def fight(ann, best, dump=False):
board = game.getEmptyBoard()
current = 'O'
action = random.randint(0, game.BOARD_SIZE * game.BOARD_SIZE - 1)
board, won = game.move(board, action, current)
if dump:
print(game.render(board))
current = 'O' if current == 'X' else 'X'
while True:
entry = state_lists_to_batch([board], current)
probs = best(entry)[0] if current == 'O' else ann(entry)[0]
while True:
action = np.argmax(probs)
if action not in game.possible_moves(board):
probs[action] = -np.inf
else:
break
board, won = game.move(board, action, current)
if dump:
print(game.render(board))
if len(game.possible_moves(board)) == 0:
return None
if won:
break
current = 'O' if current == 'X' else 'X'
return best if current == 'O' else ann
def expectimax(game, depth, value_function):
""" Searches the game tree down to given depth
to find the maximum expected value of
value_function. value_function takes the game
at the leaf nodes and returns a number value.
Returns tuple (value, move) with the expected value
and the maximizing move.
"""
# base case: reached depth
if depth == 0:
return value_function(game), 'null'
possible = possible_moves(game)
# base case: no more moves
if not possible:
return value_function(game), 'null'
# get the possible outcomes of each possible move
expanded = [(move, move_made_all(game, move)) for move in possible]
move_value = []
# calculate expected value of each possible move
for move, games in expanded:
value = sum(p * expectimax(g, depth - 1, value_function)[0] for p, g in games)
move_value.append((value, move))
return max(move_value)
def expectimax(game, depth, value_function):
""" Searches the game tree down to given depth
to find the maximum expected value of
value_function. value_function takes the game
at the leaf nodes and returns a number value.
Returns tuple (value, move) with the expected value
and the maximizing move.
"""
# base case: reached depth
if depth == 0:
return value_function(game), 'null'
possible = possible_moves(game)
# base case: no more moves
if not possible:
return value_function(game), 'null'
# get the possible outcomes of each possible move
expanded = [(move, move_made_all(game, move)) for move in possible]
move_value = []
# calculate expected value of each possible move
for move, games in expanded:
value = sum(p * expectimax(g, depth - 1, value_function)[0]
for p, g in games)
move_value.append((value, move))
return max(move_value)
def get_move_randomly(active_player, board):
piece_location_options = locations_of_pieces_with_valid_moves(
active_player, board)
if not piece_location_options:
raise GameOver()
location_of_piece_to_move = random.choice(piece_location_options)
move_options = possible_moves(location_of_piece_to_move, active_player,
board)
return random.choice(move_options)
def get_move_from_stdin(active_player, board):
piece_location_options = locations_of_pieces_with_valid_moves(
active_player, board)
if not piece_location_options:
raise GameOver()
location_of_piece_to_move = get_choice_from_stdin(piece_location_options,
pretty_location_str)
move_options = possible_moves(location_of_piece_to_move, active_player,
board)
return get_choice_from_stdin(move_options, pretty_move_str)
def best_move_silvia(board):
moves = possible_moves(board)
move_std = dict(zip(moves, [0] * len(moves)))
for move in moves:
new_board, _ = move_made((board, 0), move)
move_std[move] = get_neighbor_devsum(new_board)
move = min(move_std, key=move_std.get)
return move
def best_move_silvia(board):
moves = possible_moves(board)
move_std = dict(zip(moves, [0] * len(moves)))
for move in moves:
new_board, _ = move_made((board, 0), move)
move_std[move] = get_neighbor_devsum(new_board)
move = min(move_std, key=move_std.get)
return move
def ai_mode(number_of_games=1):
""" Just let the AI play a number of times
without showing the UI.
"""
scores = []
for i in range(number_of_games):
board, score = new_game(4)
possible = possible_moves(board)
while possible:
move = game_ai.expectimax_move((board, score), 'gradient')
score += make_move(board, move)
possible = possible_moves(board)
scores.append(score)
print("Game {0}: Score = {1}".format(i+1, score))
print()
print("Mean score = " + str(sum(scores)/float(len(scores))))
def main(playmode):
view = GameView()
board, score = new_game(4)
view.draw(board, score)
clock = pygame.time.Clock()
while True:
clock.tick(20) # limit fps
redraw = False
# get move from player
for event in pygame.event.get():
if event.type == QUIT:
print("Exiting...")
return
if event.type == KEYDOWN and playmode:
if event.key == K_LEFT:
score += make_move(board, 'left')
redraw = True
if event.key == K_RIGHT:
score += make_move(board, 'right')
redraw = True
if event.key == K_UP:
score += make_move(board, 'up')
redraw = True
if event.key == K_DOWN:
score += make_move(board, 'down')
redraw = True
if redraw:
view.draw(board, score)
# get move from ai
if not playmode:
best_move = game_ai.expectimax_move((board, score), 'gradient')
# make the move and redraw
score += make_move(board, best_move)
view.draw(board, score)
pygame.time.wait(0)
# check if it is game over
if len(possible_moves(board)) == 0:
print("Game over! Score: " + str(score) + " Max square: " +
str(max_square(board)))
print("Starting new game...\n")
pygame.time.wait(3000)
pygame.event.clear()
board, score = new_game(4)
view.draw(board, score)
def ai_mode(number_of_games=1):
""" Just let the AI play a number of times
without showing the UI.
"""
scores = []
for i in range(number_of_games):
board, score = new_game(4)
possible = possible_moves(board)
while possible:
move = game_ai.expectimax_move((board, score), 'gradient')
score += make_move(board, move)
possible = possible_moves(board)
scores.append(score)
print("Game {0}: Score = {1}".format(i + 1, score))
print()
print("Mean score = " + str(sum(scores) / float(len(scores))))
def main(playmode):
view = GameView()
board, score = new_game(4)
view.draw(board,score)
clock = pygame.time.Clock()
while True:
clock.tick(20) # limit fps
redraw = False
# get move from player
for event in pygame.event.get():
if event.type == QUIT:
print("Exiting...")
return
if event.type == KEYDOWN and playmode:
if event.key == K_LEFT:
score += make_move(board, 'left')
redraw = True
if event.key == K_RIGHT:
score += make_move(board, 'right')
redraw = True
if event.key == K_UP:
score += make_move(board, 'up')
redraw = True
if event.key == K_DOWN:
score += make_move(board, 'down')
redraw = True
if redraw:
view.draw(board, score)
# get move from ai
if not playmode:
best_move = game_ai.expectimax_move((board, score), 'gradient')
# make the move and redraw
score += make_move(board, best_move)
view.draw(board, score)
pygame.time.wait(0)
# check if it is game over
if len(possible_moves(board)) == 0:
print("Game over! Score: " + str(score) + " Max square: " + str(max_square(board)))
print("Starting new game...\n")
pygame.time.wait(3000)
pygame.event.clear()
board, score = new_game(4)
view.draw(board, score)
def get_move_from_stdin_old_style(active_player, board):
location_of_piece_to_move = None
reason_cant_move = "Select a piece to move."
while reason_cant_move:
print(reason_cant_move)
try:
location_of_piece_to_move = parse_location_from_str(input("> "))
reason_cant_move = reason_piece_at_location_cant_move(
location_of_piece_to_move, active_player, board)
except InvalidLocationString:
reason_cant_move = "That's not a valid square. Try something like \"h6\"..."
options = possible_moves(location_of_piece_to_move, active_player, board)
return get_choice_from_stdin(options, pretty_move_str)
def max_score_search(game, depth, val):
""" Simple AI that searches the game tree down
a given depth and returns the maximum value
of the leaf nodes at level depth. Value
is given by the function val(game)
Signature: game, int depth, function val => int score
"""
if depth == 0:
return val(game)
new_games = [move_made(game, move) for move in possible_moves(game)]
if len(new_games) == 0:
return 0
return max(max_score_search(g, depth - 1, val) for g in new_games)
def max_score_search(game, depth, val):
""" Simple AI that searches the game tree down
a given depth and returns the maximum value
of the leaf nodes at level depth. Value
is given by the function val(game)
Signature: game, int depth, function val => int score
"""
if depth == 0:
return val(game)
new_games = [move_made(game, move) for move in possible_moves(game)]
if len(new_games) == 0:
return 0
return max(max_score_search(g, depth - 1, val) for g in new_games)
def best_move(game, method='score'):
"""
val should be function game -> int
"""
board, score = game
if method == 'score':
def val(g):
return g[1]
elif method == 'empty':
val = empty_squares
elif method == 'gradient':
val = gradient_value
else:
print('Invalid method given to best_move function')
exit(1)
if score < 10000:
depth = 4
elif score < 20000:
depth = 5
elif score < 30000:
depth = 7
else:
depth = 7
moves = possible_moves(board)
# dict with score for each move
move_score = dict(zip(moves, [0] * len(moves)))
# search for max score for each move
for move in moves:
move_score[move] += max_score_search(move_made(game, move), depth, val)
move = max(move_score, key=move_score.get)
return move
def best_move(game, method='score'):
"""
val should be function game -> int
"""
board, score = game
if method == 'score':
def val(g):
return g[1]
elif method == 'empty':
val = empty_squares
elif method == 'gradient':
val = gradient_value
else:
print('Invalid method given to best_move function')
exit(1)
if score < 10000:
depth = 4
elif score < 20000:
depth = 5
elif score < 30000:
depth = 7
else:
depth = 7
moves = possible_moves(board)
# dict with score for each move
move_score = dict(zip(moves, [0] * len(moves)))
# search for max score for each move
for move in moves:
move_score[move] += max_score_search(move_made(game, move), depth, val)
move = max(move_score, key=move_score.get)
return move
