def battle(player1: Player = RandomPlayer(),
player2: Player = RandomPlayer(),
num_games: int = 100000,
silent: bool = False):
board = Board()
draw_count = 0
cross_count = 0
naught_count = 0
for _ in range(num_games):
result = play_game(board, player1, player2)
if result == GameResult.CROSS_WIN:
cross_count += 1
elif result == GameResult.NAUGHT_WIN:
naught_count += 1
else:
draw_count += 1
if not silent:
print(
"After {} game we have draws: {}, Player 1 wins: {}, and Player 2 wins: {}."
.format(num_games, draw_count, cross_count, naught_count))
print(
"Which gives percentages of draws: {:.2%}, Player 1 wins: {:.2%}, and Player 2 wins: {:.2%}"
.format(draw_count / num_games, cross_count / num_games,
naught_count / num_games))
return cross_count, naught_count, draw_count
from tic_tac_toe.RndMinMaxAgent import RndMinMaxAgent
from tic_tac_toe.DirectPolicyAgent import DirectPolicyAgent
TENSORLOG_DIR = './graphs'
if tf.gfile.Exists(TENSORLOG_DIR):
tf.gfile.DeleteRecursively(TENSORLOG_DIR)
tf.reset_default_graph()
nnplayer = DirectPolicyAgent("PolicyLearner1")
# nn2player = EGreedyNNQPlayer("QLearner2", win_value=100.0, loss_value=-100.0)
# nnplayer = EGreedyNNQPlayer("QLearner1")#, learning_rate=0.001, win_value=10.0, loss_value=-10.0)
# nn2player = EGreedyNNQPlayer("QLearner2")#, learning_rate=0.001, win_value=10.0, loss_value=-10.0)
mm_player = MinMaxAgent()
rndplayer = RandomPlayer()
rm_player = RndMinMaxAgent()
TFSessionManager.set_session(tf.Session())
sess = TFSessionManager.get_session()
writer = tf.summary.FileWriter(TENSORLOG_DIR, sess.graph)
nnplayer.writer = writer
sess.run(tf.global_variables_initializer())
# game_number, p1_wins, p2_wins, draws = evaluate_players(rndplayer, nnplayer, num_battles=10000) #, num_battles = 20)
# game_number, p1_wins, p2_wins, draws = evaluate_players(rndplayer, nnplayer) #, num_battles = 20)
# game_number, p1_wins, p2_wins, draws = evaluate_players( mm_player, nnplayer, num_battles=300) # , num_battles = 20)
game_number, p1_wins, p2_wins, draws = evaluate_players(
nnplayer, rm_player, num_battles=1000,
silent=True)
p1_wins.append(p1win * 100.0 / games_per_battle)
p2_wins.append(p2win * 100.0 / games_per_battle)
draws.append(draw * 100.0 / games_per_battle)
counter += 1
game_number.append(counter)
plt.ylabel('Game outcomes in %')
plt.xlabel('Game number')
plt.plot(game_number, draws, 'r-', label='Draw')
plt.plot(game_number, p1_wins, 'g-', label='Player 1 wins')
plt.plot(game_number, p2_wins, 'b-', label='Player 2 wins')
plt.legend(loc=loc, shadow=True, fancybox=True, framealpha=0.7)
plt.show()
# TFSessionManager.set_session(None)
return game_number, p1_wins, p2_wins, draws
# tf.reset_default_graph()
nnplayer = NNQPlayer("QLearner1")
randPlayer = RandomPlayer()
nnplayer2 = NNQPlayer("QLearner2")
greedyplayer = EGreedyNNQPlayer("Greedy")
game_number, p1_wins, p2_wins, draws = eval_players(greedyplayer, randPlayer)
p = plt.plot(game_number, draws, 'r-', game_number, p1_wins, 'g-', game_number,
p2_wins, 'b-')
from tic_tac_toe.Board import Board, GameResult, CROSS, NAUGHT, EMPTY from util import print_board, play_game, battle from tic_tac_toe.RandomPlayer import RandomPlayer from tic_tac_toe.MinMaxAgent import MinMaxAgent from tic_tac_toe.RndMinMaxAgent import RndMinMaxAgent from tic_tac_toe.TabularQPlayer import TQPlayer from tic_tac_toe.SimpleNNQPlayer import NNQPlayer from tic_tac_toe.TFSessionManager import TFSessionManager import matplotlib.pyplot as plt import tensorflow as tf import random board = Board() #tf.reset_default_graph() player1 = RandomPlayer() player2 = RandomPlayer() p1_wins = [] p1count = 0 p2_wins = [] p2count = 0 draws = [] drawcount = 0 count = [] num_battles = 100 games_per_battle = 10 TFSessionManager.set_session(tf.Session()) TFSessionManager.get_session().run(tf.global_variables_initializer())