def run_game():
gm = game.Game()
boardState = Gameplay.board_state_from_board(gm.board)
while (not gm.is_over()):
Gameplay.show_board(boardState)
print("Current Player: " +
("White" if gm.whose_turn() == 1 else "Black"))
possible_moves_coord = list(
map(lambda x: Gameplay.move_from_pos(x, gm),
gm.get_possible_moves()))
print("Valid Moves: ")
for i in range(len(possible_moves_coord)):
print(str(i + 1) + ": " + (str(possible_moves_coord[i])))
for bd_st in Gameplay.board_states_from_possible_moves(gm.board):
Gameplay.show_board(bd_st)
move_idx = -1
while move_idx not in range(len(possible_moves_coord)):
usr_input = input("Pick a move: ")
move_idx = -1 if (usr_input == '') else (int(usr_input) - 1)
if move_idx not in range(len(possible_moves_coord)):
print("Illegal move")
move = gm.get_possible_moves()[move_idx]
if (move in gm.board.get_possible_capture_moves()):
print('capture move!')
piece_was_king = gm.board.searcher.get_piece_by_position(
move[0]).king
boardState = Gameplay.make_move(gm, move)
if (not piece_was_king
) and gm.board.searcher.get_piece_by_position(move[1]).king:
print('king move!')
print("Game Over! ")
if gm.move_limit_reached():
print("It's a tie!!")
else:
print("Winner is: " +
("White" if gm.get_winner() == 1 else "Black"))
def run_agent_duel(agt1, agt2, verbose=False):
players = [agt1, agt2]
gm = game.Game()
boardState = Gameplay.board_state_from_board(gm.board)
random.shuffle(players)
while (not gm.is_over()):
if verbose:
Gameplay.show_board(boardState)
print("Current Player: " +
("White" if gm.whose_turn() == 1 else "Black"))
possible_moves_coord = list(
map(lambda x: Gameplay.move_from_pos(x, gm),
gm.get_possible_moves()))
print("Valid Moves: ")
for i in range(len(possible_moves_coord)):
print(str(i + 1) + ": " + (str(possible_moves_coord[i])))
curr_agt = players[gm.whose_turn() - 1]
move, q_val = Gameplay.get_QAgent_move_pp(curr_agt, gm)
if verbose:
print("Agent {} picks {}: ".format(curr_agt.name, move + 1))
print("Agent {} Q-value {}: ".format(curr_agt.name, q_val))
boardState = Gameplay.make_move(gm, gm.get_possible_moves()[move])
if verbose:
print("Game Over! ")
if gm.move_limit_reached():
result = 'tie'
if verbose:
print("It's a tie!!")
else:
result = players[gm.get_winner() - 1].name
if verbose:
print("Winner is: {}".format(players[gm.get_winner() -
1].name))
return result
def run_game_with_agent(agent):
players = [HUMAN, AGENT]
gm = game.Game()
boardState = Gameplay.board_state_from_board(gm.board)
random.shuffle(players)
while (not gm.is_over()):
Gameplay.show_board(boardState)
print("Current Player: " +
("White" if gm.whose_turn() == 1 else "Black"))
possible_moves_coord = list(
map(lambda x: Gameplay.move_from_pos(x, gm),
gm.get_possible_moves()))
print("Valid Moves: ")
for i in range(len(possible_moves_coord)):
print(str(i + 1) + ": " + (str(possible_moves_coord[i])))
if (players[gm.whose_turn() - 1] == HUMAN):
move = -1
while move not in range(len(possible_moves_coord)):
usr_input = input("Pick a move: ")
move = -1 if (usr_input == '') else (int(usr_input) - 1)
if move not in range(len(possible_moves_coord)):
print("Illegal move")
print("Human picks {}: ".format(move))
else:
move, q_val = Gameplay.get_QAgent_move_pp(agent, gm)
print("Agent {} picks {}: ".format(agent.name, move + 1))
print("Agent {} Q-value {}: ".format(agent.name, q_val))
boardState = Gameplay.make_move(gm, gm.get_possible_moves()[move])
print("Game Over! ")
if gm.move_limit_reached():
print("It's a tie!!")
else:
print("Winner is: " +
("White" if gm.get_winner() == 1 else "Black"))
def auto_play(self, n_episodes):
plt.ion()
plt.xlabel('Episodes')
plt.ylabel('{} mean error'.format(self.name))
x, y = [], []
line, = plt.plot(x, y)
plt.xlim(0, n_episodes)
plt.ylim(0, config.PLOT_Y_LIM)
for i in range(n_episodes):
print("Episode {}".format(i))
turns_hist = {
1: [],
2: []
}
gm = game.Game()
boardState = Gameplay.board_state_from_board(gm.board)
while (not gm.is_over()):
player = gm.whose_turn()
possible_board_states = Gameplay.board_states_from_possible_moves(gm.board)
move_idx, q_val = Gameplay.get_QAgent_move_pp(self, gm)
draw_counter = gm.moves_since_last_capture
if (player == 2):
boardState = Gameplay.invert_board(boardState)
possible_board_states = np.array(
list(map(lambda x: Gameplay.invert_board(x), possible_board_states)))
# Updating previous history
if len(turns_hist[player]) > 0:
turns_hist[player][-1]['next_board_state'] = boardState
turns_hist[player][-1]['next_draw_counter'] = draw_counter
turns_hist[player][-1]['next_possible_board_states'] = possible_board_states
move = gm.get_possible_moves()[move_idx]
reward = 0
if (move in gm.board.get_possible_capture_moves()):
reward += config.CAPTURE_REWARD
piece_was_king = gm.board.searcher.get_piece_by_position(move[0]).king
new_boardState = Gameplay.make_move(gm, move)
if (not piece_was_king) and gm.board.searcher.get_piece_by_position(move[1]).king:
reward += config.KING_REWARD
if len(turns_hist[Gameplay.get_other_player(player)]) > 0:
turns_hist[Gameplay.get_other_player(player)][-1]['reward'] -= reward
# New history
turns_hist[player].append({
'board_state': boardState,
'draw_counter': draw_counter,
'board_state_action': new_boardState,
'reward': reward,
'next_board_state': None,
'next_draw_counter': None,
'next_possible_board_states': None,
'done': False
})
if (player == 2):
turns_hist[player][-1]['board_state_action'] = Gameplay.invert_board(new_boardState)
boardState = new_boardState
print("Game Over! ")
if gm.move_limit_reached():
print("It's a tie!!")
for j in range(2):
turns_hist[j + 1][-1]['reward'] += config.DRAW_REWARD
turns_hist[j + 1][-1]['done'] = True
else:
print("Winner is: {}".format(gm.get_winner()))
turns_hist[gm.get_winner()][-1]['reward'] += config.WIN_REWARD
turns_hist[gm.get_winner()][-1]['done'] = True
turns_hist[Gameplay.get_other_player(gm.get_winner())][-1]['reward'] -= config.WIN_REWARD
turns_hist[Gameplay.get_other_player(gm.get_winner())][-1]['done'] = True
for k, v in turns_hist.items():
print("Reward sum for {}: {}".format(k, sum(list(map(lambda x: x['reward'], v)))))
for k, v in turns_hist.items():
for turn_hist in v:
self.remember(turn_hist['board_state'], turn_hist['draw_counter'],
turn_hist['board_state_action'], turn_hist['reward'],
turn_hist['next_board_state'], turn_hist['next_draw_counter'],
turn_hist['next_possible_board_states'], turn_hist['done'])
if (len(self.memory) > self.replay_batch_size):
self.replay_memory()
y.append(self.loss_mean)
x.append(i)
line.set_data(x, y)
plt.draw()
plt.pause(0.000000001)
return self