Game Theory and Its Applications

Nur Aini Masruroh

 What is game theory?

Game theory is a study of how to

mathematically determine the best strategy for
given conditions in order to optimize the
outcome
 Game Theory
Finding acceptable, if not optimal, strategies in
conflict situations.
Abstraction of real complex situation
Game theory is highly mathematical
Game theory assumes all human interactions
can be understood and navigated by
presumptions.
 Why is game theory
important?
All intelligent beings make decisions all the
time.
AI needs to perform these tasks as a result.
Helps us to analyze situations more rationally
and formulate an acceptable alternative with
respect to circumstance.
 Definitions

In game conflict, two opponents known as players

Each player has a finite number of strategies
Associated with each pair of strategies is a payoff
that one player pays to the other
 Zero Sum Game
The gain by one player is equal to the loss to
the other
Players may co-operate or compete
Being well informed may harm a player.
Basic assumption: each player chooses a
strategy that enables him to do the best he can,
given that his opponent knows the strategy he
is following
 Saddle point condition
Max(row minimum)=Min(column maximum)
If the game has a saddle point, the VALUE of the
game equals to the the common value of both sides

A saddle point can also be thought as an

equilibrium point in that neither player can
benefit from a unilateral change in strategy.

A saddle point is stable in that neither player has an

incentive to move away from it
 Two person zero sum
game:example
Two companies, A and B, sell two brands of flu
vaccine.

Company A advertises in radio (A1), TV (A2), and

newspaper (A3).

Company B advertises in radio (B1), TV (B2),

newspaper (B3), and mail brochures (B4).

The payoff matrix summarizes the percentage of the

market captured or lost by Company A
 Payoff Matrix
Company B
Maximin

B1 B2 B3 B4 Row min

A1 8 -2 9 -3 -3
Company A
A2 6 6 8 5
5
A3
-2 4 -9 5 -9

Column max 8 5 9 8
Minimax
 Analysis
Optimal solution: select strategies A2 and B2
Value of the game: 5%
A and B use pure strategy SADDLE POINT solution
Saddle point solution guarantees that neither
company is tempted to select a better strategy
 If you have a Dominating strategy, use
it

Use strategy Opponent

1 Strategy Strategy
A B

Strategy 1 150 1000

You

Strategy 2 25 - 10
 Eliminate any Dominated strategy
Eliminate
strategy 2 as Opponent
its dominated
by strategy 1 Strategy 1 Strategy 2

Strategy 1 150 1000

You
Strategy 2 25 - 10

Strategy 3 160 -15

 Non-Zero Sum Game

The sum of payoffs is not constant during the

course of game play.
Players may co-operate or compete
Being well informed may harm a player.
Example: two person constant-sum game
During 8 to 9 pm, two network vying for audience of 100
million viewers.

The network must simultaneously announce the type of

show they will air in that time slot.

The possible choices for each network are shown in the

Table. For example, if both networks choose a western, the
matrix indicates that 35 million people will watch network 1
and 100-35=65 million people will watch network 2

Thus we have a two-person constant-sum game with c=100.

Does this game have a saddle point?
 Payoff Matrix
Network 2
Maximin
Soap
Western opera Comedy Row min
Western
35 15 60 15
Network 1 Soap
opera 45 58 50 45

Comedy 38 14 70 14

Column max 45 58 70
Minimax
 Two-Person Nonconstant-
Sum Games
Most game-theoretic models of business situations
are not constant-sum games because it is unusual for
business competitors to be in conflict
Prisoner's dilemma
 Prisoner's dilemma
Two suspects arrested for a crime
Prisoners decide whether to confess or not to
confess
If both confess, both sentenced to 3 months of jail
If both do not confess, then both will be sentenced
to 1 month of jail
If one confesses and the other does not, then the
confessor gets freed (0 months of jail) and the
non-confessor sentenced to 9 months of jail
What should each prisoner do?
 Prisoners Dilemma

Prisoner 2

Confess Don't
Confess
-3 , -3 0 , -9
Prisoner 1
Don't
-9 , 0 -1 , -1
 Nash equilibrium

If there is a set of strategies with the property

that no player can benefit by changing her
strategy while the other players keep their
strategies unchanged, then that set of strategies
and the corresponding payoffs constitute the
Nash Equilibrium.
Source: http://www.lebow.drexel.edu/economics/mccain/game/game.html
 Nash equilibrium
Each players predicted strategy is the best
response to the predicted strategies of other
players
No incentive to deviate unilaterally
Strategically stable or self-enforcing
Prisoner 2

Confess Not Confess

Prisoner 1 Confess -3,-3 0,-9

Not Confess -9,0 -1,-1
 General Prisoner's Dilemma
Reward Matrix
Player 2

NC C
NC
P, P T, S
Player 1

C S, T R, R
C=cooperative action
NC=noncooperative action

R=reward for cooperating T=temptation for double- P=punishment for not S=payoff to person who is
if both players cooperates crossing opponent cooperating double-crossed
 Analysis of prisoner's
dilemma game
(P, P) is an equilibrium point -- P>S
For (R, R) is not equilibrium point -- T>R
The game is reasonable only if R>P
The prisoner's dilemma game is of interest because it
explains why two adversaries often fail to cooperate
with each other
 Example
Competing restaurant: Hot Dog King and Hot Dog Chef are
attempting to determine their advertising budget next year.

The two restaurant will have a combined sales of $240 million and
can spend either $6 million or $10 million on advertising.

If one restaurant spend more money than other, the restaurant that
spends money more will have sales of $190 million.

If both companies spend the same amount on advertising, then they

have equal sales.

Suppose each restaurant is interested in maximizing (contribution of

sales to profit)-(advertising cost), find an equilibrium point for this
game!
 Reward matrix for
advertising game
Hot Dog Chef

Hot Dog King Spend $10million Spend $6 million

Spend $10 million (2, 2) (9,-1)

Spend $6 million (-1,9) (6,6)

 Mixed strategy
A probability distribution over the pure strategies of
the games.

Example: two players (called Odd and Even)

simultaneously choose the number of fingers (1 or 2)
to put out. If the sum of the fingers put out by both
players is odd, the Odd wins $1 from Even. If the sum
of fingers is even, then Even wins $1 from Odd.
Consider the row player to be Odd and the Column
player to be Even.

This is a zero sum game but no saddle point.

 Reward matrix for Odd and Even
Even
Row
1 finger 2 finger
minimum

1 finger -1 +1 -1
Odd
2 finger +1 -1 -1

Column
+1 +1
maximum
 Mixed strategy: Odd's Optimal
Strategy
Define x1 = probability that Odd puts out one finger
x2 = probability that Odd puts out two fingers
y1 = probability that Even puts out one finger
y2 = probability that Even puts out two fingers
And x1, x2, y1, y2 >= 0
x1+x2= 1
y1+y2=1
Odd's mixed strategy becomes (x1, 1-x1)
If Even puts out 1 finger, then Odd's expected value = (-1)x1 + (+1)(1-
x1) = 1-2x1
Similarly, if Even puts out 2 fingers, Odd's expected value = (+1)(x1)
+ (-1)(1-x1) = 2x1 - 1
 Mixed strategy: Even's Optimal
Strategy

Even's mixed strategy (y1, 1-y1)

Basic assumption: Even should choose y1 to minimize her expected
losses (or equivalently, minimize Odd's expected reward) under the
assumption that Odd knows the value of y1.

If Odd puts out one finger, expected reward=(-1)y1 + (+1)(1-y1) =

1-2y1

If Odd puts out two fingers, expected reward=(+1)(y1) + (-1)(1-y1) =

2y1 -1
More on the idea of value and
optimal strategies
When each player is allowed to choose mixed strategies, the row
player's floor will always equal yo the column player's ceiling --> the
value of the game to the row player

Any mixed strategy for the row player that guarantees that the row
player gets an expected reward at least equal to the value of the game
is an optimal strategy for the row player.

Similarly, any mixed strategy for the column player that guarantees
that the column player's expected loss is no more than the value of
the game is an optimal strategy for the column player
 Mixed strategy
Rock-paper-scissors game
Each player simultaneously forms his or her hand into the
shape of either a rock, a piece of paper, or a pair of scissors
Rule: rock beats (breaks) scissors, scissors beats (cuts)
paper, and paper beats (covers) rock
No pure strategy Nash equilibrium
One mixed strategy Nash equilibrium each
player plays rock, paper and scissors each with 1/3
probability
 Nash's theorem
Existence
Any finite game will have at least one Nash equilibrium
possibly involving mixed strategies
Finding a Nash equilibrium is not easy
Not efficient from an algorithmic point of view
 Dynamic games
Sequential moves
One player moves
Second player observes and then moves

Examples
Industrial Organization a new entering firm in the market
versus an incumbent firm; a leader-follower game in
quantity competition
Sequential bargaining game - two players bargain over the
division of a pie of size 1 ; the players alternate in making
offers
 Game tree example: Bargaining

Period 2:
B offers x2.
(x1,1-x1) A responds. (x3,1-x3)

1 Y 1 1 Y

x1 x3
N
B B (0,0)
N
A B x2 A A
N

Y
0 0 0
Period 1: Period 3:
A offers x1. (x2,1-x2) A offers x3.
B responds. B responds.
 Economic applications of
game theory
The study of oligopolies (industries containing only
a few firms)
The study of cartels, e.g., OPEC
The study of externalities, e.g., using a common
resource such as a fishery
The study of military strategies
The study of international negotiations
Bargaining
 Auction
Games of incomplete information
First Price Sealed Bid Auction
Buyers simultaneously submit their bids
Buyers valuations of the good unknown to each other
Highest Bidder wins and gets the good at the amount he bid
Nash Equilibrium: Each person would bid less than what the
good is worth to you
Second Price Sealed Bid Auction
Same rules
Exception Winner pays the second highest bid and gets
the good
Nash equilibrium: Each person exactly bids the goods
valuation
 Second price auction
Suppose you value an item at 100
You should bid 100 for the item
If you bid 90
Someone bids more than 100: you lose anyway
Someone bids less than 90: you win anyway and pay second-
price
Someone bids 95: you lose; you could have won by paying 95

If you bid 110

Someone bids more than 11o: you lose anyway
Someone bids less than 100: you win anyway and pay
second-price
Someone bids 105: you win; but you pay 105, i.e., 5 more
than what you value
 Game programming
Counting game does not depend on opponents
choice
Tic-tac-toe, chess, etc. depend on opponents moves
You want a move that has the best chance of winning
However, chances of winning depend on opponents
subsequent moves
You choose a move where the worst-case winning
chance (opponents best play) is the best: max-
min
Minmax principle says that this strategy is equal to
opponents min-max strategy
The worse your opponents best move is, the better is your move
Thank you