def heuristic_value(self, board, move):
data = [{'state': board.fen(), 'move': move}]
prec = md.preprocessor()
prec.fit(raw_data=data)
data = prec.transform()
theta = self.model.predict_proba(data)
return theta * eval.evaluate_board(board), theta
def pvs(self, depth, alpha, beta, color):
if self.hash_board(depth, color) in self.cache:
return self.cache[self.hash_board(depth, color)]
if depth == 0 or not self.board.legal_moves:
self.cache[self.hash_board(depth,
color)] = evaluate_board(self.board)
return self.cache[self.hash_board(depth, color)]
if time.time() - self.start > self.total_time:
return alpha
moves = list(self.board.legal_moves)
for move in moves:
self.board.push(move)
if moves.index(move) == 0:
score = -self.pvs(depth - 1, -beta, -alpha, -color)
else:
score = -self.pvs(depth - 1, -alpha - 1, -alpha, -color)
if alpha < score < beta:
score = -self.pvs(depth - 1, -beta, -score, -color)
if alpha <= score:
alpha = score
self.board.pop()
if alpha >= beta:
break
return alpha
def record(self, board, score):
self.visits["total"] = self.visits.get("total", 1) + 1
self.visits[board.fen()] = self.visits.get(board.fen(), 0) + 1
#dataset = {'input': np.asarray(list(board.fen().encode('utf8'))), 'target': score}
#self.data.append(dataset)
self.log('Visit Recorded', msg_type=2)
#return self.model.fit(dataset = self.data)
return eval.evaluate_board(board)
def play_value(board, movehistory=None):
print('play_value')
if board.is_checkmate():
record(board, eval.evaluate_board(board))
return eval.evaluate_board(board)
heuristic_vals = {}
for move in board.pseudo_legal_moves:
board.push(move)
heuristic_vals[move] = -heuristic_value(board)
board.pop()
move = max(heuristic_vals, key=heuristic_vals.get)
board.push(move)
value = -play_value(board)
board.pop()
record(board, value)
return value
def play_value(self, board, depth=5):
if board.is_checkmate() or depth == 0:
self.record(board, eval.evaluate_board(board))
return eval.evaluate_board(board)
heuristic_vals = {}
for move in board.legal_moves:
board.push(move)
val = self.heuristic_value(board, move)
if val is not None:
heuristic_vals[move] = val[0]
board.pop()
move = max(heuristic_vals, key=heuristic_vals.get)
board.push(move)
value = self.play_value(board, depth=depth - 1)
#print("value" + str(value))
board.pop()
self.record(board, value)
#self.log('Playout Complete')
return value
def quiesce(alpha, beta, board):
# need to import evaluate.py
stand_pat = eval.evaluate_board(board)
if (stand_pat >= beta):
return beta
if (alpha < stand_pat):
alpha = stand_pat
for move in board.legal_moves:
if board.is_capture(move):
board.push(move)
score = -quiesce(-beta, -alpha, board)
board.pop()
if (score >= beta):
return beta
if (score > alpha):
alpha = score
return alpha