ADVERSARIAL SEARCH

MINIMAX ALGORITHM WITH ALPHA BETA PRUNING

Search algorithms for game playing
• Higher branching factor, and
• number and type of moves,
• timing
• Contingencies introduced by other agents
makes for an optimal path infeasible/inappropriate for the
search algorithms
Game-playing research has therefore spawned a number of
interesting ideas on how to make the best possible use of time.
E.g Pruning – ignoring portions of the search tree that
makes no difference to the final choice, Heuristic evaluation
functions approximating utility functions
ALPHA-BETA PRUNING
• When applied to a standard minimax tree, it returns the same move as minimax
would, but prunes away branches that cannot possibly influence the final
decision
• Example

X Y
ALPHA-BETA PRUNING
• When applied to a standard minimax tree, it returns the same move as minimax
would, but prunes away branches that cannot possibly influence the final
decision
• Example

MIN(2, X, Y)

X Y
ALPHA-BETA PRUNING
• When applied to a standard minimax tree, it returns the same move as minimax
would, but prunes away branches that cannot possibly influence the final
decision
• Example
MAX(3, Z, 2) = 3
Z = MIN(2, X, Y) which is < 2

MINIMAX DECISION ARE

MIN(2, X, Y) INDEPENDENT OF THE
VALUES OF THE PRUNED LEAVES X
AND Y.

X Y
ALPHA BETA PRUNING
• APPLIED TO TREES OF ANY DEPTH, AND IT IS OFTEN POSSIBLE TO PRUNE ENTIRE SUBTREES
RATHER THAN JUST LEAVES
• General Principle
 Alpha – Beta Pruning Parameters
• Minimax search is depth-first, so at any one time we just have to consider the
nodes along a single path in the tree. Alpha–beta pruning gets its name from the
following two parameters that describe bounds on the backed-up values that
appear anywhere along the path:

α = the value of the best (i.e., highest-value) choice we have found so far at any choice
point along the path for MAX.
β = the value of the best (i.e., lowest-value) choice we have found so far at any choice
point along the path for MIN.
Parameters Update and pruning based on
parameters
• Alpha–beta search updates the values of α and β as it goes
along and prunes the remaining branches at a node (i.e.,
terminates the recursive call) as soon as the value of the
current node is known to be worse than the current α or β
value for MAX or MIN, respectively.
• MIN-VALUE returns when a result is found less than α else
updates β with the minimum value found
• MAX-VALUE returns when a result is found greater than β
else updates α with the maximum value found
S.No Node Type v α β
Example 1- Solution
1 A Max -∞ -∞ ∞
v – local variable
2 B Min +∞ -∞ ∞
α, β – function arguments
3 T1 T 3 -∞ ∞
4 B Min 3 -∞ 3
5 T2 T 12 -∞ 3
6 B Min 3 -∞ 3
7 T3 T 8 -∞ 3
8 B Min 3 -∞ 3
9 A Max 3 3 ∞
10 C Min +∞ 3 ∞
S.No Node Type V α β
11 T4 T 2 3 ∞
18 D Min 5 3 5
12 C Min 2 3 ∞ (v<α)
19 T9 T9 2 3 5
13 A Max 3 3 ∞
20 D Min 2 3 5 (v<α)
14 D Min +∞ 3 ∞
21 A Max 3 3 ∞
15 T7 T 14 3 ∞
16 D Min 14 3 ∞
17 T8 T 5 3 ∞
Example 1- Solution contd…

OPTIMAL MOVE
v=3
α=3
β=+∞
v=2
v=3 α=3
α = -∞ v=2 β=5
β=3 α=3
β=+∞
Example 2
S.No Node Type v α β Example 2 - Solution

1 A Max -∞ -∞ +∞
2 B Min +∞ -∞ +∞
3 D Max -∞ -∞ +∞
4 H T 1 -∞ +∞
5 D Max 1 1 +∞
6 I T 2 1 +∞
7 D Max 2 2 +∞
8 B Min 2 -∞ 2
9 E Max -∞ -∞ 2
10 J T 5 -∞ 2 S.No Node Type v α β
11 E Max 5 -∞ 2 (v>β) 18 M T 1 0 2
12 B Min 2 -∞ 2 19 F Max 1 1 +∞
13 A Max 2 2 +∞ 20 C Min 1 2 1 (v<α)
14 C Min +∞ 2 +∞ 21 A Max 2 2 +∞
15 F Max -∞ 2 +∞
16 L T 0 2 +∞
17 F Max 0 2 +∞
Example 2 Solution contd…

Optimal move
Performance of Minimax algorithm with and
without Alpha Beta Pruning
• Time Complexity of search using minimax is O(bm)
• Effectiveness of pruning depends on the order of the
child nodes in the tree With an accurate ordering
(best first search) of the successors efficiency is
O(bm/2) (i.e effective branching factor of )
• With a random choice of successors the efficiency was
found to be