def test_within_bounds_returns_true_if_move_is_within_index_range():
new_board = Board()
new_board.board = ["X", " ", " ", " ", " ", " ", " ", " ", " "]
assert not new_board.within_bounds(10)
assert new_board.within_bounds(1)
def test_human_player_can_make_a_move():
new_board = Board()
player = Player(token=["X", "O"])
new_board.move(1, new_board.board, player.token[0])
assert new_board.board[1] == "X"
def test_win_can_detect_a_winning_combination_present_on_the_game_board(
board_state, win_status):
new_board = Board()
new_board.board = board_state
assert new_board.win() == win_status
def test_winner_saves_the_winner(board_state, winner_token):
new_board = Board()
new_board.board = board_state
assert new_board.winner()
assert new_board.winner() == winner_token
def test_game_over_returns_true_when_a_game_is_won(board_state,
game_over_status):
new_board = Board()
new_board.board = board_state
assert new_board.game_over() == game_over_status
def test_turn_count_keeps_track_of_number_of_turns(board_state, turn_count):
new_board = Board()
new_board.board = board_state
player = Player(token=["X", "O"])
assert new_board.turn_count(player) == turn_count
def test_position_taken_returns_true_if_square_is_already_occupied_and_false_if_not(
):
new_board = Board()
new_board.board = ["X", " ", " ", " ", " ", " ", " ", " ", " "]
assert new_board.position_taken(0)
assert not new_board.position_taken(1)
def test_board_raises_error_when_positions_are_taken():
new_board = Board()
player = Player(token=["X", "O"])
new_board.board = ["X"]
with pytest.raises(PositionAlreadyTakenError):
new_board.turn(0, player)
def test_current_player_returns_current_player_token(board_state,
current_player):
new_board = Board()
new_board.board = board_state
player = Player(token=["X", "O"])
assert new_board.current_player(player) == current_player
def test_user_interface_congratulates_winner():
spy = PrinterSpy()
user_message = UserMessages(spy)
board = Board()
board.board = ["X", "O", " ", "X", "O", " ", "X", " ", " "]
board.win()
user_message.who_won(board)
assert spy.printed_item == "X, you're the winner!"
def init_board_states():
state = ['0'] * board_size
for i in range(3):
state[0] = str(i)
for j in range(3):
state[1] = str(j)
for k in range(3):
state[2] = str(k)
for l in range(3):
state[3] = str(l)
for m in range(3):
state[4] = str(m)
for n in range(3):
state[5] = str(n)
for o in range(3):
state[6] = str(o)
for p in range(3):
state[7] = str(p)
for q in range(3):
state[8] = str(q)
if valid_state(state):
state_str = "".join(state)
states[state_str] = Board(copy.deepcopy(state))
set_board_actions()
def find_possible_states(self, output_path, export=True):
'''
Get all possible states
:param output_path:
:param export:
:return:
'''
b = Board()
print('get all possible states: player x plays at first')
self.get_all_possible_states(b, Board.x_id)
print('get all possible states: player o plays at first')
self.get_all_possible_states(b, Board.o_id)
print(f'Num of possible states = {len(t.states)}')
# export to file
if export:
print(f'export possible states to file')
with open(output_path, mode='w') as employee_file:
writer = csv.writer(employee_file, delimiter=',', quotechar='"', quoting=csv.QUOTE_MINIMAL)
writer.writerow(['State_id', 'is_over', 'winner'])
WINNER_COLUMN = 2
OVER_COLUMN = 1
STATE_ID_COLUMN = 0
for pair in self.pairs:
if pair[WINNER_COLUMN] == None:
# in case that the game is not ended, there is no winner
writer.writerow([pair[STATE_ID_COLUMN], pair[OVER_COLUMN], 'None'])
else:
writer.writerow(pair)
def main():
"""main function to ran the game"""
first, second = introduction()
if isinstance(first, Player):
player = 'x_symbol'
else:
player = 'o_symbol'
board = Board()
while True:
if board.play:
board = first.make_move(board, Board.CROSS)
board.check_status()
if board.play:
board = second.make_move(board, Board.ZERO)
board.check_status()
else:
break
if board.draw:
print('The game ended in a draw (ಥ_ಥ)')
# representing end of the game
if board.have_winner:
winner = board.winner()
if (winner == '✗' and player == 'x_symbol'):
print('Cool, {}, you won the battle, but what about the the '
'war'.format(first.player_name))
elif (winner == 'o' and player == 'o_symbol'):
print('Cool, {}, you won the battle, but what about the the '
'war'.format(second.player_name))
else:
print('You lost the battle, but what about the war')
def test_tic_tac_toe_display_board_with_user_token():
new_board = Board()
view = CommandLineBoardPresenter()
spy = PrinterSpy()
new_board.board = ["X", " ", " ", " ", "X", " ", " ", " ", "X"]
view.display_board(new_board, spy)
expected = """\
X | |
--------------
| X |
--------------
| | X
"""
assert spy.printed() == expected
def test_tic_tac_toe_display_board():
new_board = Board()
view = CommandLineBoardPresenter()
spy = PrinterSpy()
view.display_board(new_board, spy)
expected = """\
| |
--------------
| |
--------------
| |
"""
assert spy.printed() == expected
def main():
print("Choosing player X...")
player_x = _pick_agent(Player.X)
print("Choosing player O...")
player_o = _pick_agent(Player.O)
size = int(input("Select size of the board >= 3: "))
print("Select number in a row to win, <=",size,": ", end = '')
to_win = int(input(""))
random_board = input("Random board state? y/[n]:")
play = "y"
games = 0
while play == "y":
if(random_board == "y"):
game_board = Board(size, to_win)
game_board.randomize()
winner = game_board.winner
while(winner is not None):
game_board = Board(size, to_win)
game_board.randomize()
winner = game_board.winner
else:
game_board = None
game = Game(player_x, player_o, size, to_win, game_board)
res = game.play()
games += 1
RESULTS[res + 1][1] += 1
for line in RESULTS:
print(*line)
print("")
print("X average runtime: ",player_x.average_runtime)
print("O average runtime: ",player_o.average_runtime)
print("States visited X", player_x.states_visited)
print("States visited O", player_o.states_visited)
play = input("Play again? y/[n]: ")
def run():
board = Board()
player = Player(token=["X", "O"])
view = CommandLineBoardPresenter()
printer = Printer()
user_messages = UserMessages(printer)
errors = Errors(printer)
user_messages.countdown()
user_messages.logo()
user_messages.instructions_option()
if player.get_input() == 0:
user_messages.display_instructions()
view.display_board(board, printer)
else:
view.display_board(board, printer)
user_messages.whos_turn(player, board)
while not board.game_over():
try:
selection = player.get_input()
board.turn(selection, player)
except InputNotNumericError:
errors.input_not_numeric_error_message()
except InvalidBoardIndexError:
errors.invalid_board_index_error_message()
except PositionAlreadyTakenError:
errors.position_already_taken_error_message()
finally:
view.display_board(board, printer)
if not board.win() and not board.tie():
user_messages.whos_turn(player, board)
if board.win():
user_messages.who_won(board)
if board.tie():
user_messages.its_a_tie()
def read_board_states():
with open('tic_tac_toe_x_actions.csv') as csv_file:
csv_reader = csv.reader(csv_file, delimiter=',')
for row in csv_reader:
board = row[0]
states[board] = Board(list(board))
with open('tic_tac_toe_x_actions.csv') as csv_file:
csv_reader = csv.reader(csv_file, delimiter=',')
for row in csv_reader:
board = row[0]
row_length = len(row)
for i in range(1, row_length):
action = states[row[i]]
states[board].add_action(action, player_dict['X'])
with open('tic_tac_toe_o_actions.csv') as csv_file:
csv_reader = csv.reader(csv_file, delimiter=',')
for row in csv_reader:
board = row[0]
row_length = len(row)
for i in range(1, row_length):
action = states[row[i]]
states[board].add_action(action, player_dict['O'])
def play_with_machine(V_machine, human=Board.x_symbol, machine=Board.o_symbol):
'''
:param V: value functions of machine
:param human: 'x' or 'o'
:param machine: 'o' or 'x'
:return:
'''
current_turn = human
b = Board()
board = b.board
states = []
states_reward = []
while not b.is_game_over():
if current_turn == human:
print(f'>>> Human turn ({human})')
else:
print(f'>>> Machine turn ({machine})')
if current_turn == human:
# ask human for move
available_move = False
while not available_move:
print('Let choose the cell you want to play (e.g., 1 2): ')
inp = input().split(' ')
row = int(inp[0])
col = int(inp[1])
if row > b.height or col > b.width or row < 0 or col < 0:
available_move = False
print('Wrong move!')
elif board[row, col] == Board.empty_id:
board[row, col] = b.convert_turn_symbol2id(current_turn)
available_move = True
else:
available_move = False
print('Wrong move!')
else:
# find potential moves
potential_states = []
potential_moves = []
for i in range(b.height):
for j in range(b.width):
if board[i, j] == b.empty_id:
board[i, j] = b.convert_turn_symbol2id(current_turn)
potential_states.append(b.get_state())
potential_moves.append([i, j])
board[i, j] = b.empty_id
# find the best move
if len(potential_moves) > 0:
best_move_value = 0
best_move = []
#print(f'potential_moves = {potential_moves}')
for idx, state in enumerate(potential_states):
print(f'potential move {potential_moves[idx]} = {np.round(V_machine[potential_states[idx]],2)}')
for move, state in zip(potential_moves, potential_states):
if V_machine[state] > best_move_value:
best_move_value = V_machine[state]
best_move = move
# play
print(f'best move: {best_move}')
board[int(best_move[0]), int(best_move[1])] = b.convert_turn_symbol2id(current_turn)
# store the state of the board
states.append(b.get_state())
states_reward.append(Training.AVERAGE_REWARD)
# change turn
if current_turn == human:
current_turn = machine
else:
current_turn = human
b.draw_board()
print(f'Winner = {b.winner_symbol}')
class SuboptimalAgent(Agent):
board = Board()
def get_appended_board_state(self, board_state, cell_value, i, j):
board_state[i][j] = cell_value
return board_state
def minimax(self, board_state, maximizer, depth):
self.board._board = board_state
# self.board.print_board()
current_utility = None
chosen_utility = None
if self._player == Player.X:
if self.board.winner == Player.X:
return 1 / depth
if self.board.winner == Player.O:
return -1 / depth
else:
if self.board.winner == Player.O:
return 1 / depth
if self.board.winner == Player.X:
return -1 / depth
if depth == self.board.size**2:
return 0
# self.board.print_board()
for i in range(self.board.size):
for j in range(self.board.size):
if self.board.cell(i, j) == CellState.EMPTY:
if self._player == Player.X:
if maximizer and chosen_utility != 1:
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 0, i, j), not maximizer,
depth + 1)
#here the state would be restored, but instead the board simply fills up
self.board._board = board_state
if not maximizer and chosen_utility != -1:
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 1, i, j), not maximizer,
depth + 1)
self.board._board = board_state
else:
if maximizer and chosen_utility != 1:
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 1, i, j), not maximizer,
depth + 1)
self.board._board = board_state
if not maximizer and chosen_utility != -1:
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 0, i, j), not maximizer,
depth + 1)
self.board._board = board_state
if current_utility is not None:
if maximizer:
if chosen_utility is None or current_utility > chosen_utility:
chosen_utility = current_utility
else:
if chosen_utility is None or current_utility < chosen_utility:
chosen_utility = current_utility
return chosen_utility
def next_move(self, given_board):
self.board._size = given_board._size
self.board._k = given_board._k
best_move_i = -1
best_move_j = -1
maximum = -100
depth = given_board.size**2
maximizer = True
for i in range(given_board.size):
for j in range(given_board.size):
if given_board.cell(i, j) == CellState.EMPTY:
depth -= 1
for i in range(given_board.size):
for j in range(given_board.size):
if given_board.cell(i, j) == CellState.EMPTY:
if self._player == Player.X:
current = self.minimax(
self.get_appended_board_state(
deepcopy(given_board._board), 0, i, j),
not maximizer, depth + 1)
else:
current = self.minimax(
self.get_appended_board_state(
deepcopy(given_board._board), 1, i, j),
not maximizer, depth + 1)
if current is not None:
if current > maximum:
maximum = current
best_move_i = i
best_move_j = j
return Move(self._player, best_move_i, best_move_j)
def test_tie_can_detect_a_full_board_with_no_winners():
new_board = Board()
new_board.board = ["X", "O", "X", "O", "X", "O", "O", "X", "O"]
assert new_board.tie()
class BruteForceAgent(Agent):
#initializes the board object and the state space
board = Board()
state_space = 0
#returns a board state with a changed cell value at the given coordinates
def get_appended_board_state(self, board_state, cell_value, i, j):
board_state[i][j] = cell_value
return board_state
#minimax algorithm for brute force, taking a board configuration integer array, a maximizer boolean, and a depth integer as parameters
def minimax(self, board_state, maximizer, depth):
#increments state space and fixes the new board state to the board object at each recursive call
self.state_space += 1
self.board._board = board_state
#initialize current and chosen utility
current_utility = None
chosen_utility = None
#this code block catches any winning boards and returns their evaluations; it minimizes calculation by only calculating board.winner once and setting it
#to a variable; calculation is reduced further by only calculating board.winner in cases where the depth is deep enough for a winner to be possible,
#i.e. the depth is greater than or equal to the depth of the winning sequence
winner = None
if depth > self.board._k - 1:
winner = self.board.winner
if self._player == Player.X:
if winner == Player.X:
#depth is accounted for in evaluating the winning board states; this allows the agent to find the quickest path to victory within the game,
#but it does not change the winrate of the agent (since it will only cause the agent to win in less moves, not more frequently), and it
#increases calculation time fairly significantly
return 1 / depth
if winner == Player.O:
return -1 / depth
else:
if winner == Player.O:
return 1 / depth
if winner == Player.X:
return -1 / depth
#accounts for the case of a drawn board, returning an evaluation of 0
if depth == self.board.size**2:
return 0
#this code block is the main body of the algorithm; it loops through the given board and calculates minimax on every empty square
for i in range(self.board.size):
for j in range(self.board.size):
if self.board.cell(i, j) == CellState.EMPTY:
#this if duplicates the functionality of the loop for players X and O
if self._player == Player.X:
#this if duplicates the functionality of the loop for the maximizer and the minimizer
if maximizer and chosen_utility != 1:
#sets the currently examined utility equal to the minimax evaluation of the current blank square; the current player's
#symbol (X or O) is placed in the coordinates of the blank, and the resulting board is passed as a new board state;
#the maximizer/minimizer is swapped and passed; the new depth is passed
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 0, i, j), not maximizer,
depth + 1)
#once the recursion is exited, the value that was changed and tested is returned to a blank
board_state[i][j] = -1
#the board's integer array is reinitialized to the restored board state
self.board._board = board_state
if not maximizer and chosen_utility != -1:
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 1, i, j), not maximizer,
depth + 1)
board_state[i][j] = -1
self.board._board = board_state
else:
#again, duplicates the functionality of the loop for the maximizer and minimizer
if maximizer and chosen_utility != 1:
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 1, i, j), not maximizer,
depth + 1)
board_state[i][j] = -1
self.board._board = board_state
if not maximizer and chosen_utility != -1:
current_utility = self.minimax(
self.get_appended_board_state(
board_state, 0, i, j), not maximizer,
depth + 1)
board_state[i][j] = -1
self.board._board = board_state
#this code block controls the algorithm's decisionmaking; in the case of the maximizer, each utility is compared to the current best and the
#greater of the two is stored; the minimizer does the opposite, storing the lowest utility
if current_utility is not None:
if maximizer:
if chosen_utility is None or current_utility > chosen_utility:
chosen_utility = current_utility
else:
if chosen_utility is None or current_utility < chosen_utility:
chosen_utility = current_utility
#when each non-leaf recursive call is complete, the chosen utility is returned instead of a raw value
return chosen_utility
def next_move(self, given_board):
#initializes the state space, the board size, and the k-value
self.state_space = 0
self.board._k = given_board._k
self.board._size = given_board._size
#initializes the local variables for the best i and j, the chosen utility, and the depth
best_move_i = -1
best_move_j = -1
chosen = -100
#the player making a move should maximize utility, so they begin as the maximizer
maximizer = True
#sets the depth equal to the number of occupied squares by decrementing from a full board of size^2 based on how
#many empty squares there are
depth = given_board.size**2
for i in range(given_board.size):
for j in range(given_board.size):
if given_board.cell(i, j) == CellState.EMPTY:
depth -= 1
#this code block duplicates the function of the main loop in the recursion (but with only the maximizer accounted for)
#this serves as the preliminary step to start off the recursion
for i in range(given_board.size):
for j in range(given_board.size):
if given_board.cell(i, j) == CellState.EMPTY:
if self._player == Player.X:
current = self.minimax(
self.get_appended_board_state(
deepcopy(given_board._board), 0, i, j),
not maximizer, depth + 1)
else:
current = self.minimax(
self.get_appended_board_state(
deepcopy(given_board._board), 1, i, j),
not maximizer, depth + 1)
if current is not None:
#keeps track not only of the best utility, but of the coordinates where it was found
if current > chosen:
chosen = current
best_move_i = i
best_move_j = j
#prints the state space after each move
print("State space traversed contained " + str(self.state_space) +
" states")
#returns the best move
return Move(self._player, best_move_i, best_move_j)
def recurse(p):
"""
make new board for branch
:param p: BSTNode
:return: None
"""
if p is not None and p.data.check_win() is None:
new_board = Board()
for i in range(3):
for j in range(3):
new_board.add_value(i, j,
p.data._game_board[i, j])
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
while not new_board.is_good_move(move1_x, move1_y):
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
new_board.second_player_move(move1_x, move1_y)
if new_board.check_win() is None:
while not new_board.is_good_move(move2_x,
move2_y) or (
move1_x == move2_x and move1_y == move2_y):
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
new_board.first_player_move(move2_x, move2_y)
self._builded_tree._add_right(p, new_board)
recurse(self._builded_tree.super_find(new_board))
if new_board.check_win() is None:
for i in range(3):
for j in range(3):
new_board.add_value(i, j,
p.data._game_board[i, j])
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
while not new_board.is_good_move(move1_x, move1_y):
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
new_board.second_player_move(move1_x, move1_y)
if new_board.check_win() is None:
while not new_board.is_good_move(move2_x,
move2_y) or (
move1_x == move2_x and move1_y == move2_y):
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
new_board.first_player_move(move2_x, move2_y)
self._builded_tree._add_left(p, new_board)
recurse(self._builded_tree.super_find(new_board))
return None
def play_to_train(self, Vo, Vx, n_iterations=10000, initial_player=Board.x_symbol):
'''
Train
:param Vo: the value functions of player 'o' (is a dictionary).
Example: Vo[1000010000] = 0.35,
where 1000010000 is the id of the state,
0.35 is the value function of the state 1000010000.
:param Vx: the value functions of player 'x' (is a dictionary)
:param n_iterations: the number of times we play
:return:
'''
assert (initial_player == Board.x_symbol or initial_player == Board.o_symbol)
assert (len(Vx) == len(Vo))
assert (n_iterations > 0)
current_turn = initial_player
for iteration in range(n_iterations):
if iteration % 1000 == 0:
print(f'iteration = {iteration}')
b = Board()
board = b.board
current_Vx = []
current_Vo = []
current_id_states = []
draw = False
game_over = False
while not draw and not game_over:
# print('++++++++++++++++++++++++++++++')
# retrieve all empty cells
empty_cells = []
for i in range(b.height):
for j in range(b.width):
if board[i, j] == 0:
empty_cells.append([i, j])
# print(f'Empty cells = {empty_cells}')
# just choose one empty cell to play
if len(empty_cells) > 0:
played_cell = np.random.choice(len(empty_cells))
board[empty_cells[played_cell][0], empty_cells[played_cell][1]] = b.convert_turn_symbol2id(
current_turn)
# print(f'Play {current_turn} on {empty_cells[played_cell]}')
# update states
state = b.get_state()
current_id_states.append(state)
current_Vx.append(self.AVERAGE_REWARD)
current_Vo.append(self.AVERAGE_REWARD)
game_over = b.is_game_over()
# print(f'Over: {over}')
# print(f'Winner: {b.winner}')
# print(f'State id: {b.get_state()}')
# update turn
if current_turn == Board.x_symbol:
current_turn = Board.o_symbol
else:
current_turn = Board.x_symbol
else:
draw = True
# b.draw()
# update the value function of states
if game_over:
winner = b.get_winner()
if winner == Board.x_symbol:
current_Vx[-1] = self.HIGHEST_REWARD
current_Vo[-1] = self.LOWEST_REWARD
elif winner == Board.o_symbol:
current_Vx[-1] = self.LOWEST_REWARD
current_Vo[-1] = self.HIGHEST_REWARD
Vx = self.update_value_function(Vx, current_id_states, current_Vx)
Vo = self.update_value_function(Vo, current_id_states, current_Vo)
else:
# print('Draw')
pass
# change the turn in the next play
# 50% the game starts with the player 1
# 50% the game starts with the player 2
if current_turn == Board.x_symbol:
current_turn = Board.o_symbol
else:
current_turn = Board.x_symbol
# Author: Kyle Olsen <*****@*****.**>
from __future__ import print_function, absolute_import
from tic_tac_toe.board import Board
if __name__ == '__main__':
game = Board()
current_player = 1
err_msg = None
while True:
game.display()
# Figure out who the current player is and gather input
mark = 'X'
if current_player == 2:
mark = 'O'
if err_msg is not None:
print(err_msg)
err_msg = None
try:
row = int(raw_input('Player {} ({}s), select a row: '.format(current_player, mark)))
cell = int(raw_input('Player {} ({}s), select a cell: '.format(current_player, mark)))
except ValueError:
err_msg = 'Row and Cell must be integers'
continue
# Try to mark the cell the player has chosen
try:
game.mark_cell(row, cell, mark)
except IndexError:
def test_board_raises_error_if_index_not_on_board():
new_board = Board()
player = Player(token=["X", "O"])
with pytest.raises(InvalidBoardIndexError):
new_board.turn(9, player)
def __init__(self):
self._board = Board()
class AlphaBetaAgent(Agent):
board = Board()
state_space = 0
def get_appended_board_state(self, board_state, cell_value, i, j):
board_state[i][j] = cell_value
return board_state
#minimax algorithm for pruning, taking the same parameters as the brute force algorithm but this time with alpha and beta
def minimax(self, board_state, maximizer, depth, alpha, beta):
self.state_space += 1
self.board._board = board_state
current_utility = None
chosen_utility = None
#static evaluations
winner = None
if depth > self.board._k - 1:
winner = self.board.winner
if self._player == Player.X:
if winner == Player.X:
return 1 / depth
if winner == Player.O:
return -1 / depth
else:
if winner == Player.O:
return 1 / depth
if winner == Player.X:
return -1 / depth
if depth == self.board.size ** 2:
return 0
#main recursion block
for i in range(self.board.size):
for j in range(self.board.size):
if self.board.cell(i, j) == CellState.EMPTY:
if self._player == Player.X:
#in any case where alpha < beta, the recursion will not be entered; this is because if the current greatest choice for the maximizer is
#less than the current least choice for the minimizer, there is no other step that can be taken by either player to prevent the outcome
#of the current branch, so the rest of the branch can be pruned
if maximizer and chosen_utility != 1 and alpha < beta:
current_utility = self.minimax(self.get_appended_board_state(board_state, 0, i, j), not maximizer, depth+1, alpha, beta)
board_state[i][j] = -1
self.board._board = board_state
#in the case of the maximizer, alpha is stored as the current greatest choice
if current_utility > alpha:
alpha = current_utility
if not maximizer and chosen_utility != -1 and alpha < beta:
current_utility = self.minimax(self.get_appended_board_state(board_state, 1, i, j), not maximizer, depth+1, alpha, beta)
board_state[i][j] = -1
self.board._board = board_state
#in the case of the minimizer, beta is stored as the current least choice
if current_utility < beta:
beta = current_utility
else:
if maximizer and chosen_utility != 1 and alpha < beta:
current_utility = self.minimax(self.get_appended_board_state(board_state, 1, i, j), not maximizer, depth+1, alpha, beta)
board_state[i][j] = -1
self.board._board = board_state
if current_utility > alpha:
alpha = current_utility
if not maximizer and chosen_utility != -1 and alpha < beta:
current_utility = self.minimax(self.get_appended_board_state(board_state, 0, i, j), not maximizer, depth+1, alpha, beta)
board_state[i][j] = -1
self.board._board = board_state
if current_utility < beta:
beta = current_utility
if current_utility is not None:
if maximizer:
if chosen_utility is None or current_utility > chosen_utility:
chosen_utility = current_utility
else:
if chosen_utility is None or current_utility < chosen_utility:
chosen_utility = current_utility
return chosen_utility
def next_move(self, given_board):
self.state_space = 0
self.board._k = given_board._k
self.board._size = given_board._size
best_move_i = -1
best_move_j = -1
maximum = -100
depth = given_board.size ** 2
#alpha and beta are initialized to negative and positive infinity
alpha = float("-inf")
beta = float("inf")
maximizer = True
for i in range(given_board.size):
for j in range(given_board.size):
if given_board.cell(i, j) == CellState.EMPTY:
depth -= 1
for i in range(given_board.size):
for j in range(given_board.size):
if given_board.cell(i, j) == CellState.EMPTY:
if self._player == Player.X:
#alpha and beta are passed into minimax
current = self.minimax(self.get_appended_board_state(deepcopy(given_board._board), 0, i, j), not maximizer, depth+1, alpha, beta)
else:
#alpha and beta are passed into minimax
current = self.minimax(self.get_appended_board_state(deepcopy(given_board._board), 1, i, j), not maximizer, depth+1, alpha, beta)
if current is not None:
if current > maximum:
maximum = current
best_move_i = i
best_move_j = j
print("State space traversed contained " + str(self.state_space) + " states")
#the best move is returned
return Move(self._player, best_move_i, best_move_j)
def test_tic_tac_toe_initialized_with_a_3_x_3_board():
new_board = Board()
assert len(new_board.board) == 9
class Game:
"""
represent game in tic tac toe
"""
def __init__(self):
self._board = Board()
def _build_tree(self):
"""
build tree with random move for computer
:return: None
"""
def recurse(p):
"""
make new board for branch
:param p: BSTNode
:return: None
"""
if p is not None and p.data.check_win() is None:
new_board = Board()
for i in range(3):
for j in range(3):
new_board.add_value(i, j,
p.data._game_board[i, j])
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
while not new_board.is_good_move(move1_x, move1_y):
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
new_board.second_player_move(move1_x, move1_y)
if new_board.check_win() is None:
while not new_board.is_good_move(move2_x,
move2_y) or (
move1_x == move2_x and move1_y == move2_y):
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
new_board.first_player_move(move2_x, move2_y)
self._builded_tree._add_right(p, new_board)
recurse(self._builded_tree.super_find(new_board))
if new_board.check_win() is None:
for i in range(3):
for j in range(3):
new_board.add_value(i, j,
p.data._game_board[i, j])
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
while not new_board.is_good_move(move1_x, move1_y):
move1_x = random.randint(0, 2)
move1_y = random.randint(0, 2)
new_board.second_player_move(move1_x, move1_y)
if new_board.check_win() is None:
while not new_board.is_good_move(move2_x,
move2_y) or (
move1_x == move2_x and move1_y == move2_y):
move2_x = random.randint(0, 2)
move2_y = random.randint(0, 2)
new_board.first_player_move(move2_x, move2_y)
self._builded_tree._add_left(p, new_board)
recurse(self._builded_tree.super_find(new_board))
return None
self._builded_tree = LinkedBST()
self._builded_tree.add(self._board)
recurse(self._builded_tree.find(self._board))
def search_max(self):
"""
calculate which move computer is better
:return: None
"""
def recurse(item):
"""
calculate how many points have branch
:param item: BSTNode
:return: int
"""
if item is None:
return 0
if item.data.check_win() is not None and item.right is None and item.left is None:
if item.data.check_win() == "draw":
return 0
elif item.data.check_win()[1] == "first player win":
return -1
else:
return 1
else:
try:
item.right
try:
item.left
return recurse(item.right) + recurse(item.left)
except:
return recurse(item.right)
except:
try:
item.left
return recurse(item.left)
except:
return 0
item = self._builded_tree._root
left_item = item.left
rigth_item = item.right
print(recurse(left_item), recurse(rigth_item))
if recurse(left_item) >= recurse(rigth_item):
return "left"
else:
return "right"
def move_computer(self):
"""
generate computer move
:return: None
"""
self._build_tree()
max = self.search_max()
if max == "left" and self._builded_tree._root.left is not None:
self._board = self._builded_tree._root.left.data
row = self._builded_tree._root.left.data._last_move[0]
col = self._builded_tree._root.left.data._last_move[1]
if self._board._game_board[row, col] == "x":
self._board.add_value(row, col, None)
elif self._builded_tree._root.right is not None:
self._board = self._builded_tree._root.right.data
row = self._builded_tree._root.right.data._last_move[0]
col = self._builded_tree._root.right.data._last_move[1]
if self._board._game_board[row, col] == "x":
self._board.add_value(row, col, None)
def run(self):
"""
run game
:return: None
"""
def check_win():
"""
check if someone win
:return: None
"""
if self._board.check_win() is not None:
if self._board.check_win() == "draw":
print("Draw!")
elif self._board.check_win()[1] == "first player win":
print("You win!")
else:
print("You lose!")
sys.exit()
print("Hello!")
print("Let`s play game/ You will be move first because if you move "
"second you will lose")
while not self._board.check_win():
move1 = int(input("Input your row move: "))
move2 = int(input("Input your col move: "))
while not self._board.is_good_move(move1, move2):
move1 = int(input("Reinput your row move: "))
move2 = int(input("Reinput your col move: "))
self._board.first_player_move(move1, move2)
check_win()
print("Your move\n", self._board)
self.move_computer()
print("Computer move\n", self._board)
check_win()
