import time
from tictactoe import TicTacToe
from player import Player
from ai_player import AIPlayer
p1 = Player()
p2 = AIPlayer()
# Train with 10000 games against itself.
print('Learning...')
t = time.time()
for _ in range(10000):
game = TicTacToe(p2, p2)
game.start()
print('Learned {} unique states from {:.0f} games in {:.2f} seconds'.format(
len(p2.states['x']) + len(p2.states['o']),
p2.games_played() / 2,
time.time() - t))
# Only perform best moves.
p2.epsilon = 0
# Play against human player.
while True:
game = TicTacToe(p1, p2)
game.start()
p1 = AIPlayer()
p2 = SoftmaxPlayer()
# Train e-greedy player against itself.
for _ in range(2000):
game = TicTacToe(p1, p1)
game.start()
# Train softmax player against itself.
for _ in range(2000):
game = TicTacToe(p2, p2)
game.start()
# Set parameters so that optimal moves are preferred.
p1.epsilon = 0
p2.theta = 1
# Remove stats from training matches.
p1.draws = 0
p1.wins = 0
p1.losses = 0
p2.draws = 0
p2.wins = 0
p2.losses = 0
# E-greedy vs softmax.
for _ in range(100):
game = TicTacToe(p1, p2)
game.start()