def __init__(self, varname, player_1='human', player_2='random',
width=300, height=350, cid=None):
valid_players = ('human', 'random', 'alphabeta')
if player_1 not in valid_players or player_2 not in valid_players:
raise TypeError("Players must be one of {}".format(valid_players))
Canvas.__init__(self, varname, width, height, cid)
self.ttt = TicTacToe()
self.state = self.ttt.initial
self.turn = 0
self.strokeWidth(5)
self.players = (player_1, player_2)
self.font("20px Arial")
self.draw_board()
async def tic_tac_toe(self, ctx, player2: discord.Member):
def _confirm_reply(m):
return m.channel == ctx.message.channel and m.author == player2 and m.content.lower(
) in ['y', 'n']
await ctx.send(
f"{player2.mention} play a game of tic tac toe with {ctx.message.author}?"
)
reply_msg = await self.client.wait_for('message',
timeout=40.0,
check=_confirm_reply)
if reply_msg.content == 'n':
await ctx.send('Ok!, game cancelled')
return
def _move_reply(m):
return m.channel == ctx.message.channel and m.author == players[
player] and m.content in [str(x) for x in range(1, 10)]
player = 'O'
player1 = ctx.message.author
players = {
'X': player1,
'O': player2,
}
game = TicTacToe()
await ctx.send(game.rules)
await ctx.send(f"{player1.mention} ~ X\n{player2.mention} ~ O")
while True:
player = 'O' if player == 'X' else 'X'
while True:
try:
await ctx.send(
f"{players[player].mention}: enter a position")
reply_msg = await self.client.wait_for('message',
timeout=40.0,
check=_move_reply)
position = int(reply_msg.content)
game.playerMove(player, position)
break
except ValueError:
await ctx.send(f"{players[player].mention} Invalid move")
await ctx.send(game.display())
if game.checkWin(player):
await ctx.send(f"{players[player].mention} won!")
break
if game.checkDraw():
await ctx.send(
f"{players['X'].mention} {players['O'].mention} Draw!")
break
def __init__(self, varname, player_1='human', player_2='random',
width=300, height=350, cid=None):
valid_players = ('human', 'random', 'alphabeta')
if player_1 not in valid_players or player_2 not in valid_players:
raise TypeError("Players must be one of {}".format(valid_players))
Canvas.__init__(self, varname, width, height, cid)
self.ttt = TicTacToe()
self.state = self.ttt.initial
self.turn = 0
self.strokeWidth(5)
self.players = (player_1, player_2)
self.font("20px Arial")
self.draw_board()
async def msg_challenge(message):
p1 = message.author
if len(message.mentions) != 1:
await client.send_message(message.channel, 'Please mention one player.')
return
p2 = message.mentions[0]
for g in games:
if p1 in g.players or p2 in g.players:
await client.send_message(message.channel, 'One player is already in a game.')
return
if p2 == p1:
await client.send_message(message.channel, 'Can\'t play against yourself.')
return
cnt = message.content.lower()
game = TicTacToe([p1, p2])
await game.start(client, message)
print('Created game ' + str(game))
games.append(game)
import numpy as np
from games import TicTacToe
from players import RandomPlayer, MCTSPlayer
from montecarlo import MCTS
game = TicTacToe(3)
print(game.board())
player_random = RandomPlayer(game)
mcts_play = MCTS(game=game, player=player_random, episodes=800)
player = MCTSPlayer(game=game, mcts=mcts_play)
mcts = MCTS(game=game, player=player_random, episodes=800)
for i in range(1000):
while game.winner() == 0:
mcts.train()
p = mcts.get_policy(prop=1)
a = np.random.choice(mcts.action_space, p=p)
print(p, a, mcts.get_Qsa(game.state(), mcts.action_space),
sum(mcts.get_Nsa(game.state(), mcts.action_space)))
game.move(a)
print(game.board())
mcts.update()
game.set_state(game.start())
print(i)
print(game.board())
from tkinter import *
import sys
import os.path
sys.path.append(os.path.join(os.path.dirname(__file__), '..'))
from games import minimax_decision, alphabeta_player, random_player, TicTacToe
# "gen_state" can be used to generate a game state to apply the algorithm
from tests.test_games import gen_state
tttgraph = __import__("tic-tac-toe_graph")
ttt = TicTacToe()
root = None
buttons = []
frames = []
x_pos = []
o_pos = []
count = 0
sym = ""
result = None
choices = None
graph_display = None
def create_frames(root):
"""
This function creates the necessary structure of the game.
"""
frame1 = Frame(root)
frame2 = Frame(root)
frame3 = Frame(root)
frame4 = Frame(root)
create_buttons(frame1)
class Canvas_TicTacToe(Canvas):
"""Play a 3x3 TicTacToe game on HTML canvas"""
def __init__(self, varname, player_1='human', player_2='random',
width=300, height=350, cid=None):
valid_players = ('human', 'random', 'alphabeta')
if player_1 not in valid_players or player_2 not in valid_players:
raise TypeError("Players must be one of {}".format(valid_players))
Canvas.__init__(self, varname, width, height, cid)
self.ttt = TicTacToe()
self.state = self.ttt.initial
self.turn = 0
self.strokeWidth(5)
self.players = (player_1, player_2)
self.font("20px Arial")
self.draw_board()
def mouse_click(self, x, y):
player = self.players[self.turn]
if self.ttt.terminal_test(self.state):
if 0.55 <= x/self.width <= 0.95 and 6/7 <= y/self.height <= 6/7+1/8:
self.state = self.ttt.initial
self.turn = 0
self.draw_board()
return
if player == 'human':
x, y = int(3*x/self.width) + 1, int(3*y/(self.height*6/7)) + 1
if (x, y) not in self.ttt.actions(self.state):
# Invalid move
return
move = (x, y)
elif player == 'alphabeta':
move = alphabeta_player(self.ttt, self.state)
else:
move = random_player(self.ttt, self.state)
self.state = self.ttt.result(self.state, move)
self.turn ^= 1
self.draw_board()
def draw_board(self):
self.clear()
self.stroke(0, 0, 0)
offset = 1/20
self.line_n(0 + offset, (1/3)*6/7, 1 - offset, (1/3)*6/7)
self.line_n(0 + offset, (2/3)*6/7, 1 - offset, (2/3)*6/7)
self.line_n(1/3, (0 + offset)*6/7, 1/3, (1 - offset)*6/7)
self.line_n(2/3, (0 + offset)*6/7, 2/3, (1 - offset)*6/7)
board = self.state.board
for mark in board:
if board[mark] == 'X':
self.draw_x(mark)
elif board[mark] == 'O':
self.draw_o(mark)
if self.ttt.terminal_test(self.state):
# End game message
utility = self.ttt.utility(self.state, self.ttt.to_move(self.ttt.initial))
if utility == 0:
self.text_n('Game Draw!', offset, 6/7 + offset)
else:
self.text_n('Player {} wins!'.format("XO"[utility < 0]), offset, 6/7 + offset)
# Find the 3 and draw a line
self.stroke([255, 0][self.turn], [0, 255][self.turn], 0)
for i in range(3):
if all([(i + 1, j + 1) in self.state.board for j in range(3)]) and \
len({self.state.board[(i + 1, j + 1)] for j in range(3)}) == 1:
self.line_n(i/3 + 1/6, offset*6/7, i/3 + 1/6, (1 - offset)*6/7)
if all([(j + 1, i + 1) in self.state.board for j in range(3)]) and \
len({self.state.board[(j + 1, i + 1)] for j in range(3)}) == 1:
self.line_n(offset, (i/3 + 1/6)*6/7, 1 - offset, (i/3 + 1/6)*6/7)
if all([(i + 1, i + 1) in self.state.board for i in range(3)]) and \
len({self.state.board[(i + 1, i + 1)] for i in range(3)}) == 1:
self.line_n(offset, offset*6/7, 1 - offset, (1 - offset)*6/7)
if all([(i + 1, 3 - i) in self.state.board for i in range(3)]) and \
len({self.state.board[(i + 1, 3 - i)] for i in range(3)}) == 1:
self.line_n(offset, (1 - offset)*6/7, 1 - offset, offset*6/7)
# restart button
self.fill(0, 0, 255)
self.rect_n(0.5 + offset, 6/7, 0.4, 1/8)
self.fill(0, 0, 0)
self.text_n('Restart', 0.5 + 2*offset, 13/14)
else: # Print which player's turn it is
self.text_n("Player {}'s move({})".format("XO"[self.turn], self.players[self.turn]),
offset, 6/7 + offset)
self.update()
def draw_x(self, position):
self.stroke(0, 255, 0)
x, y = [i-1 for i in position]
offset = 1/15
self.line_n(x/3 + offset, (y/3 + offset)*6/7, x/3 + 1/3 - offset, (y/3 + 1/3 - offset)*6/7)
self.line_n(x/3 + 1/3 - offset, (y/3 + offset)*6/7, x/3 + offset, (y/3 + 1/3 - offset)*6/7)
def draw_o(self, position):
self.stroke(255, 0, 0)
x, y = [i-1 for i in position]
self.arc_n(x/3 + 1/6, (y/3 + 1/6)*6/7, 1/9, 0, 360)
from games import TicTacToe
from algorithms import MCTS
import numpy as np
import tqdm
## 10000 searches per move is pretty strong. Vary the quality of
## decision making by changing number of searches from 10000 to
## something else. Or, try changing the board size and playing a game.
# Create environment
env = TicTacToe(3)
# Reset
obs = env.reset()
# whose turn? True if you want to play first
human_play = True
player = 'x'
other_player = 'o'
while True:
# Print board and check if game over
env.print_board()
if env.won('x'):
print('Game over, x wins')
break
if env.won('o'):
print('Game over, o wins')
break
if env.draw():
print('Draw')
break
if human_play:
# ask for action
def game():
return TicTacToe()
class Canvas_TicTacToe(Canvas):
"""Play a 3x3 TicTacToe game on HTML canvas"""
def __init__(self, varname, player_1='human', player_2='random',
width=300, height=350, cid=None):
valid_players = ('human', 'random', 'alphabeta')
if player_1 not in valid_players or player_2 not in valid_players:
raise TypeError("Players must be one of {}".format(valid_players))
Canvas.__init__(self, varname, width, height, cid)
self.ttt = TicTacToe()
self.state = self.ttt.initial
self.turn = 0
self.strokeWidth(5)
self.players = (player_1, player_2)
self.font("20px Arial")
self.draw_board()
def mouse_click(self, x, y):
player = self.players[self.turn]
if self.ttt.terminal_test(self.state):
if 0.55 <= x/self.width <= 0.95 and 6/7 <= y/self.height <= 6/7+1/8:
self.state = self.ttt.initial
self.turn = 0
self.draw_board()
return
if player == 'human':
x, y = int(3*x/self.width) + 1, int(3*y/(self.height*6/7)) + 1
if (x, y) not in self.ttt.actions(self.state):
# Invalid move
return
move = (x, y)
elif player == 'alphabeta':
move = alphabeta_player(self.ttt, self.state)
else:
move = random_player(self.ttt, self.state)
self.state = self.ttt.result(self.state, move)
self.turn ^= 1
self.draw_board()
def draw_board(self):
self.clear()
self.stroke(0, 0, 0)
offset = 1/20
self.line_n(0 + offset, (1/3)*6/7, 1 - offset, (1/3)*6/7)
self.line_n(0 + offset, (2/3)*6/7, 1 - offset, (2/3)*6/7)
self.line_n(1/3, (0 + offset)*6/7, 1/3, (1 - offset)*6/7)
self.line_n(2/3, (0 + offset)*6/7, 2/3, (1 - offset)*6/7)
board = self.state.board
for mark in board:
if board[mark] == 'X':
self.draw_x(mark)
elif board[mark] == 'O':
self.draw_o(mark)
if self.ttt.terminal_test(self.state):
# End game message
utility = self.ttt.utility(self.state, self.ttt.to_move(self.ttt.initial))
if utility == 0:
self.text_n('Game Draw!', offset, 6/7 + offset)
else:
self.text_n('Player {} wins!'.format("XO"[utility < 0]), offset, 6/7 + offset)
# Find the 3 and draw a line
self.stroke([255, 0][self.turn], [0, 255][self.turn], 0)
for i in range(3):
if all([(i + 1, j + 1) in self.state.board for j in range(3)]) and \
len({self.state.board[(i + 1, j + 1)] for j in range(3)}) == 1:
self.line_n(i/3 + 1/6, offset*6/7, i/3 + 1/6, (1 - offset)*6/7)
if all([(j + 1, i + 1) in self.state.board for j in range(3)]) and \
len({self.state.board[(j + 1, i + 1)] for j in range(3)}) == 1:
self.line_n(offset, (i/3 + 1/6)*6/7, 1 - offset, (i/3 + 1/6)*6/7)
if all([(i + 1, i + 1) in self.state.board for i in range(3)]) and \
len({self.state.board[(i + 1, i + 1)] for i in range(3)}) == 1:
self.line_n(offset, offset*6/7, 1 - offset, (1 - offset)*6/7)
if all([(i + 1, 3 - i) in self.state.board for i in range(3)]) and \
len({self.state.board[(i + 1, 3 - i)] for i in range(3)}) == 1:
self.line_n(offset, (1 - offset)*6/7, 1 - offset, offset*6/7)
# restart button
self.fill(0, 0, 255)
self.rect_n(0.5 + offset, 6/7, 0.4, 1/8)
self.fill(0, 0, 0)
self.text_n('Restart', 0.5 + 2*offset, 13/14)
else: # Print which player's turn it is
self.text_n("Player {}'s move({})".format("XO"[self.turn], self.players[self.turn]),
offset, 6/7 + offset)
self.update()
def draw_x(self, position):
self.stroke(0, 255, 0)
x, y = [i-1 for i in position]
offset = 1/15
self.line_n(x/3 + offset, (y/3 + offset)*6/7, x/3 + 1/3 - offset, (y/3 + 1/3 - offset)*6/7)
self.line_n(x/3 + 1/3 - offset, (y/3 + offset)*6/7, x/3 + offset, (y/3 + 1/3 - offset)*6/7)
def draw_o(self, position):
self.stroke(255, 0, 0)
x, y = [i-1 for i in position]
self.arc_n(x/3 + 1/6, (y/3 + 1/6)*6/7, 1/9, 0, 360)
from games import (GameState, Game, Fig52Game, TicTacToe, query_player,
random_player, alphabeta_player, minimax_decision,
alphabeta_full_search, alphabeta_search)
if __name__ == '__main__':
ttt = TicTacToe()
print('Initial game state:')
print(ttt.display(ttt.initial))
for _ in range(10):
print(ttt.play_game(alphabeta_player, random_player))
)
adaptive_trainer_robot = PolicyGradientPlayer(
learner_brain,
discount_rate=DISCOUNT_RATE,
batch_iterations=BATCH_ITERATIONS,
experience_batch_size=EXPERIENCE_BATCH_SIZE,
experience_buffer_size=EXPERIENCE_BUFFER_SIZE,
)
static_trainer_robot = PolicyGradientPlayer(
static_trainer_brain,
discount_rate=DISCOUNT_RATE,
batch_iterations=BATCH_ITERATIONS,
experience_buffer_size=EXPERIENCE_BUFFER_SIZE,
)
adaptive_train_game = TicTacToe((learner_robot, adaptive_trainer_robot))
static_train_game = TicTacToe((learner_robot, static_trainer_robot))
# Initialize plot data
game_counts = []
wins = []
losses = []
scores = []
mean_experience_values = []
mean_confidences = []
brain_costs = []
brain_costs_ema = []
weight_ranges = []
weight_means = []
plot_data = {
"score": {
discount_factor=DISCOUNT_FACTOR,
reward_factor=REWARD_FACTOR,
batch_iterations=BATCH_ITERATIONS,
experience_batch_size=EXPERIENCE_BATCH_SIZE,
experience_buffer_size=EXPERIENCE_BUFFER_SIZE,
)
train_player = PolicyGradientPlayer(
robot_brain,
discount_factor=DISCOUNT_FACTOR,
reward_factor=REWARD_FACTOR,
batch_iterations=1,
experience_batch_size=EXPERIENCE_BATCH_SIZE,
experience_buffer_size=EXPERIENCE_BUFFER_SIZE,
)
human_game = TicTacToe((human, robot))
training = TicTacToe((robot, train_player))
random_training = TicTacToe((robot, RandomPlayer()))
robot.act_greedy = True
robot.show_action_probabilities = True
playing = True
while playing:
# Gain experience, no learning to keep it fast
robot.learn_while_playing = False
random_training.play(32)
training.play(32)
# Learn on every move of the human game
action='store_true',
default=False)
parser.add_argument(
"--no-clear",
help="If True, the screen is not cleared between rounds",
action='store_true',
default=False)
parser.add_argument(
"--cores",
help="The number of threads that will be run concurrently",
default=4)
args = parser.parse_args()
# Get game to play
if args.game == 'tictactoe':
game_rules = TicTacToe()
elif args.game == 'connect4':
game_rules = Connect4()
elif args.game == 'draughts':
game_rules = Draughts()
# Decide whether to show probabilities from argument
print('Welcome to the Monte-Carlo tree search AI')
print()
print('AI plays first')
print()
while True:
interface = Interface(game_rules, parallel=not args.no_parallel)
interface.set_iterations(int(args.iter))
interface.print_path(args.print_path)
interface.show_probabilities = args.show_probs
from games import TicTacToe # task 1.2.1 ttt = TicTacToe() #ttt.run(1000, 'move at random') #ttt.draw(True) ttt.run(1000, 'move at probability x_board') ttt.draw(False) ttt.run(1000, 'move at probability x_o_board') ttt.draw(False) ttt.run(1000, 'move at heuristic') ttt.draw(False) ttt.run(1000, 'move at heuristic forward') ttt.draw(False)
import pickle
from games import TicTacToe
from brain import Brain
from players import PolicyGradientPlayer, HumanPlayer, RandomPlayer
from brain.activation_functions import ReLU, Softmax
human = HumanPlayer()
BRAIN_FILEPATH = "brain/saved/tictactoe-brain.pickle"
playing = True
while playing:
robot_brain = pickle.load(open(BRAIN_FILEPATH, "rb"))
robot = PolicyGradientPlayer(robot_brain)
robot.act_greedy = True
robot.show_action_probabilities = 0.3
human_game = TicTacToe((human, robot))
# robot_game = TicTacToe((robot, robot_opponent))
playing = human_game.play(2, render=True, pause=0.5)
# robot_game.play(1, render=True, pause=0.7)
from games import TicTacToe, minimax_decision, alphabeta_search, GameState, alphabeta_player, random_player
print("Enter the first row.")
row1 = input().split(' ')
print("Enter the second row.")
row2 = input().split(' ')
print("Enter the third row.")
row3 = input().split(' ')
ticTacToe = TicTacToe()
board = {}
moves = []
row1.extend(row2)
row1.extend(row3)
table = row1
rowIndex = 1
columnIndex = 1
# Creating the current state of the game, from the user inputs
for element in table:
if (element in ['X', 'O']):
board[(rowIndex, columnIndex)] = element
else:
moves.append((rowIndex, columnIndex))
columnIndex += 1
if (columnIndex > 3):
rowIndex += 1
columnIndex = 1
# Code to find who has to move next, considering X starts the game
nextToMove = 'X'
(64, ReLU),
(9, Softmax),
)
robot_brain = Brain(BRAIN_TOPOLOGY,
learning_rate=LEARNING_RATE,
regularization=REGULARIZATION)
learning_robot = PolicyGradientPlayer(
robot_brain,
discount_rate=DISCOUNT_RATE,
batch_iterations=BATCH_ITERATIONS,
experience_batch_size=EXPERIENCE_BATCH_SIZE,
experience_buffer_size=EXPERIENCE_BUFFER_SIZE,
)
random_game = TicTacToe((learning_robot, RandomPlayer()))
# Initialize plot data
game_counts = []
wins = []
losses = []
scores = []
mean_experience_values = []
mean_confidences = []
brain_costs = []
brain_costs_ema = []
weight_ranges = []
weight_means = []
# Create a plot figure
plot_data = {