def test_simulate_random():
for _ in range(20):
players = [RandomPolicy(), RandomPolicy(), RandomPolicy()]
starting_stacks = choose_starting_stacks()
deal: FullDeal = dealer.deal_cards(len(players))
game, results = simulate(players, starting_stacks, deal)
result.get_result(deal, game.view())
def test_loss():
deal = FullDeal(
hole_cards=[mkhand("AcAd"), mkhand("AsKh")],
board=Board(flop=mkflop("KsJs3d"),
turn=mkcard("7s"),
river=mkcard("6s")),
)
game = GameRunner(starting_stacks=[300, 400])
game.start_game()
# Preflop
game.bet_raise(to=10)
game.call()
# Flop
game.bet_raise(to=25)
game.call()
# Turn
game.check()
game.bet_raise(to=40)
game.call()
# River
game.check()
game.check()
results = result.get_result(deal, game.game_view())
model = heads_up.HeadsUpModel("HeadsUp")
player_index = 1
example = make_forward_backward_example(
player_index,
game.game_view(),
deal.hole_cards[player_index],
deal.board,
results,
)
(
feature_tensors,
target_tensors,
) = model.feature_config.make_features_and_target_tensors(
[example.SerializeToString()])
print(tf.cast(target_tensors["reward__cumulative_reward"], tf.float32))
loss = model.loss(feature_tensors, target_tensors)
batch_size = 1
num_steps = 9
assert loss.shape == [batch_size, num_steps]
def test_game_result():
deal = FullDeal(
hole_cards=[mkhand("AcAh"),
mkhand("KdKs"),
mkhand("JhJd")],
board=Board(flop=mkflop("6dQc2s"),
turn=mkcard("6s"),
river=mkcard("3c")),
)
game = GameRunner(starting_stacks=[100, 200, 300])
game.start_game()
# Preflop
game.bet_raise(to=10)
game.call()
game.call()
# Flop
game.bet_raise(to=20)
game.call()
game.call()
# Turn
game.bet_raise(to=30)
game.call()
game.call()
# River
game.check()
game.check()
game.check()
result = get_result(deal, game.game_view())
assert result == Result(
won_hand=[True, False, False],
hand_results=[
EvaluationResult(
hand_type=HandType.TWO_PAIR,
kicker=33686528,
),
EvaluationResult(hand_type=HandType.TWO_PAIR, kicker=16909312),
EvaluationResult(hand_type=HandType.TWO_PAIR, kicker=4326400),
],
went_to_showdown=[True, True, True],
remained_in_hand=[True, True, True],
earned_from_pot=[180, 0, 0],
profits=[120, -60, -60],
)
def simulate(players: List[Policy], starting_stacks: List[int],
deal: FullDeal) -> Tuple[Game, Result]:
"""
Players are ordered by Small Blind, Big Blind, ..., Button
:param players:
:param starting_stacks:
:param deal:
:return:
"""
logger.debug("Simulating game with %s players", len(players))
game_runner = GameRunner(starting_stacks=starting_stacks)
logger.debug("Hole Cards: %s", deal.hole_cards)
game_runner.start_game()
while True:
board = deal.board.at_street(game_runner.street())
player_index = game_runner.current_player()
player = players[player_index]
hand = deal.hole_cards[player_index]
action = player.select_action(player_index, game_runner.game_view(),
hand, board)
if action is None:
raise Exception("Invalid Action")
action_result = game_runner.advance(action)
if action_result.street == Street.HAND_OVER:
logger.debug("Hand Over")
break
try:
result = get_result(deal, game_runner.game_view())
except Exception as e:
for event in game_runner.game.events:
print(event)
raise e
logger.debug("Result: %s", result)
return game_runner.game, result
def test_game_with_tie():
deal = FullDeal(
hole_cards=[mkhand("Ac3h"),
mkhand("Ad4s"),
mkhand("5s5d")],
board=Board(flop=mkflop("AsAhKc"),
turn=mkcard("Kd"),
river=mkcard("7h")),
)
game = GameRunner(starting_stacks=[100, 200, 300])
game.start_game()
# Preflop
game.bet_raise(to=30)
game.call()
game.call()
# Flop
game.bet_raise(to=50)
game.call()
game.call()
# Turn
game.bet_raise(to=20)
game.call()
game.call()
# River
game.bet_raise(to=100)
game.call()
result = get_result(deal, game.game_view())
assert result == Result(
won_hand=[True, True, False],
hand_results=[
EvaluationResult(hand_type=HandType.FULL_HOUSE, kicker=33556480),
EvaluationResult(hand_type=HandType.FULL_HOUSE, kicker=33556480),
EvaluationResult(hand_type=HandType.TWO_PAIR, kicker=50331680),
],
remained_in_hand=[True, True, True],
went_to_showdown=[True, True, True],
earned_from_pot=[150, 350, 0],
profits=[50, 150, -200],
)
def test_fold_preflop() -> None:
# Player 3 wins with a flopped set of jacks.
deal = FullDeal(
hole_cards=[mkhand("AcAd"),
mkhand("AsKh"),
mkhand("JcJh")],
board=Board(flop=mkflop("KsJs3d"),
turn=mkcard("7s"),
river=mkcard("6s")),
)
game = GameRunner(starting_stacks=[200, 250, 300])
game.start_game()
game.bet_raise(to=10)
game.fold()
game.fold()
results = result.get_result(deal, game.game_view())
s = Stats()
s.update_stats(game.game_view(), results, player_id=0)
assert s == Stats(
num_hands=1,
reward=-1,
num_wins=0,
num_losses=1,
num_flops=0,
num_turns=0,
num_rivers=0,
num_showdowns=0,
num_decisions=1,
num_check=0,
num_call=0,
total_amount_called=0,
num_bet=0,
total_amount_bet=0,
num_fold=1,
)
s = Stats()
s.update_stats(game.game_view(), results, player_id=1)
assert s == Stats(
num_hands=1,
reward=-2,
num_wins=0,
num_losses=1,
num_flops=0,
num_turns=0,
num_rivers=0,
num_showdowns=0,
num_decisions=1,
num_check=0,
num_call=0,
total_amount_called=0,
num_bet=0,
total_amount_bet=0,
num_fold=1,
)
s = Stats()
s.update_stats(game.game_view(), results, player_id=2)
assert s == Stats(
num_hands=1,
reward=3,
num_wins=1,
num_losses=0,
num_flops=0,
num_turns=0,
num_rivers=0,
num_showdowns=0,
num_decisions=1,
num_check=0,
num_call=0,
total_amount_called=0,
num_bet=1,
total_amount_bet=10,
num_fold=0,
)
def test_full_hand() -> None:
# Player 3 wins with a flopped set of jacks.
deal = FullDeal(
hole_cards=[mkhand("AhKs"), mkhand("Kc10d"), mkhand("5h5c"), mkhand("7s3d")],
board=Board(flop=mkflop("KdJs3d"), turn=mkcard("7s"), river=mkcard("6s")),
)
game = GameRunner(starting_stacks=[300, 300, 300, 300])
game.start_game()
# Preflop
game.bet_raise(to=10)
game.fold()
game.bet_raise(to=25)
game.call()
game.call()
# Flop
game.bet_raise(to=40)
game.call()
game.fold()
# Turn
game.check()
game.bet_raise(to=60)
game.call()
# River
game.check()
game.bet_raise(to=100)
game.call()
results = result.get_result(deal, game.game_view())
player_index = 1
example = make_example(
public_context=make_public_context(game.game_view()),
private_context=make_private_context(deal.hole_cards[1]),
target=make_target(deal, results),
public_states=make_public_states(game.game_view(), deal.board),
player_states=make_player_states(
player_index, game.game_view(), deal.hole_cards[player_index], deal.board
),
last_actions=make_last_actions(game.game_view()),
next_actions=make_next_actions(game.game_view()),
rewards=make_rewards(game.game_view(), results),
)
example_dict = seq_example_to_dict(example)
assert example_dict == {
"context": {
"num_steps": [14],
"private_context__hand_encoded": [140],
"private_context__hole_card_0_rank": [13],
"private_context__hole_card_0_suit": [2],
"private_context__hole_card_1_rank": [10],
"private_context__hole_card_1_suit": [3],
"public_context__num_players": [4],
"public_context__starting_stack_sizes": [300, 300, 300, 300],
"target__hole_cards": [168, 140, 3, 36],
"target__total_rewards": [250, -225, -25, 0],
},
"features": {
"last_action__action_encoded": [
[-1],
[3],
[0],
[3],
[1],
[1],
[5],
[1],
[0],
[1],
[7],
[1],
[1],
[14],
],
"last_action__amount_added": [
[-1],
[10],
[0],
[24],
[23],
[15],
[40],
[40],
[0],
[0],
[60],
[60],
[0],
[100],
],
"last_action__amount_added_percent_of_remaining": [
[-1.0],
[0.03333333507180214],
[0.0],
[0.08026755601167679],
[0.07718120515346527],
[0.0517241396009922],
[0.145454540848732],
[0.145454540848732],
[0.0],
[0.0],
[0.25531914830207825],
[0.25531914830207825],
[0.0],
[0.5714285969734192],
],
"last_action__amount_raised": [
[-1],
[8],
[0],
[15],
[0],
[0],
[40],
[0],
[0],
[0],
[60],
[0],
[0],
[100],
],
"last_action__amount_raised_percent_of_pot": [
[-1.0],
[2.6666667461395264],
[0.0],
[1.1538461446762085],
[0.0],
[0.0],
[0.5333333611488342],
[0.0],
[0.0],
[0.0],
[0.3870967626571655],
[0.0],
[0.0],
[0.3636363744735718],
],
"last_action__move": [
[-1],
[5],
[3],
[5],
[4],
[4],
[5],
[4],
[3],
[4],
[5],
[4],
[4],
[5],
],
"next_action__action_encoded": [
[3],
[0],
[3],
[1],
[1],
[5],
[1],
[0],
[1],
[7],
[1],
[1],
[14],
[1],
],
"next_action__amount_added": [
[10],
[0],
[24],
[23],
[15],
[40],
[40],
[0],
[0],
[60],
[60],
[0],
[100],
[100],
],
"next_action__amount_raised": [
[8],
[0],
[15],
[0],
[0],
[40],
[0],
[0],
[0],
[60],
[0],
[0],
[100],
[0],
],
"next_action__move": [
[5],
[3],
[5],
[4],
[4],
[5],
[4],
[3],
[4],
[5],
[4],
[4],
[5],
[4],
],
"next_action__new_total_bet": [
[10],
[10],
[25],
[25],
[25],
[40],
[40],
[40],
[0],
[60],
[60],
[0],
[100],
[100],
],
"player_state__current_hand_rank": [
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[3652],
[-1],
[-1],
[3648],
[-1],
[-1],
[3648],
[-1],
],
"player_state__current_hand_strength": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.5105869770050049],
[-1.0],
[-1.0],
[0.51112300157547],
[-1.0],
[-1.0],
[0.51112300157547],
[-1.0],
],
"player_state__current_hand_type": [
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[2],
[-1],
[-1],
[2],
[-1],
[-1],
[2],
[-1],
],
"player_state__current_player_offset": [
[1],
[2],
[-1],
[0],
[1],
[-1],
[0],
[1],
[-1],
[0],
[-1],
[-1],
[0],
[-1],
],
"player_state__frac_hands_better": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.046253468841314316],
[-1.0],
[-1.0],
[0.07439613342285156],
[-1.0],
[-1.0],
[0.17171716690063477],
[-1.0],
],
"player_state__frac_hands_tied": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.005550416186451912],
[-1.0],
[-1.0],
[0.005797101650387049],
[-1.0],
[-1.0],
[0.005050505045801401],
[-1.0],
],
"player_state__frac_hands_worse": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.9481961131095886],
[-1.0],
[-1.0],
[0.9198067784309387],
[-1.0],
[-1.0],
[0.8232323527336121],
[-1.0],
],
"player_state__is_current_player": [
[0],
[0],
[0],
[1],
[0],
[0],
[1],
[0],
[0],
[1],
[0],
[0],
[1],
[0],
],
"player_state__win_prob_vs_any": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.8297058343887329],
[-1.0],
[-1.0],
[0.8359622955322266],
[-1.0],
[-1.0],
[0.8232323527336121],
[-1.0],
],
"player_state__win_prob_vs_better": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.17900000512599945],
[-1.0],
[-1.0],
[0.09700000286102295],
[-1.0],
[-1.0],
[0.0],
[-1.0],
],
"player_state__win_prob_vs_tied": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.052000001072883606],
[-1.0],
[-1.0],
[0.0],
[-1.0],
[-1.0],
[0.0],
[-1.0],
],
"player_state__win_prob_vs_worse": [
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[-1.0],
[0.8659999966621399],
[-1.0],
[-1.0],
[0.9010000228881836],
[-1.0],
[-1.0],
[1.0],
[-1.0],
],
"public_state__all_in_player_mask": [
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 0],
],
"public_state__amount_to_call": [
[1, 0, 2, 2],
[9, 8, 0, 10],
[9, 8, 0, 10],
[0, 23, 15, 25],
[0, 0, 15, 25],
[0, 0, 0, 0],
[0, 40, 40, 40],
[0, 0, 40, 40],
[0, 0, 0, 0],
[0, 0, 0, 0],
[60, 0, 60, 60],
[0, 0, 0, 0],
[0, 0, 0, 0],
[100, 0, 100, 100],
],
"public_state__current_player_mask": [
[0, 0, 1, 0],
[0, 0, 0, 1],
[1, 0, 0, 0],
[0, 1, 0, 0],
[0, 0, 1, 0],
[1, 0, 0, 0],
[0, 1, 0, 0],
[0, 0, 1, 0],
[1, 0, 0, 0],
[0, 1, 0, 0],
[1, 0, 0, 0],
[1, 0, 0, 0],
[0, 1, 0, 0],
[1, 0, 0, 0],
],
"public_state__flop_0_rank": [
[-1],
[-1],
[-1],
[-1],
[-1],
[13],
[13],
[13],
[13],
[13],
[13],
[13],
[13],
[13],
],
"public_state__flop_0_suit": [
[-1],
[-1],
[-1],
[-1],
[-1],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
],
"public_state__flop_1_rank": [
[-1],
[-1],
[-1],
[-1],
[-1],
[11],
[11],
[11],
[11],
[11],
[11],
[11],
[11],
[11],
],
"public_state__flop_1_suit": [
[-1],
[-1],
[-1],
[-1],
[-1],
[1],
[1],
[1],
[1],
[1],
[1],
[1],
[1],
[1],
],
"public_state__flop_2_rank": [
[-1],
[-1],
[-1],
[-1],
[-1],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
],
"public_state__flop_2_suit": [
[-1],
[-1],
[-1],
[-1],
[-1],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
[3],
],
"public_state__folded_player_mask": [
[0, 0, 0, 0],
[0, 0, 0, 0],
[0, 0, 0, 1],
[0, 0, 0, 1],
[0, 0, 0, 1],
[0, 0, 0, 1],
[0, 0, 0, 1],
[0, 0, 0, 1],
[0, 0, 1, 1],
[0, 0, 1, 1],
[0, 0, 1, 1],
[0, 0, 1, 1],
[0, 0, 1, 1],
[0, 0, 1, 1],
],
"public_state__min_raise_amount": [
[2],
[8],
[8],
[15],
[15],
[2],
[40],
[40],
[2],
[2],
[60],
[2],
[2],
[100],
],
"public_state__pot_size": [
[3],
[13],
[13],
[37],
[60],
[75],
[115],
[155],
[155],
[155],
[215],
[275],
[275],
[375],
],
"public_state__river_rank": [
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[6],
[6],
[6],
],
"public_state__river_suit": [
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[1],
[1],
[1],
],
"public_state__stack_sizes": [
[299, 298, 300, 300],
[299, 298, 290, 300],
[299, 298, 290, 300],
[275, 298, 290, 300],
[275, 275, 290, 300],
[275, 275, 275, 300],
[235, 275, 275, 300],
[235, 235, 275, 300],
[235, 235, 275, 300],
[235, 235, 275, 300],
[235, 175, 275, 300],
[175, 175, 275, 300],
[175, 175, 275, 300],
[175, 75, 275, 300],
],
"public_state__street": [
[1],
[1],
[1],
[1],
[1],
[2],
[2],
[2],
[3],
[3],
[3],
[4],
[4],
[4],
],
"public_state__turn_rank": [
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[7],
[7],
[7],
[7],
[7],
[7],
],
"public_state__turn_suit": [
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[-1],
[1],
[1],
[1],
[1],
[1],
[1],
],
"reward__cumulative_reward": [
[-25],
[0],
[251],
[-223],
[-15],
[275],
[-200],
[0],
[315],
[-160],
[315],
[375],
[-100],
[375],
],
"reward__instant_reward": [
[-10],
[0],
[-24],
[-23],
[-15],
[-40],
[-40],
[0],
[0],
[-60],
[-60],
[0],
[-100],
[375],
],
"reward__is_players_last_action": [
[0],
[1],
[0],
[0],
[0],
[0],
[0],
[1],
[0],
[0],
[0],
[0],
[1],
[1],
],
"reward__won_hand": [
[0],
[0],
[1],
[0],
[0],
[1],
[0],
[0],
[1],
[0],
[1],
[1],
[0],
[1],
],
},
}
def test_fold_preflop() -> None:
# Player 3 wins with a flopped set of jacks.
deal = FullDeal(
hole_cards=[mkhand("AcAd"), mkhand("AsKh"), mkhand("JcJh")],
board=Board(flop=mkflop("KsJs3d"), turn=mkcard("7s"), river=mkcard("6s")),
)
game = GameRunner(starting_stacks=[200, 250, 100])
game.start_game()
game.bet_raise(to=10)
game.fold()
game.fold()
results = result.get_result(deal, game.game_view())
player_index = 1
example = make_example(
public_context=make_public_context(game.game_view()),
private_context=make_private_context(deal.hole_cards[1]),
target=make_target(deal, results),
public_states=make_public_states(game.game_view(), deal.board),
player_states=make_player_states(
player_index, game.game_view(), deal.hole_cards[player_index], deal.board
),
last_actions=make_last_actions(game.game_view()),
next_actions=make_next_actions(game.game_view()),
rewards=make_rewards(game.game_view(), results),
)
example_dict = seq_example_to_dict(example)
assert example_dict == {
"context": {
"num_steps": [3],
"private_context__hand_encoded": [168],
"private_context__hole_card_0_rank": [14],
"private_context__hole_card_0_suit": [1],
"private_context__hole_card_1_rank": [13],
"private_context__hole_card_1_suit": [4],
"public_context__num_players": [3],
"public_context__starting_stack_sizes": [200, 250, 100],
"target__hole_cards": [12, 168, 9],
"target__total_rewards": [-1, -2, 3],
},
"features": {
"last_action__action_encoded": [[-1], [4], [0]],
"last_action__amount_added": [[-1], [10], [0]],
"last_action__amount_added_percent_of_remaining": [
[-1.0],
[0.10000000149011612],
[0.0],
],
"last_action__amount_raised": [[-1], [8], [0]],
"last_action__amount_raised_percent_of_pot": [
[-1.0],
[2.6666667461395264],
[0.0],
],
"last_action__move": [[-1], [5], [3]],
"next_action__action_encoded": [[4], [0], [0]],
"next_action__amount_added": [[10], [0], [0]],
"next_action__amount_raised": [[8], [0], [0]],
"next_action__move": [[5], [3], [3]],
"next_action__new_total_bet": [[10], [10], [10]],
"player_state__current_hand_rank": [[-1], [-1], [-1]],
"player_state__current_hand_strength": [[-1.0], [-1.0], [-1.0]],
"player_state__current_hand_type": [[-1], [-1], [-1]],
"player_state__current_player_offset": [[1], [-1], [0]],
"player_state__frac_hands_better": [[-1.0], [-1.0], [-1.0]],
"player_state__frac_hands_tied": [[-1.0], [-1.0], [-1.0]],
"player_state__frac_hands_worse": [[-1.0], [-1.0], [-1.0]],
"player_state__is_current_player": [[0], [0], [1]],
"player_state__win_prob_vs_any": [[-1.0], [-1.0], [-1.0]],
"player_state__win_prob_vs_better": [[-1.0], [-1.0], [-1.0]],
"player_state__win_prob_vs_tied": [[-1.0], [-1.0], [-1.0]],
"player_state__win_prob_vs_worse": [[-1.0], [-1.0], [-1.0]],
"public_state__all_in_player_mask": [[0, 0, 0], [0, 0, 0], [0, 0, 0]],
"public_state__amount_to_call": [[1, 0, 2], [9, 8, 0], [9, 8, 0]],
"public_state__current_player_mask": [[0, 0, 1], [1, 0, 0], [0, 1, 0]],
"public_state__flop_0_rank": [[-1], [-1], [-1]],
"public_state__flop_0_suit": [[-1], [-1], [-1]],
"public_state__flop_1_rank": [[-1], [-1], [-1]],
"public_state__flop_1_suit": [[-1], [-1], [-1]],
"public_state__flop_2_rank": [[-1], [-1], [-1]],
"public_state__flop_2_suit": [[-1], [-1], [-1]],
"public_state__folded_player_mask": [[0, 0, 0], [0, 0, 0], [1, 0, 0]],
"public_state__min_raise_amount": [[2], [8], [8]],
"public_state__pot_size": [[3], [13], [13]],
"public_state__river_rank": [[-1], [-1], [-1]],
"public_state__river_suit": [[-1], [-1], [-1]],
"public_state__stack_sizes": [
[199, 248, 100],
[199, 248, 90],
[199, 248, 90],
],
"public_state__street": [[1], [1], [1]],
"public_state__turn_rank": [[-1], [-1], [-1]],
"public_state__turn_suit": [[-1], [-1], [-1]],
"reward__cumulative_reward": [[3], [0], [0]],
"reward__instant_reward": [[3], [0], [0]],
"reward__is_players_last_action": [[1], [1], [1]],
"reward__won_hand": [[1], [0], [0]],
},
}