利用上一篇的框架，再写了个翻转棋的程序，为了调试minimax算法，花了两天的时间。
几点改进说明：
- 拆分成四个文件：board.py，player.py，ai.py，othello.py。使得整个结构更清晰，更通用，更易于维护。
- AI 的水平跟 minimax 的递归深度，以及评价函数有关。基于此，我把 minimax 和评价函数都放到 AI 类里面
- AIPlayer 使用了多重继承。继承了 Player 与 AI 两个类
Game 类中把原run函数里的生成两个玩家的部分提出来，写成一个函数
make_two_players，使得 run函数结构更清晰AI 玩家等级不要选择 0：beginer。会报错，还没调试好
board.py
'''
作者：hhh5460
时间：2017年7月1日
'''
class Board(object):
def __init__(self):
self.empty = '.'
self._board = [[self.empty for _ in range(8)] for _ in range(8)] # 规格：8*8
self._board[3][4], self._board[4][3] = 'X', 'X'
self._board[3][3], self._board[4][4] = 'O', 'O'
# 增加 Board[][] 索引语法
def __getitem__(self, index):
return self._board[index]
# 打印棋盘
def print_b(self):
board = self._board
print(' ', ' '.join(list('ABCDEFGH')))
for i in range(8):
print(str(i+1),' '.join(board[i]))
# 棋局终止
def teminate(self):
list1 = list(self.get_legal_actions('X'))
list2 = list(self.get_legal_actions('O'))
return [False, True][len(list1) == 0 and len(list2) == 0]
# 判断赢家
def get_winner(self):
s1, s2 = 0, 0
for i in range(8):
for j in range(8):
if self._board[i][j] == 'X':
s1 += 1
if self._board[i][j] == 'O':
s2 += 1
if s1 > s2:
return 0 # 黑胜
elif s1 < s2:
return 1 # 白胜
elif s1 == s2:
return 2 # 平局
# 落子
def _move(self, action, color):
x,y = action
self._board[x][y] = color
return self._flip(action, color)
# 翻子（返回list）
def _flip(self, action, color):
flipped_pos = []
for line in self._get_lines(action):
for i,p in enumerate(line):
if self._board[p[0]][p[1]] == self.empty:
break
elif self._board[p[0]][p[1]] == color:
flipped_pos.extend(line[:i])
break
for p in flipped_pos:
self._board[p[0]][p[1]] = color
return flipped_pos
# 撤销
def _unmove(self, action, flipped_pos, color):
self._board[action[0]][action[1]] = self.empty
uncolor = ['X', 'O'][color=='X']
for p in flipped_pos:
self._board[p[0]][p[1]] = uncolor
# 生成8个方向的下标数组，方便后续操作
def _get_lines(self, action):
'''说明：刚开始我是用一维棋盘来考虑的，后来改为二维棋盘。偷懒，不想推倒重来，简单地修改了一下'''
board_coord = [(i,j) for i in range(8) for j in range(8)] # 棋盘坐标
r,c = action
ix = r*8 + c
r, c = ix//8, ix%8
left = board_coord[r*8:ix] # 要反转
right = board_coord[ix+1:(r+1)*8]
top = board_coord[c:ix:8] # 要反转
bottom = board_coord[ix+8:8*8:8]
if r <= c:
lefttop = board_coord[c-r:ix:9] # 要反转
rightbottom = board_coord[ix+9:(7-(c-r))*8+7+1:9]
else:
lefttop = board_coord[(r-c)*8:ix:9] # 要反转
rightbottom = board_coord[ix+9:7*8+(7-(c-r))+1:9]
if r+c<=7:
leftbottom = board_coord[ix+7:(r+c)*8:7]
righttop = board_coord[r+c:ix:7] # 要反转
else:
leftbottom = board_coord[ix+7:7*8+(r+c)-7+1:7]
righttop = board_coord[((r+c)-7)*8+7:ix:7] # 要反转
# 有四个要反转，方便判断
left.reverse()
top.reverse()
lefttop.reverse()
righttop.reverse()
lines = [left, top, lefttop, righttop, right, bottom, leftbottom, rightbottom]
return lines
# 检测，位置是否有子可翻
def _can_fliped(self, action, color):
flipped_pos = []
for line in self._get_lines(action):
for i,p in enumerate(line):
if self._board[p[0]][p[1]] == self.empty:
break
elif self._board[p[0]][p[1]] == color:
flipped_pos.extend(line[:i])
break
return [False, True][len(flipped_pos) > 0]
# 合法走法
def get_legal_actions(self, color):
uncolor = ['X', 'O'][color=='X']
uncolor_near_points = [] # 反色邻近的空位
board = self._board
for i in range(8):
for j in range(8):
if board[i][j] == uncolor:
for dx,dy in [(-1,0),(-1,1),(0,1),(1,1),(1,0),(1,-1),(0,-1)]:
x, y = i+dx, j+dy
if 0 <= x <=7 and 0 <= y <=7 and board[x][y] == self.empty and (x, y) not in uncolor_near_points:
uncolor_near_points.append((x, y))
for p in uncolor_near_points:
if self._can_fliped(p, color):
yield p
# 测试
if __name__ == '__main__':
board = Board()
board.print_b()
print(list(board.get_legal_actions('X')))player.py
from ai import AI
'''
作者：hhh5460
时间：2017年7月1日
'''
# 玩家
class Player(object):
def __init__(self, color):
self.color = color
# 思考
def think(self, board):
pass
# 落子
def move(self, board, action):
flipped_pos = board._move(action, self.color)
return flipped_pos
# 悔子
def unmove(self, board, action, flipped_pos):
board._unmove(action, flipped_pos, self.color)
# 人类玩家
class HumanPlayer(Player):
def __init__(self, color):
super().__init__(color)
def think(self, board):
while True:
action = input("Turn to '{}'. \nPlease input a point.(such as 'A1'): ".format(self.color)) # A1~H8
r, c = action[1], action[0].upper()
if r in '12345678' and c in 'ABCDEFGH': # 合法性检查1
x, y = '12345678'.index(r), 'ABCDEFGH'.index(c)
if (x,y) in board.get_legal_actions(self.color): # 合法性检查2
return x, y
# 电脑玩家（多重继承）
class AIPlayer(Player, AI):
def __init__(self, color, level_ix=0):
super().__init__(color) # init Player
super(Player, self).__init__(level_ix) # init AI
def think(self, board):
print("Turn to '{}'. \nPlease wait a moment. AI is thinking...".format(self.color))
uncolor = ['X','O'][self.color=='X']
opfor = AIPlayer(uncolor) # 假想敌，陪练
action = self.brain(board, opfor, 4)
return actionai.py
import random
'''
作者：hhh5460
时间：2017年7月1日
'''
class AI(object):
'''
三个水平等级：初级（beginner）、中级（intermediate）、高级（advanced）
'''
def __init__(self, level_ix =0):
# 玩家等级
self.level = ['beginner','intermediate','advanced'][level_ix]
# 棋盘位置权重，参考：https://github.com/k-time/ai-minimax-agent/blob/master/ksx2101.py
self.board_weights = [
[120, -20, 20, 5, 5, 20, -20, 120],
[-20, -40, -5, -5, -5, -5, -40, -20],
[ 20, -5, 15, 3, 3, 15, -5, 20],
[ 5, -5, 3, 3, 3, 3, -5, 5],
[ 5, -5, 3, 3, 3, 3, -5, 5],
[ 20, -5, 15, 3, 3, 15, -5, 20],
[-20, -40, -5, -5, -5, -5, -40, -20],
[120, -20, 20, 5, 5, 20, -20, 120]
]
# 评估函数（仅根据棋盘位置权重）
def evaluate(self, board, color):
uncolor = ['X','O'][color=='X']
score = 0
for i in range(8):
for j in range(8):
if board[i][j] == color:
score += self.board_weights[i][j]
elif board[i][j] == uncolor:
score -= self.board_weights[i][j]
return score
# AI的大脑
def brain(self, board, opponent, depth):
if self.level == 'beginer': # 初级水平
_, action = self.randomchoice(board)
elif self.level == 'intermediate': # 中级水平
_, action = self.minimax(board, opponent, depth)
elif self.level == 'advanced': # 高级水平
_, action = self.minimax_alpha_beta(board, opponent, depth)
assert action is not None, 'action is None'
return action
# 随机选（从合法走法列表中随机选）
def randomchoice(self, board):
color = self.color
action_list = list(board.get_legal_actions(color))
return None, random.choice(action_list)
# 极大极小算法，限制深度
def minimax(self, board, opfor, depth=4): # 其中 opfor 是假想敌、陪练
'''参考：https://github.com/k-time/ai-minimax-agent/blob/master/ksx2101.py'''
color = self.color
if depth == 0:
return self.evaluate(board, color), None
action_list = list(board.get_legal_actions(color))
if not action_list:
return self.evaluate(board, color), None
best_score = -100000
best_action = None
for action in action_list:
flipped_pos = self.move(board, action) # 落子
score, _ = opfor.minimax(board, self, depth-1) # 深度优先，轮到陪练
self.unmove(board, action, flipped_pos) # 回溯
score = -score
if score > best_score:
best_score = score
best_action = action
return best_score, best_action
# 极大极小算法，带alpha-beta剪枝
def minimax_alpha_beta(self, board, opfor, depth=8, my_best=-float('inf'), opp_best=float('inf')):
'''参考：https://github.com/k-time/ai-minimax-agent/blob/master/ksx2101.py'''
color = self.color
if depth == 0:
return self.evaluate(board, color), None
action_list = list(board.get_legal_actions(color))
if not action_list:
return self.evaluate(board, color), None
best_score = my_best
best_action = None
for action in action_list:
flipped_pos = self.move(board, action) # 落子
score, _ = opfor.minimax_alpha_beta(board, self, depth-1, -opp_best, -best_score) # 深度优先，轮到陪练
self.unmove(board, action, flipped_pos) # 回溯
score = -score
if score > best_score:
best_score = score
best_action = action
if best_score > opp_best:
break
return best_score, best_actionothello.py
from board import Board
from player import HumanPlayer, AIPlayer
'''
作者：hhh5460
时间：2017年7月1日
'''
# 游戏
class Game(object):
def __init__(self):
self.board = Board()
self.current_player = None
# 生成两个玩家
def make_two_players(self):
ps = input("Please select two player's type:\n\t0.Human\n\t1.AI\nSuch as:0 0\n:")
p1, p2 = [int(p) for p in ps.split(' ')]
if p1 == 1 or p2 == 1: # 至少有一个AI玩家
level_ix = int(input("Please select the level of AI player.\n\t0: beginner\n\t1: intermediate\n\t2: advanced\n:"))
if p1 == 0:
player1 = HumanPlayer('X')
player2 = AIPlayer('O', level_ix)
elif p2 == 0:
player1 = AIPlayer('X', level_ix)
player2 = HumanPlayer('O')
else:
player1 = AIPlayer('X', level_ix)
player2 = AIPlayer('O', level_ix)
else:
player1, player2 = HumanPlayer('X'), HumanPlayer('O') # 先手执X，后手执O
return player1, player2
# 切换玩家（游戏过程中）
def switch_player(self, player1, player2):
if self.current_player is None:
return player1
else:
return [player1, player2][self.current_player == player1]
# 打印赢家
def print_winner(self, winner): # winner in [0,1,2]
print(['Winner is player1','Winner is player2','Draw'][winner])
# 运行游戏
def run(self):
# 生成两个玩家
player1, player2 = self.make_two_players()
# 游戏开始
print('\nGame start!\n')
self.board.print_b() # 显示棋盘
while True:
self.current_player = self.switch_player(player1, player2) # 切换当前玩家
action = self.current_player.think(self.board) # 当前玩家对棋盘进行思考后，得到招法
if action is not None:
self.current_player.move(self.board, action) # 当前玩家执行招法，改变棋盘
self.board.print_b() # 显示当前棋盘
if self.board.teminate(): # 根据当前棋盘，判断棋局是否终止
winner = self.board.get_winner() # 得到赢家 0,1,2
break
self.print_winner(winner)
print('Game over!')
self.board.print_history()
if __name__ == '__main__':
Game().run()
效果图
