Adversarial search

Game Playing Progress

 Checkers: 1950: First computer player. 1994: First
computer champion: Chinook ended 40-year-reign
of human champion Marion Tinsley using complete
8-piece endgame. 2007: Checkers solved!

 Chess: 1997: Deep Blue defeats human champion

Gary Kasparov in a six-game match. Deep Blue
examined 200M positions per second, used very
sophisticated evaluation and undisclosed methods
for extending some lines of search up to 40 ply.
Current programs are even better, if less historic.

 Go: AlphaGo defeats human in 2016. Uses Monte

Carlo Tree Search and learned evaluation function.

 Pacman

 Why play games? Helps track progress in AI

 Types of Games
 Many different kinds of games!

 Axes:
 Deterministic or stochastic?
 One, two, or more players?
 Zero sum?
 Perfect information (can you see the state)?

 Want algorithms for calculating a strategy (policy) which recommends a

move from each state
 Deterministic Games
 Many possible formalizations, one is:
 States: S (start at s0)
 Players: P={1...N} (usually take turns)
 Actions: A (may depend on player / state)
 Transition Function: SxA  S
 Terminal Test: S  {t,f}
 Terminal Utilities: SxP  R

 Solution for a player is a policy: S  A

 Zero-Sum Games

 Zero-Sum Games  General Games

 Agents have opposite utilities (values on  Agents have independent utilities (values on
outcomes) outcomes)
 Lets us think of a single value that one  Cooperation, indifference, competition, and
maximizes and the other minimizes more are all possible
 Adversarial, pure competition  More later on non-zero-sum games
Solving Zero-Sum Games
 Single-Agent Search Trees

2 0 … 2 6 … 4 6
 Value of a State
Value of a state: Non-Terminal States:
The best achievable
outcome (utility)
from that state

2 0 … 2 6 … 4 6
Terminal States:
 Adversarial Game Trees

-20 -8 … -18 -5 … -10 +4 -20 +8

 Minimax Values
States Under Agent’s Control: States Under Opponent’s Control:

-8 -5 -10 +8

Terminal States:
Tic-Tac-Toe Game Tree
 Adversarial Search (Minimax)
 Deterministic, zero-sum games: Minimax values:
computed recursively
 Tic-tac-toe, chess, checkers
 One player maximizes result 5 max
 The other minimizes result

2 5 min
 Minimax search:
 A state-space search tree
 Players alternate turns
 Compute each node’s minimax value: 8 2 5 6

the best achievable utility against a

Terminal values:
rational (optimal) adversary part of the game
 Minimax Implementation

def max-value(state): def min-value(state):

initialize v = -∞ initialize v = +∞
for each successor of for each successor of
state: state:
v = max(v, min-value(successor)) v = min(v, max-value(successor))
return v return v
 Minimax Implementation
def value(state):
if the state is a terminal state: return the state’s utility
if the next agent is MAX: return max-value(state)
if the next agent is MIN: return min-value(state)

def max-value(state): def min-value(state):

initialize v = -∞ initialize v = +∞
for each successor of for each successor of
state: state:
v = max(v, value(successor)) v = min(v, value(successor))
return v return v
 Minimax Example

3 12 8 2 4 6 14 5 2
 Minimax Properties

max

min

10 10 9 100

Optimal against a perfect player. Otherwise?

[Demo: min vs exp (L6D2, L6D3)]

 Minimax Efficiency
 How efficient is minimax?
 Just like (exhaustive) DFS
 Time: O(bm)
 Space: O(bm)

 Example: For chess, b  35, m 

100 For Go, b  250-300, m  150
 Exact solution is completely infeasible
 But, do we need to explore the whole
tree?
Overcoming Resource Limits
Game Tree Pruning
 Minimax Example

3 12 8 2 4 6 14 5 2
 Minimax Pruning

3 12 8 2 14 5 2
 Alpha-Beta Pruning
 General configuration (MIN version)
 We’re computing the MIN-VALUE at some node n MAX
 We’re looping over n’s children
 n’s estimate of the childrens’ min is dropping MIN a
 Who cares about n’s value? MAX
 Let a be the best value that MAX can get at any choice
point along the current path from the root
 If n becomes worse than a, MAX will avoid it, so we can MAX
stop considering n’s other children (it’s already bad
enough that it won’t be played) MIN n

 MAX version is symmetric

 Alpha-Beta Implementation

α: MAX’s best option on path to root

β: MIN’s best option on path to root

def max-value(state, α, β): def min-value(state , α, β):

initialize v = -∞ initialize v = +∞
for each successor of for each successor of
state: state:
v = max(v, value(successor, α, β)) if v v = min(v, value(successor, α, β)) if v
≥ β return v ≤ α return v
α = max(α, v) β = min(β, v)
return v return v
 Alpha-Beta Pruning Properties
 This pruning has no effect on minimax value computed for the root!

 Values of intermediate nodes might be wrong

 Important: children of the root may have the wrong value max
 So the most naïve version won’t let you do action selection

min
 Good child ordering improves effectiveness of pruning

 With “perfect ordering”:

 Time complexity drops to O(bm/2) 10 10 0
 Doubles solvable depth!
 Full search of, e.g. chess, is still hopeless…

 This is a simple example of metareasoning (computing about what to compute)

Alpha-Beta Quiz
Alpha-Beta Quiz 2
Overcoming Resource Limits
 Limiting Depth
 Problem: In realistic games, cannot search to leaves! max
4
 Solution: Depth-limited search -2 4 min
 Instead, search only to a limited depth in the tree
-1 -2 4 9
 Replace terminal utilities with an evaluation function for
non-terminal positions

 Example:
 Suppose we have 100 seconds, can explore 10K nodes /
sec
 So can check 1M nodes per move
 - reaches about depth 8 – decent chess program

 Guarantee of optimal play is gone

 More plies makes a BIG difference ? ? ? ?

 Use iterative deepening for an anytime algorithm

Why Pacman Starves (d=2)
Why Pacman Starves (d=2)
 Why Pacman Starves (d=2)

 A danger of replanning agents!

 He knows his score will go up by eating the dot now (west, east)
 He knows his score will go up just as much by eating the dot later (east, west)
 There are no point-scoring opportunities after eating the dot (within the horizon, two here)
 Therefore, waiting seems just as good as eating: he may go east, then back west in the next
round of replanning!
Evaluation Functions
 Evaluation Functions
 Evaluation functions score non-terminals in depth-limited search

 Ideal function: returns the actual minimax value of the position

 In practice: typically weighted linear sum of features:

 e.g. f1(s) = (num white queens – num black queens), etc.

 Evaluation for Pacman

[Demo: thrashing d=2, thrashing d=2 (fixed evaluation function), smart ghosts coordinate (L6D6,7,8,10)]
 Depth Matters
 Evaluation functions are always
imperfect
 The deeper in the tree the
evaluation function is buried, the
less the quality of the evaluation
function matters
 An important example of the
tradeoff between complexity of
features and complexity of
computation

[Demo: depth limited (L6D4, L6D5)]

 What we did today
 Introduced multi-agent games
 Come up in many places in AI
 Different types of games (focused on zero-sum today)
 Reviewed single-agent search trees (from previous lectures)
 Value of a state is an important concept that will come up in the future
 Modified search trees to include opponent actions
 Assumed opponent acts in a way that is worst for you
 Called this Minimax Search
 Looked at efficiency (not good) and suggested two ways to improve it
 Alpha-Beta pruning (exact, some gains but not huge)
 Limiting tree depth (big gains, but not exact and needs heuristics)