def __run__(self, player1, player2):
"""
Runs an episode of the game
:param player1:
:param player2:
:return: The original color of the winning player
"""
self.board = OthelloBoard()
players = player1, player2
if self.gui:
self.gui.show_game(self.board)
while True:
if len(self.board.get_valid_moves(players[0].color)) > 0:
move = players[0].get_move(self.board.copy())
self.board.apply_move(move, players[0].color)
if self.gui:
self.gui.flash_move(move, players[0].color)
self.gui.update(self.board, players[1])
winner = self.board.game_won()
if winner is not None:
if self.gui:
self.gui.show_winner(winner, self.board)
return config.get_label_from_winner_color(
player1, player2, winner)
players = list(reversed(players))
def test_Board_CountStones(self):
board = OthelloBoard()
self.assertEqual((2, 2), board.count_stones())
board.apply_move((2, 3), config.BLACK)
board.apply_move((2, 2), config.WHITE)
board.apply_move((2, 1), config.BLACK)
board.apply_move((1, 1), config.WHITE)
self.assertEqual((4, 4), board.count_stones())
def generate_supervised_training_data(cls, games, labeling_strategy):
"""
Generates training data by applying random moves to a board and labeling each sample with the move that :param labeling_strategy would have taken given the board.
:param games: The number of games to be simulated
:param labeling_strategy: The strategy used to label each sample. The label equals labeling_strategy.get_move(board)
:return: a list of tuples(board_sample, move_label)
"""
labeling_strategy.color = cls.config.BLACK
generator = RandomPlayer()
color_iterator = OthelloBaseExperiment.AlternatingColorIterator()
start = datetime.now()
training_set = []
for game in range(games):
board = OthelloBoard()
while board.game_won() is None:
# generate training pair
expert_move = labeling_strategy.get_move(board)
training_set.append((board.copy(), expert_move))
# prepare for next sample
color = color_iterator.__next__()
generator.color = color
move = generator.get_move(board)
if move is not None:
board.apply_move(move, color)
print("Generated %s training pairs form %s games in %s" %
(len(training_set), games, datetime.now() - start))
return training_set
def test_Board_ApplyIllegalMove(self):
board = OthelloBoard()
board.apply_move((2,3), config.BLACK)
self.assertEqual(board.illegal_move, None)
board.apply_move((1, 1), config.BLACK)
self.assertEqual(board.illegal_move, config.BLACK)
def test_Board_GameWon(self):
# Case 1: Full board
board = OthelloBoard()
self.assertIsNone(board.game_won(), msg="Empty Board")
board.apply_move((2, 3), config.BLACK)
board.apply_move((2, 2), config.WHITE)
board.apply_move((2, 1), config.BLACK)
board.apply_move((1, 1), config.WHITE)
board.apply_move((5, 4), config.BLACK)
board.apply_move((5, 5), config.WHITE)
board.apply_move((5, 6), config.BLACK)
board.apply_move((6, 6), config.WHITE)
self.assertIsNone(board.game_won(), msg="Empty Board")
for col in range(len(board.board)):
for tile in range(len(board.board)):
if board.board[col, tile] == config.EMPTY:
board.board[col, tile] = config.BLACK
board.legal_moves = {} # This is required because moves were directly set to the board instead of using apply_move
self.assertEqual(board.game_won(), config.BLACK, msg="Black wins by stone count")
# Case 2: No valid moves
board = OthelloBoard()
board.apply_move((3, 2), config.BLACK)
board.apply_move((4, 5), config.BLACK)
self.assertEqual(board.game_won(), config.BLACK, msg="Black wins by stone count after no players could perform any legal moves")
# Case 3: Regular, deterministic game
for i in range(32):
game = Othello((DeterministicPlayer(), DeterministicPlayer()))
game.run_simulations(1)
winner = game.board.game_won()
self.assertEqual(winner, config.WHITE, "Winner of deterministic game was not Black")
def test_Board_applyValidMoves(self):
board = OthelloBoard()
self.assertEqual(board.get_valid_moves(config.BLACK), {(2, 3), (3, 2), (4, 5), (5, 4)}, msg="Valid moves incorrect")
self.assertEqual(board.get_valid_moves(config.WHITE), {(2, 4), (4, 2), (3, 5), (5, 3)}, msg="Valid moves incorrect")
board.apply_move((3, 2), config.BLACK)
self.assertEqual(board.get_valid_moves(config.BLACK), {(4, 5), (5, 4), (5, 5)}, msg="Valid moves incorrect")
self.assertEqual(board.get_valid_moves(config.WHITE), {(4, 2), (2, 4), (2, 2)}, msg="Valid moves incorrect")
b = [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((2, 2), config.WHITE)
b = [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((1, 2), config.BLACK)
b = [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((1, 1), config.WHITE)
b = [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((1, 0), config.BLACK)
b = [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((0, 0), config.WHITE)
b = [[-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((4, 5), config.BLACK)
b = [[-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((5, 5), config.WHITE)
b = [[-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, -1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((6, 5), config.BLACK)
b = [[-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((6, 6), config.WHITE)
b = [[-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, -1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 1.0, -1.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((6, 7), config.BLACK)
b = [[-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, -1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]
self.assertTrue((board.board == b).all())
board.apply_move((7, 7), config.WHITE)
b = [[-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [1.0, -1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, -1.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 1.0, -1.0, 1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, -1.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 1.0, -1.0, 1.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -1.0]]
self.assertTrue((board.board == b).all())
def test_Board_Representation(self):
iterator = OthelloBaseExperiment.AlternatingColorIterator()
boards = []
inverses = []
for i in range(10):
board = OthelloBoard()
inverse_board = OthelloBoard()
inverse_board.board[3, 3] = config.BLACK
inverse_board.board[3, 4] = config.WHITE
inverse_board.board[4, 4] = config.BLACK
inverse_board.board[4, 3] = config.WHITE
for j in range(30):
color = iterator.__next__()
legal_moves = board.get_valid_moves(color)
if legal_moves:
move = random.choice(list(legal_moves))
board.apply_move(move, color)
boards.append(board.copy())
inverse_board.apply_move(move, board.other_color(color))
inverses.append((inverse_board.copy()))
for i in range(len(boards)):
rep = boards[i].get_representation(config.WHITE)
self.assertTrue((rep == inverses[i].board).all(), msg="Inverting board failed")
class Othello(TwoPlayerGame):
def __init__(self, players, gui=None):
super(Othello, self).__init__(players=players, config=config, gui=gui)
self.player1.color = config.BLACK
self.player2.color = config.WHITE
for player in players:
player.original_color = player.color
def __run__(self, player1, player2):
"""
Runs an episode of the game
:param player1:
:param player2:
:return: The original color of the winning player
"""
self.board = OthelloBoard()
players = player1, player2
if self.gui:
self.gui.show_game(self.board)
while True:
if len(self.board.get_valid_moves(players[0].color)) > 0:
move = players[0].get_move(self.board.copy())
self.board.apply_move(move, players[0].color)
if self.gui:
self.gui.flash_move(move, players[0].color)
self.gui.update(self.board, players[1])
winner = self.board.game_won()
if winner is not None:
if self.gui:
self.gui.show_winner(winner, self.board)
return config.get_label_from_winner_color(
player1, player2, winner)
players = list(reversed(players))
def run_simulations(self,
episodes,
switch_colors=True,
switch_players=True):
"""
Runs a number of games using the given players and returns statistics over all games run.
If both :param switch_colors and :param switch_players are set, all four possible starting positions will iterated through.
:param episodes: The number of games to run
:param switch_colors: Flag specifying whether to alternate the players colors during play
:param switch_players: Flag specifying whether to alternate the starting player
:return: The results and average losses per episode where results is a list of the original colors of the winning player ([original_winning_color])
"""
simulation_players = [self.player1, self.player2]
results = []
losses = []
for episode in range(episodes):
if switch_colors and episode != 0 and episode % 2 == 0:
simulation_players[0].color, simulation_players[
1].color = simulation_players[1].color, simulation_players[
0].color
# Alternate starting color, effectively resulting in 4 starting positions rather than 2.
if switch_players and episode != 0 and episode + 1 % 2:
simulation_players = list(reversed(simulation_players))
winner = self.__run__(simulation_players[0], simulation_players[1])
player_losses = []
for player in simulation_players:
loss = player.register_winner(winner)
if loss is not None:
player_losses.append(loss)
losses += player_losses
results.append(winner)
for player in simulation_players:
player.color = player.original_color
return results, losses