def recurse(state: State, alpha, beta) -> Tuple[int, object]:
"""
根据当前状态返回一个当前最佳效应和所对应的动作
:param state: 当前的状态
:param alpha: 到当前状态的最开始玩家的收益下界
:param beta: 到结束状态的最对手玩家的收益上界
:return: 返回一个元组 (utility action)
"""
is_over, winner = state.get_state_result()
if is_over:
if winner == self.player:
return 1, None
elif winner == get_opponent(self.player):
return -1, None
else:
return 0, None
available_actions = state.get_available_actions()
if state.player == self.player:
max_value = (float("-inf"), None)
for action in available_actions:
max_value = max(max_value, (recurse(
state.get_next_state(action), alpha, beta)[0], action),
key=lambda x: x[0])
alpha = max(alpha, max_value[0])
if beta <= alpha:
break
return max_value
elif state.player == get_opponent(self.player):
min_value = (float("inf"), None)
for action in available_actions:
min_value = min(min_value, (recurse(
state.get_next_state(action), alpha, beta)[0], action),
key=lambda x: x[0])
beta = min(beta, min_value[0])
if beta <= alpha:
break
return min_value
def recurse(state: State) -> Tuple[int, object]:
"""
根据当前状态返回一个当前最佳效应和所对应的动作
:param state: 当前的状态
:return: 返回一个元组 (utility action)
"""
is_over, winner = state.get_state_result()
if is_over:
if winner == state.player:
return 1, None
elif winner == get_opponent(state.player):
return -1, None
else:
return 0, None
available_actions = state.get_available_actions()
values = [
-recurse(state.get_next_state(action))[0]
for action in available_actions
]
kws = pd.Series(data=values, index=available_actions)
action = kws.idxmax()
return kws[action], action
