into a continuously varying top bracket, whose formula involved a
natural logarithm. To be a teacher often involves simplifying the world so your
students can grasp it; here, it seems to entail the reverse.
• Donald Duck: The New Spirit (Walt Disney Studios, 1942). To see the
magnificent original, go to https://www.youtube.com/watch?v=eMU-KGKK6q8.
• their greatest album: Yes, Revolver is their best. Come at me, Rubber Soul
fans.
• one for you, nineteen for me: This corresponds to a marginal rate of 95%, so
George Harrison was slightly understating his tax burden.
• a woman named Pomperipossa: Astrid Lindgren, “Pomperipossa in
Monismania,” Expressen, March 3, 1976. I used Lennart Bilén’s helpful 2009
translation: https://lenbilen.com/2012/01/24 pomperipossa-in-monismania.
• caught fire in Sweden: “Influencing Public Opinion,” AstridLindgren.se,
accessed September 2017, http://www.astridlindgren.se/en/person/influencing-
public-opinion.
• higher income levels faced a lower marginal tax rate: Of course, this is how
the actual payroll tax, designed to fund Social Security and Medicare, works.
You pay 15% on your first $120,000, and nothing on income above that level.
Chapter 23: One State, Two State, Red State, Blue State

• democracy is an act of mathematics: After writing this cheerful line, I came

across its evil twin. My beloved Jorge Luis Borges, in a moment of
uncharacteristic cynicism, once called democracy “an abuse of statistics.”
• “men chosen by the people”: Federalist, no. 68 (Alexander Hamilton). While
I’m name-dropping beloved American figures, I shall thank Geoff Koslig (a man
so constitutionally patriotic that his brain’s background music is “You’re a
Grand Old Flag”) for his help with this chapter. (I’d also like to thank Jeremy
Kun and Zhi Chen for their capable help.)
• Who Chose Electors?: I drew this line of argument, like 93% of my
knowledge about the Constitution, from Akhil Reed Amar. In particular, from
this book: Akhil Reed Amar, America’s Constitution: A Biography (New York:
Random House, 2005), 148–52. The data in the graph comes from (where else?)
Wikipedia.
• this system encoded an advantage for slavery: This argument comes from
Amar, America’s Constitution, 155–59, 342–47. He goes further, arguing that
the Electoral College not only advantaged slave states (which is pretty
indisputable) but was designed in part to do so. This position has drawn some
criticism, including Gary L. Gregg II’s energetically titled: “No! The Electoral
College Was Not about Slavery!” Law and Liberty, January 3, 2017,
http://www.libertylawsite.org/2017/01/03/no-the-electoral-college-was-not-
about-slavery/.
My $0.02: It would be silly to argue that slavery was the only factor that
shaped the Electoral College. But I don’t see anyone saying that. Here, Amar
writes: “three main factors—information barriers, federalism, and slavery—
doomed direct presidential election in 1787.” Meanwhile, the idea that slavery
was a nonfactor strikes my amateur eyes as quite wrong. On July 19, 1787,
James Madison raised the issue:

There was one difficulty… attending an immediate choice by the people.

The right of suffrage was much more diffusive in the Northern than the
 Southern States; and the latter could have no influence in the election on
the score of the Negroes. The substitution of electors obviated this
difficulty…

On July 25, Madison raised the point again—although this time, he endorsed
direct election as his favorite method, saying: “As an individual from the S.
States he was willing to make the sacrifice.” Read the details for yourself in
“Notes on the Debates in the Federal Convention”:
http://avalon.law.yale.edu/subject_menus/debcont.asp.
Further evidence that the slavery advantage was a feature, not a bug: in his
first inaugural address in 1833, Southerner Andrew Jackson proposed replacing
the Electoral College with a direct election—but only if the irregularities in vote
weighting were preserved. Source: Amar, 347.
• Here’s the elector count from 1800: I drew census data from Wikipedia.
Then, I redistributed the number of representatives in the House in proportion to
the free population of each state.
• bucket-filled with red and blue: Koslig informed me that the “Democrat =
blue, Republican = red” color scheme was once the opposite. It shifted in the
1990s and was cemented in 2000. For more, you can read a fun Smithsonian
piece: Jodi Enda, “When Republicans Were Blue and Democrats Were Red,”
Smithsonian.com, October 31, 2012, https://www.smithsonianmag.com/history
when-republicans-were-blue-and-democrats-were-red-104176297.
• 60% of the vote gets six electors: Even this is not straightforward as it sounds.
Take the 2016 election, where the Democrats’ 46.44% rounds up to 50% (for
five electors), and the Republicans’ 44.92% rounds down to 40% (for four
electors). That’s only nine. Who gets the last elector? You could give it to the
Libertarians, but their 3.84% is pretty far from 10%. It’s probably more logical
to give the last elector to the party that was closest to earning it outright—in this
case, the Republicans, who were just 0.08% away.
• where 96% of the country sits today: As Koslig points out, this dynamic is
different in states that favor one party in presidential elections and the other for
state government. In recent years, a few such states have considered joining
Nebraska and Maine in their way of allocating electors, as a way for the state-
government-ruling party to seize some electors from its rival.
• a beautiful method for answering this question: Nate Silver, “Will the
Electoral College Doom the Democrats Again?” FiveThirtyEight, November 14,
2016, https://fivethirtyeight.com/features/will-the-electoral-college-doom-the-
democrats-again/.
• no correlation between which party has the Electoral College advantage:
“The major exception,” Silver points out,

was in the first half of the 20th century, when Republicans persistently
had an Electoral College advantage because Democrats racked up huge
margins in the South, yielding a lot of wasted votes.… The question is
whether Democrats are reentering something akin to the “Solid South”
era, except with their votes concentrated in more urban coastal states….

Silver’s guess seems to be “no.” Time will tell.

• the Amar Plan: It originated with Professor Akhil Reed Amar (who taught the
superb Constitutional Law class I took in college) and his brother, Professor
Vikram Amar.
Chapter 24: The Chaos of History

• 8 inches of snow fell on the eve of President Kennedy’s inauguration:

Source: Wikipedia, of course. I also take thank David Klumpp and Valur
Gunnarsson for brain-tickling feedback on this chapter.
• a researcher named Edward Lorenz found something funny: This
anecdote, the ensuing graph, and indeed most of the mathematics in this chapter
come from the indispensable: James Gleick, Chaos: Making a New Science
(New York: Viking Press, 1987). The anecdote is found here.
• The pendulum obeyed a simple equation: Okay, they didn’t have “meters” in
the 17th century. This approximation is a modern restatement, and it holds only
for fairly small-angle swings. Still, it’s striking that the cycle length doesn’t
depend on the angle: a wide, fast oscillation will take about the same time as a
slower, narrower one.
• The double pendulum: For an awesome simulator, check out:
https://www.myphysicslab.com/pendulum/double-pendulum-en.html. Seriously,
if you’ve been skimming the endnotes in search of some unknown inspiration,
then this is it.
• The Game of Life: Siobhan Roberts, Genius at Play: The Curious Mind of
John Horton Conway (New York: Bloomsbury, 2015). On pages xiv–xv,
Roberts quotes musician Brian Eno on the Game of Life: “The whole system is
so transparent that there should be no surprises at all, but in fact there are plenty:
the complexity and ‘organicness’ of the evolution of the dot patterns completely
beggars prediction.” Here, she quotes philosopher Daniel Dennett: “I think that
Life should be a thinking tool in everybody’s kit.”
• “The big choices we make”: This Amos Tversky quote comes from: Michael
Lewis, The Undoing Project: A Friendship That Changed Our Minds (New
York: W. W. Norton, 2016), 101.
• Kim Stanley Robinson’s novella: Kim Stanley Robinson, The Lucky Strike
(1984). Excellent story. I read it in: Harry Turtledove, ed., The Best Alternate
History Stories of the 20th Century (New York: Random House, 2002).
Also, I can’t help mentioning here my favorite alternate-history novel: Orson
Scott Card, Pastwatch: The Redemption of Christopher Columbus (New York:
Tor Books, 1996).
• Kyoto had remained atop the target list: Mariko Oi, “The Man Who Saved
Kyoto from the Atomic Bomb,” BBC News, August 9, 2015,
http://www.bbc.com/news/world-asia-33755182.
• Hiroshima was a civilian target: Hard to imagine that Truman didn’t
understand this, but it seems he didn’t. Listen to the excellent episode: “Nukes,”
Radiolab, April 7, 2017, http://www.radiolab.org/story/nukes.
• Winston Churchill, not primarily known for his science fiction: In his 1931
essay “If Lee Had NOT Won the Battle of Gettysburg,” Churchill pretends that
he is a historian living in an alternate timeline, where the Confederacy won; then
he writes an “alternate” fiction imagining what life is like in our timeline. I
found his conclusions stupid and naïve, but then again, I never saved a
civilization from Nazis, so don’t feel compelled to listen to me.
https://www.winston churchill.org/publications/finest-hour-extras/qif-lee-had-
not-won-the-battle-of-gettysburgq/
• “Confederate is a shockingly unoriginal idea”: Ta-Nehisi Coates, “The Lost
Cause Rides Again,” Atlantic, August 4, 2017,
https://www.theatlantic.com/entertainment/archive/2017/08 no-
confederate535512/.
• The rules of plausibility: This discussion was inspired in part by Michael
Lewis’s The Undoing Project, 299–305.
• shed light on the current state: A key insight of chaos theory is that many
systems of towering complexity follow simple underlying rules. Maybe—and
here I slip into the realm of late-night dorm-room speculation—there’s a way to
articulate a few deterministic rules that underlie (aspects of) history. We might
develop toy models to capture history’s sensitivity and degree of randomness,
the way Edward Lorenz’s simulation did for the weather.
• Benoit Mandelbrot published a brief bombshell of a paper: Benoit
Mandelbrot, “How Long Is the Coast of Britain? Statistical Self-Similarity and
Fractional Dimension,” Science 156, no. 3775 (1967): 636–38.
• “Each new book”: Alan Moore and Eddie Campbell, From Hell (Marietta,
GA: Top Shelf Productions, 1999), appendix II, 23.
• something bottomless about the study of history: In Ursula Le Guin’s
science fiction story “A Man of the People,” the protagonist seeks to understand
the sprawling history of a civilization called Hain:

He knew now that historians did not study history. No human mind could
encompass the history of Hain: three million years of it…. there had been
uncountable kings, empires, inventions, billions of lives lived in millions
of countries, monarchies, democracies, oligarchies, anarchies, ages of
chaos and ages of order, pantheon upon pantheon of gods, infinite wars
and times of peace, incessant discoveries and forgettings, innumerable
horrors and triumphs, an endless repetition of unceasing novelty. What is
the use trying to describe the flowing of a river at any one moment, and
then at the next moment, and then at the next, and the next, and the next?
You wear out. You say: There is a great river, and it flows through this
land, and we have named it History.

Gosh, do I love this passage. Source: Ursula Le Guin, Four Ways to

Forgiveness (New York: Harper-Collins, 1995), 124–25.
• three PowerPoint presentations projected onto a single screen: I stole this
nifty image from Valur Gunnarsson.
Also, on the topic of competing historical metaphors, I can’t help quoting
Borges: “It may be that universal history is the history of the different
intonations given a handful of metaphors.”
Thank you for buying this ebook, published by Hachette Digital.

To receive special offers, bonus content, and news about our latest ebooks and
apps, sign up for our newsletters.

Sign Up

Or visit us at hachettebookgroup.com/newsletters