def monte_carlo_ai_random_playouts(game_state: GameState, possible_actions: List[Action]) -> Action:
"""
Uses a Monte Carlo Tree Search with random playouts at the simulation step to determine
the best next move
:param game_state: Current GameState
:param possible_actions: Possible actions to take
:return: Next action to take
"""
# Do not monte carlo search if there is only one potential action
if len(possible_actions) == 1:
return possible_actions[0]
NUM_SIMS = 200
Qsa = {}
Nsa = {}
Ns = {}
visited = set()
# Simulate playouts
for num_playout in range(NUM_SIMS):
mcts_search(game_state, game_state.player_turn, game_state.calculate_worm_count().get(game_state.player_turn),
Qsa, Nsa, Ns, visited)
s = str(game_state)
sorted_actions = sorted(possible_actions, key=lambda a: Nsa.get((s, str(a)), float("-inf")))
return sorted_actions.pop()
def get_change_worm_count(game_state: GameState, player: int, orig_num_worms: int) -> int:
"""
Returns the difference between the original worm count and the worm count in the current
game state for a given player
:param game_state: Game state to get current worm counts from
:param player: Player number to calculate change in worm count for
:param orig_num_worms: Original worm count to calculate against
:return: Difference in worm count for player between the current worm count in game_state and orig_num_worms
"""
worm_counts = game_state.calculate_worm_count()
return worm_counts[player] - orig_num_worms
