Tic-Tac-Toe Game Play with MCTS

main.py

#!/usr/bin/env python

import numpy as np

# from github.com/int8/monte-carlo-tree-search.git
from mctspy.tree.nodes import TwoPlayersGameMonteCarloTreeSearchNode
from mctspy.tree.search import MonteCarloTreeSearch
from mctspy.games.examples.tictactoe import TicTacToeGameState, TicTacToeMove

initial_board = np.zeros((3,3))
initial_state = TicTacToeGameState(state=initial_board, next_to_move=1)  # 1/-1
current = TwoPlayersGameMonteCarloTreeSearchNode(state=initial_state)
SIMULATIONS = 1600

def coded(c):
    return 'O' if c == 1 else ('X' if c == -1 else ' ')

def print_board(board):
    for row in board:
        line = [coded(e) for e in row]
        print('|{}|'.format(''.join(line)))


class Player(object):
    def __init__(self, index, state):
        self.move_index = index
        self.node = TwoPlayersGameMonteCarloTreeSearchNode(state)

    @property
    def state(self):
        return self.node.state

    def my_action(self):
        assert self.move_index == self.node.state.next_to_move
        best = MonteCarloTreeSearch(self.node).best_action(SIMULATIONS)
        self.node = best

    def other_action(self, state):
        while not self.node.is_fully_expanded():
            self.node.expand()

        for c in self.node.children:
            def _eq(a, b):
                return np.all(a.board == b.board) and a.next_to_move == b.next_to_move

            if _eq(c.state, state):
                self.node = c
                self.node.parent = None
                break
        else:
            raise ValueError('cannot reach state')


class RealPlayer(object):
    def __init__(self, index, state):
        self.move_index = index
        self._state = state

    @property
    def state(self):
        return self._state

    def my_action(self):
        assert self.move_index == self._state.next_to_move
        while True:
            try:
                line = input('\nwait for your input: ')
                cords = [int(e) for e in line.split()]
                move = TicTacToeMove(cords[0], cords[1], self.move_index)
                self._state = self._state.move(move)
                break
            except (ValueError,) as e:
                print('try again...')
                continue

    def other_action(self, state):
        self._state = state


# Init
print_board(current.state.board)

# Gaming
#player = [Player(1, initial_state), Player(-1, initial_state)]
player = [RealPlayer(1, initial_state), Player(-1, initial_state)]
state = initial_state

while not state.is_game_over():
    current_player, player = player[0], player[1:]
    player.append(current_player)
    current_player.my_action()
    state = current_player.state
    print('\nPlayer[{}] move:'.format(coded(-1*state.next_to_move)))
    print_board(state.board)
    player[0].other_action(state)

# Result
res = state.game_result
if res == 0:
    print('\nNobody wins')
else:
    print('\nPlayer[{}] wins'.format(coded(-1*state.next_to_move)))