def action(self, game: GameState):
if game.game_over():
return None
possibleMoves = game.legal_moves()
boards = game.children()
win_h = self.win_heuristic(boards)
if not win_h == -1:
return possibleMoves[win_h], 1
evals = self.heuristic.hs(boards)
evals = self.default_heuristic(boards, evals)
if self.variance:
if not game.turn():
evals = [1 - e for e in evals]
total = sum(evals)
evals = [e / total for e in evals]
choice = random.random()
s = 0
for idx, e in enumerate(evals):
s += e
if choice <= s:
bestIdx = idx
break
else:
if game.turn():
bestIdx = np.argmax(evals)
else:
bestIdx = np.argmin(evals)
bestMove = possibleMoves[bestIdx]
return bestMove, evals[bestIdx]
def play(self, game: GameState, verbose=False):
if self.alternate:
temp = self.actionModel1
self.actionModel1 = self.actionModel2
self.actionModel2 = temp
states = []
hs = []
if verbose:
print(game)
while not game.game_over():
actingModel = self.actionModel1 if game.turn() else self.actionModel2
move, h = actingModel.action(game)
states.append(game.__copy__())
hs.append(h)
game.push(move)
if verbose:
print("move: " + str(move) + " h: " + str(h))
print(game)
return game.winner(), states, hs
def h_score(self, game: GameState):
s = self._score(game)
return 1 - s if game.turn() else s
