def play_game():
tree = MCTS()
board = new_domineering_board()
board.to_pretty_string()
while True:
row_col = input("enter row,col: ")
row, col = map(int, row_col.split(","))
stdout.write('You choose ({}, {})'.format(row, col))
index = conf.BOARD_Y_SIZE * (row - 1) + (col - 1)
if (board.tup[index] is not None) and (
board.is_valid_move(index + conf.BOARD_Y_SIZE)):
raise RuntimeError("Invalid move")
board = board.make_move(index)
board.to_pretty_string()
if board.terminal:
stdout.write("\nWinner is {}".format(
conf.PLAYERS_NAME[board.winner]))
break
# You can train as you go, or only at the beginning.
# Here, we train as we go, doing fifty rollouts each turn.
for _ in range(conf.TRAINING_EPOCHS):
tree.do_rollout(board)
board = tree.choose(board)
board.to_pretty_string()
if board.terminal:
stdout.write("\nWinner is {}".format(
conf.PLAYERS_NAME[board.winner]))
break
def play_game():
tree = MCTS()
board = new_tic_tac_toe_board()
print(board.to_pretty_string())
while True:
row_col = input("enter row,col: ")
row, col = map(int, row_col.split(","))
index = 3 * (row - 1) + (col - 1)
if board.tup[index] is not None:
raise RuntimeError("Invalid move")
board = board.make_move(index)
print(board.to_pretty_string())
if board.terminal:
break
# You can train as you go, or only at the beginning.
# Here, we train as we go, doing fifty rollouts each turn.
for _ in range(2):
tree.do_rollout(board)
print(tree.children)
print(len(tree.children[board]))
for b in tree.children[board]:
print(colored(b.to_pretty_string(), 'green'))
board = tree.choose(board)
print(board.to_pretty_string())
if board.terminal:
break
def play_game_ocba(budget=1000, optimum=0, n0=5, sigma_0=1):
mcts = MCTS(policy='ocba',
budget=budget,
optimum=optimum,
n0=n0,
sigma_0=sigma_0)
tree = new_tree()
for _ in range(budget):
mcts.do_rollout(tree)
next_tree = mcts.choose(tree)
return (mcts, tree, next_tree)
class myPlayer(AdvancePlayer):
def __init__(self, _id):
super().__init__(_id)
self.opponent_id = None
self.tree = MCTS()
def SelectMove(self, moves, game_state):
self.opponent_id = next(
filter(lambda ps: ps.id != self.id, game_state.players)).id
self.board = new_azul_board(game_state, self.id, self.opponent_id)
for _ in range(30):
self.tree.do_rollout(self.board)
best_board = self.tree.choose(self.board)
return best_board.game_state.lastmove
def play_game_uct(budget=1000,
exploration_weight=1,
optimum=0,
n0=2,
sigma_0=1):
mcts = MCTS(policy='uct',
exploration_weight=exploration_weight,
budget=budget,
n0=n0,
sigma_0=sigma_0)
tree = new_tree()
for _ in range(budget):
mcts.do_rollout(tree)
next_tree = mcts.choose(tree)
return (mcts, tree, next_tree)