backgammon/game.py

from board import Board
from bot import Bot
from restore_bot import RestoreBot
import numpy as np

from cup import Cup


class Game:
    def __init__(self):
        self.board = Board.initial_state

        self.p1 = Bot(1)
        self.p2 = RestoreBot(1)
        self.cup = Cup()

    def roll(self):
        return self.cup.roll()

    def roll_and_find_best_for_bot(self):
        roll = self.roll()
        move_and_val = self.p1.make_move(self.board, self.p1.get_sym(), roll)
        self.board = move_and_val[0]
        return move_and_val

    def next_round(self):
        roll = self.roll()
        #print(roll)
        self.board = Board.flip(self.p2.make_move(Board.flip(self.board), self.p2.get_sym(), roll))
        return self.board

    def board_state(self):
        return self.board

    def train_model(self):
        episodes = 8000
        outcomes = []
        for episode in range(episodes):
            self.board = Board.initial_state
#            prev_board = self.board
            prev_board, prev_board_value = self.roll_and_find_best_for_bot()
            # find the best move here, make this move, then change turn as the
            # first thing inside of the while loop and then call
            # roll_and_find_best_for_bot to get V_t+1
#            self.p1.make_move(prev_board, self.p1.get_sym(), self.roll())
            while Board.outcome(self.board) is None:
                self.next_round()
                cur_board, cur_board_value = self.roll_and_find_best_for_bot()
                self.p1.get_network().train(prev_board, cur_board_value)
                prev_board = cur_board
#                self.next_round()
                # print("-"*30)
                # print(Board.pretty(self.board))
                # print("/"*30)
            print("Outcome:", Board.outcome(self.board)[1])
            outcomes.append(Board.outcome(self.board)[1])
            final_score = np.array([ Board.outcome(self.board)[1] ]).reshape((1, 1))
            self.p1.get_network().train(prev_board, final_score)
            print("trained episode {}".format(episode))
            if episode % 100 == 0:
                print("Saving...")
                self.p1.get_network().save_model()
                self.p2.restore_model()

        print(outcomes)
        print(sum(outcomes))

    def next_round_test(self):
        print(self.board)
        print()
        self.next_round()
        print("--------------------------------")
        print(self.board)
        print("--------------------------------")

    def play(self):
        count = 0
        while Board.outcome(self.board) is None:
            count += 1
            print("Turn:",count)

            roll = self.roll()

            print("type of board: ", type(self.board))
            print("Board:",self.board)
            print("{} rolled: {}".format(self.p1.get_sym(), roll))

            self.board = (self.p1.make_move(self.board, self.p1.get_sym(), roll))[0]

            print(self.board)

            print()
            
            count += 1
            
            roll = self.roll()
            print("{} rolled: {}".format(self.p2.get_sym(), roll))
            self.board = self.p2.make_move(self.board, self.p2.get_sym(), roll)


        if Board.outcome(self.board)[1] > 0:
            print_winner = "1: White, " + str(Board.outcome(self.board))
        else:
            print_winner = "-1: Black " + str(Board.outcome(self.board))
        print("The winner is {}!".format(print_winner))
        print("Final board:",Board.pretty(self.board))
        return count

highest = 0

#for i in range(100000):
#    try:
g = Game()
g.train_model()
#count = g.play()
 #       highest = max(highest,count)
 #   except KeyboardInterrupt:
 #       break
#print("\nHighest amount of turns is:",highest)