def train(epochs=10000, print_every_n=500, ep1=0.1, ep2=0.1):
exp_number = int(ep1 * 100 + ep2 * 10)
player1 = Player(epsilon=ep1, exp_number=exp_number)
player2 = Player(epsilon=ep2, exp_number=exp_number)
judge = Judge(player1, player2)
player1_win = 0.0
player2_win = 0.0
player1_win_rate_history = np.ndarray(shape=(epochs, ))
player2_win_rate_history = np.ndarray(shape=(epochs, ))
for i in range(1, epochs + 1):
winner = judge.play(print_state=False)
if winner == 1:
player1_win += 1
if winner == -1:
player2_win += 1
if i % print_every_n == 0:
print('[Exp#%2d] Epoch %d, 1 win rate: %.02f, 2 win rate: %.02f' %
(exp_number, i, player1_win / i, player2_win / i))
player1.backup()
player2.backup()
judge.reset()
player1_win_rate_history[i - 1] = player1_win / i
player2_win_rate_history[i - 1] = player2_win / i
player1.save_policy()
player2.save_policy()
return player1_win_rate_history, player2_win_rate_history
def test_minimax_2():
new_game = Game()
new_board = Board()
new_board.board = [["?", "O", "?"], ["X", "O", "?"], ["O", "X", "X"]]
new_game.board = copy.deepcopy(new_board)
new_game.player1 = Player(1, "Human", "O")
new_game.player2 = Player(2, "Computer", "X")
assert new_game.expert_ai(new_game.board, -math.inf, math.inf,
"X") == (0, 0, 2)
def test_minimax():
new_game = Game()
new_board = Board()
new_board.board = [["O", "O", "X"], ["X", "?", "O"], ["?", "?", "X"]]
new_game.board = new_board
new_game.player1 = Player(1, "Human", "O")
new_game.player2 = Player(2, "Computer", "X")
assert new_game.expert_ai(new_game.board, -math.inf, math.inf,
"X") == (1, 2, 1)
def test_minimax_3():
new_game = Game()
new_board = Board()
new_board.board = [["?", "?", "?"], ["?", "O", "?"], ["X", "O", "X"]]
new_game.board = new_board
new_game.player1 = Player(1, "Human", "O")
new_game.player2 = Player(2, "Computer", "X")
new_game.current_turn = new_game.player2
result = new_game.expert_ai(new_game.board, -math.inf, math.inf, "X")
assert result[1] == 0
assert result[2] == 1
def next_move_and_result(player, board_string):
new_board = Board()
flat_board = board_string.split()
board_array = [flat_board[0:3], flat_board[3:6], flat_board[6:9]]
new_board.board = board_array
new_game = Game()
new_game.board = new_board
new_game.player1 = Player(1, "Human", "O")
new_game.player2 = Player(2, "Computer", "X")
result, best_row, best_column = new_game.expert_ai(new_game.board, player)
board_array[best_row][best_column] = "*"
new_board_string = "\n ".join(" ".join(row) for row in board_array)
return """
Position score: {}
{}
""".format(result, new_board_string)
def compete(turns, ep1, ep2):
exp_number = int(ep1 * 100 + ep2 * 10)
player1 = Player(epsilon=ep1, exp_number=exp_number)
player2 = Player(epsilon=ep2, exp_number=exp_number)
player1.load_policy()
player2.load_policy()
judge = Judge(player1, player2)
player1_win = 0.0
player2_win = 0.0
for _ in range(turns):
winner = judge.play()
if winner == 1:
player1_win += 1
if winner == -1:
player2_win += 1
judge.reset()
print('[Exp#%02d] %d turns, player 1 win %.05f, player 2 win %.05f' %
(exp_number, turns, player1_win / turns, player2_win / turns))
return player1_win / turns, player2_win / turns
"initial_value": 0.1,
"verbose": False
}
second_player_params = {
"train": True,
"epsilon": 0.7,
"epsilon_decay": 0.9999,
"min_epsilon": 0.,
"gamma": 1.,
"alpha": 0.4,
"initial_value": 0.1,
"verbose": False
}
first_player = Player(**first_player_params)
second_player = Player(**second_player_params)
players = {1: first_player, -1: second_player}
game = Game()
train(game,
players,
games=20000,
verbose=False,
plot=True,
window_size=1000,
seed=None)
first_player.save_state_dict('states/first_player')
second_player.save_state_dict('states/second_player')
def setUp(self):
self.player = Player("Mike", sign_x)
self.player2 = Player("Computer", sign_o)
from tic_tac_toe import Game, Player
from tabulate import tabulate
games_to_play_counter = 1000
games_to_play = games_to_play_counter
all_games = []
ai_1 = "E"
ai_2 = "A"
while games_to_play_counter > 0:
game = Game()
game.player1 = Player(1, ai_1, "X")
game.player2 = Player(2, ai_2, "O")
game.current_turn = game.player1
winner = game.ai_vs_ai()
all_games.append(winner)
games_to_play_counter -= 1
print(all_games)
wins_player_1 = all_games.count("X")
percent_wins_player_1 = wins_player_1 / games_to_play
wins_player_2 = all_games.count("O")
percent_wins_player_2 = wins_player_2 / games_to_play
ties = all_games.count("tie")
percent_ties = ties / games_to_play
print(
tabulate(
[[ai_1, "1", str(wins_player_1),
str(percent_wins_player_1)],
[ai_2, "2", str(wins_player_2),
