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 _expand(self, node: GameState):
"Update the `children` dict with the children of `node`"
if node in self.children:
return # already expanded
self.children[node] = set(node.children())
