def move_coordinates():
fen = Entry.query.first().board
s = State()
s.board = chess.Board(fen)
if not s.board.is_game_over():
source = int(request.args.get('from', default=''))
target = int(request.args.get('to', default=''))
promotion = True if request.args.get('promotion',
default='') == 'true' else False
move = s.board.san(
chess.Move(source,
target,
promotion=chess.QUEEN if promotion else None))
# MONTE:
move_uci = chess.Move(source,
target,
promotion=chess.QUEEN if promotion else None)
if move is not None and move != "":
print("human moves", move)
try:
s.board.push_san(move)
bk = Entry.query.update(dict(board=s.board.fen()))
db.session.commit()
if use_mc:
# MONTE: Note monte won't work on heroku bc it stores state;
ai_mc.push_move(move_uci)
computer_move()
except Exception:
traceback.print_exc()
fen = Entry.query.first().board
s.board = chess.Board(fen)
response = app.response_class(response=s.board.fen(), status=200)
print(s.board)
return response
print("GAME IS OVER")
response = app.response_class(response="game over", status=200)
return response
def computer_move():
aimove = None
fen = Entry.query.first().board
s = State()
s.board = chess.Board(fen)
if not use_mc:
# MINIMAX
possible_moves = ai.minimax(s.board)
probs = [x[1] for x in possible_moves]
moves = [x[0] for x in possible_moves]
probs = probs / np.sum(probs)
aimove = np.random.choice(moves, p=probs)
s.board.push(aimove)
else:
# MONTE: monte carlo agent
aimove_mc, val, improved_policy = ai_mc.select_move(MC_SEARCH_ITER)
s.board.push(chess.Move.from_uci(aimove_mc.a))
bk = Entry.query.update(dict(board=s.board.fen()))
db.session.commit()