ass3: implement alpha-beta pruning

assignment3/tic_tac_toe.py

@@ -1,3 +1,4 @@
+import typing
 from copy import deepcopy
 
 State = tuple[int, list[list[int | None]]]  # Tuple of player (whose turn it is),
@@ -76,3 +77,54 @@ class Game:
                 print("The game is a draw")
         else:
             print(f"It is P{self.to_move(state)+1}'s turn to move")
+
+
+def alpha_beta_search(game: Game, state: State) -> Action | None:
+    _, result = max_value(game, state, float("-inf"), float("+inf"))
+    return result
+
+
+def max_value(
+    game: Game, state: State, alpha: float, beta: float
+) -> tuple[float, Action | None]:
+    if game.is_terminal(state):
+        return game.utility(state, player), None
+    v, move = float("-inf"), float("-inf")
+    for a in game.actions(state):
+        v2, a2 = min_value(game, game.result(state, a), alpha, beta)
+        if v2 > v:
+            v, move = v2, a
+            alpha = max(alpha, v)
+        if v >= beta:
+            return v, typing.cast(Action, move)
+    return v, typing.cast(Action, move)
+
+
+def min_value(
+    game: Game, state: State, alpha: float, beta: float
+) -> tuple[float, Action | None]:
+    if game.is_terminal(state):
+        return game.utility(state, player), None
+    v, move = float("+inf"), float("+inf")
+    for a in game.actions(state):
+        v2, a2 = max_value(game, game.result(state, a), alpha, beta)
+        if v2 < v:
+            v, move = v2, a
+            alpha = min(beta, v)
+        if v <= alpha:
+            return v, typing.cast(Action, move)
+    return v, typing.cast(Action, move)
+
+
+game = Game()
+
+state = game.initial_state()
+game.print(state)
+while not game.is_terminal(state):
+    player = game.to_move(state)
+    action = alpha_beta_search(game, state)  # The player whose turn it is
+    # is the MAX player
+    print(f"P{player+1}'s action: {action}")
+    assert action is not None
+    state = game.result(state, action)
+    game.print(state)