def test_legal_trips(self):
rnd = YanivRound(None, 1, None)
player = YanivPlayer(1, None)
player.hand = [
Card("S", "2"),
Card("D", "2"),
Card("C", "2"),
]
actions = rnd.get_legal_actions([player], 0)
self._assert_legal_actions(actions)
expectations = [
"yaniv",
"S2",
"D2",
"C2",
"C2D2",
"C2S2",
"D2S2",
"C2S2D2",
"C2D2S2",
"D2C2S2",
]
for action in expectations:
self.assertIn(action, actions)
self.assertEqual(len(expectations), len(actions))
def test_legal_straights(self):
rnd = YanivRound(None, 1, None)
player = YanivPlayer(1, None)
player.hand = [
Card("S", "9"),
Card("S", "T"),
Card("S", "J"),
Card("S", "Q"),
]
actions = rnd.get_legal_actions([player], 0)
self._assert_legal_actions(actions)
self.assertIn("S9STSJSQ", actions)
self.assertIn("STSJSQ", actions)
self.assertIn("S9STSJ", actions)
self.assertEqual(7, len(actions))
player.hand = [
Card("S", "8"),
Card("S", "9"),
Card("S", "T"),
Card("S", "J"),
Card("S", "Q"),
]
actions = rnd.get_legal_actions([player], 0)
self._assert_legal_actions(actions)
self.assertIn("S8S9STSJSQ", actions)
self.assertIn("S9STSJSQ", actions)
self.assertIn("S8S9STSJ", actions)
self.assertIn("STSJSQ", actions)
self.assertIn("S9STSJ", actions)
self.assertIn("S8S9ST", actions)
self.assertEqual(11, len(actions))
player.hand = [Card("S", "2")]
actions = rnd.get_legal_actions([player], 0)
self._assert_legal_actions(actions)
self.assertIn("yaniv", actions)
self.assertIn("S2", actions)
self.assertEqual(2, len(actions))
def test_get_legal_actions_pickup(self):
rnd = YanivRound(None, 1, None)
player = YanivPlayer(1, None)
player.hand = [
Card("S", "T"),
Card("S", "A"),
Card("D", "T"),
Card("S", "Q"),
]
rnd.discarding = False
actions = rnd.get_legal_actions([player], 0)
self._assert_legal_actions(actions)
self.assertEqual(utils.pickup_actions, actions)
def test_legal_pairs(self):
rnd = YanivRound(None, 1, None)
player = YanivPlayer(1, None)
player.hand = [
Card("S", "T"),
Card("S", "A"),
Card("D", "T"),
Card("S", "J"),
Card("S", "Q"),
]
actions = rnd.get_legal_actions([player], 0)
self._assert_legal_actions(actions)
self.assertIn("DTST", actions)
self.assertIn("STSJSQ", actions)
self.assertIn("ST", actions)
self.assertIn("DT", actions)
self.assertEqual(7, len(actions))
class YanivGame(object):
def __init__(self,
num_players=2,
single_step_actions=False,
allow_step_back=False):
self.allow_step_back = allow_step_back
self.np_random = np.random.RandomState()
self.num_players = num_players
self.payoffs = [0 for _ in range(self.num_players)]
# default config
self._end_after_n_deck_replacements = 0
self._end_after_n_steps = 0
self._early_end_reward = 0
self._use_scaled_negative_reward = False
self._use_scaled_positive_reward = False
self._max_negative_reward = -1
self._negative_score_cutoff = 50
self._starting_player = "random"
self._starting_hands = {}
self._single_step_actions = single_step_actions
def init_game(self):
"""Initialize players and state
Returns:
(tuple): Tuple containing:
(dict): The first state in one game
(int): Current player's id
"""
# Initalize payoffs
self.payoffs = [0 for _ in range(self.num_players)]
# Initialize a dealer that can deal cards
self.dealer = YanivDealer(self.np_random)
# Initialize four players to play the game
self.players = [
YanivPlayer(i, self.np_random) for i in range(self.num_players)
]
# deal with predefined starting hands
for pid, starting_hand in self._starting_hands.items():
player = self.players[pid]
if isinstance(starting_hand, list):
starting_hand = random.choices(starting_hand, k=1)[0]
assert len(starting_hand) == 10
cards = list(
filter(lambda c: str(c) in starting_hand, self.dealer.deck))
for card in cards:
player.hand.append(card)
self.dealer.deck.remove(card)
assert len(player.hand) == 5
# Deal 5 cards to each player
for player in self.players:
if player.player_id in self._starting_hands.keys():
continue
for _ in range(utils.INITIAL_NUMBER_OF_CARDS):
player.hand.append(self.dealer.draw_card())
for player in self.players:
player.save_starting_hand()
# Initialize a Round
self.round = YanivRound(
self.dealer,
self.num_players,
self.np_random,
starting_player=self.np_random.randint(0, self.num_players)
if self._starting_player == "random" else self._starting_player,
)
self.round.flip_top_card()
# Save the hisory for stepping back to the last state.
self.history = []
self.actions = []
player_id = self.round.current_player
state = self.get_state(player_id)
return state, player_id
def configure(self, config):
"""Specifiy some game specific parameters, such as player number"""
self._end_after_n_deck_replacements = config[
"end_after_n_deck_replacements"]
self._end_after_n_steps = config["end_after_n_steps"]
self._early_end_reward = config["early_end_reward"]
self._use_scaled_negative_reward = config["use_scaled_negative_reward"]
self._use_scaled_positive_reward = config["use_scaled_positive_reward"]
self._max_negative_reward = config["max_negative_reward"]
self._negative_score_cutoff = config["negative_score_cutoff"]
self._starting_player = config["starting_player"]
self._starting_hands = config["starting_hands"]
def step(self, action):
"""Get the next state
Args:
action (str): A specific action
Returns:
(tuple): Tuple containing:
(dict): next player's state
(int): next plater's id
"""
if self.allow_step_back:
# First snapshot the current state
his_dealer = deepcopy(self.dealer)
his_round = deepcopy(self.round)
his_players = deepcopy(self.players)
self.history.append((his_dealer, his_players, his_round))
self.players[self.round.current_player].actions.append(action)
self.actions.append(action)
if self._end_after_n_steps > 0 and len(
self.actions) >= self._end_after_n_steps:
self.round.winner = -1
self.round.is_over = True
if not self._single_step_actions or action == utils.YANIV_ACTION:
return self._step(action)
assert isinstance(
action, tuple), "in single step actions the action must be a tuple"
cur_player = self.round.current_player
_, next_player = self._step(action[0])
assert cur_player == next_player
return self._step(action[1])
def _step(self, action):
self.round.proceed_round(self.players, action)
player_id = self.round.current_player
state = self.get_state(player_id)
# end the game if repalce deck is required with everyone losing
if (self._end_after_n_deck_replacements > 0
and self.round.deck_replacements >=
self._end_after_n_deck_replacements):
self.round.winner = -1
self.round.is_over = True
return state, player_id
def step_back(self):
"""Return to the previous state of the game
Returns:
(bool): True if the game steps back successfully
"""
if not self.history:
return False
self.dealer, self.players, self.round = self.history.pop()
return True
def get_state(self, player_id):
"""Return player's state
Args:
player_id (int): player id
Returns:
(dict): The state of the player
"""
state = self.round.get_state(self.players, player_id)
state["player_num"] = self.get_player_num()
state["current_player"] = self.round.current_player
state["legal_actions"] = self._get_legal_actions(player_id)
return state
def get_payoffs(self):
"""Return the payoffs of the game
1 if game won
-(score / 50) otherwise
Returns:
(list): Each entry corresponds to the payoff of one player
"""
self.payoffs = []
if self.round.winner == -1:
self.payoffs = [
self._early_end_reward for _ in range(self.num_players)
]
elif self._use_scaled_negative_reward:
for score in self.round.scores:
if score == 0:
payoff = 1
else:
payoff = self._max_negative_reward * (
min(self._negative_score_cutoff, score) /
self._negative_score_cutoff)
self.payoffs.append(payoff)
else:
for score in self.round.scores:
if score == 0:
payoff = 1
else:
payoff = self._max_negative_reward
self.payoffs.append(payoff)
if self._use_scaled_positive_reward and 1 in self.payoffs:
winner = self.payoffs.index(1)
self.payoffs[winner] = abs(min(self.payoffs))
return self.payoffs
def get_legal_actions(self):
"""Return the legal actions for current player
Returns:
(list): A list of legal actions
"""
return self._get_legal_actions(self.round.current_player)
def _get_legal_actions(self, player_id):
legal_actions = self.round.get_legal_actions(self.players, player_id)
if not self._single_step_actions:
return legal_actions
joint_legal_actions = []
if utils.YANIV_ACTION in legal_actions:
joint_legal_actions.append(utils.YANIV_ACTION)
legal_actions.remove(utils.YANIV_ACTION)
for d, p in product(legal_actions, utils.pickup_actions):
joint_legal_actions.append((d, p))
return joint_legal_actions
def get_player_num(self):
"""Return the number of players in Limit Texas Hold'em
Returns:
(int): The number of players in the game
"""
return self.num_players
def get_action_num(self):
"""Return the number of applicable actions
Returns:
(int): The number of actions.
"""
if self._single_step_actions:
return len(utils.JOINED_ACTION_LIST)
else:
return len(utils.ACTION_LIST)
def get_player_id(self):
"""Return the current player's id
Returns:
(int): current player's id
"""
return self.round.current_player
def is_over(self):
"""Check if the game is over
Returns:
(boolean): True if the game is over
"""
return self.round.is_over
def render(self):
header = lambda x: "{:=^20}".format(" {} ".format(x))
print(header("STEP {}".format(len(self.actions))))
print("Discard Pile: {}".format(self.round.discard_pile))
player_f = "PLAYER {}"
cur_f = "*{}*".format(player_f)
for player in self.players:
print(
header(
(player_f if player.player_id != self.round.current_player
else cur_f).format(player.player_id)))
print("Actions: {}".format(", ".join(player.actions)))
print("Hand: " + " ".join([c.__str__() for c in player.hand]))
print()