monte_carlo.py

import random
import numpy as np
from games.games import AbstractGame

def playout_value(game):
    if game.over():
        return -game.score()
    
    move = random.choice(game.valid_moves())
    game.make_move(move)
    value = -playout_value(game))
    game.undo_move()
	
    return value

#Finds the expected value of a game by running random simulations
def monte_carlo_value(game, N=100):
    scores = [playout_value(game) for i in range(0, N)]
    return np.mean(scores)

# Chooses best valued move to play using Monte Carlo Tree search.
def ai_best_move(game):
	
    action_dict = {}
    for move in game.valid_moves():
        game.make_move(move)
        action_dict[move] = -monte_carlo_value(game)
        game.undo_move()

    return max(action_dict, key=action_dict.get)