def test_bucket_small_04(self):
# Round #4, B (93), A (-93)
# B posts the blind of 1
# A posts the blind of 2
# B dealt Qs 8c [Ts Jc]
# A dealt Js Kh [4s Ah]
# B calls
# A checks
# Flop 7d 2h 5s [Qd 2h 6s], B (2), A (2)
# A checks
# B bets 2
sb_index = 1
round_state = create_new_round(sb_index)
print("Start of preflop, no actions yet")
infoset = make_infoset(round_state, 1, True)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
round_state = round_state.proceed(CallAction())
print("SB called on preflop")
infoset = make_infoset(round_state, 0, False)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
round_state = round_state.proceed(CheckAction())
print("BB checked on preflop, first action of flop")
infoset = make_infoset(round_state, 0, False)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
round_state = round_state.proceed(CheckAction())
print("BB checked on flop")
infoset = make_infoset(round_state, 1, True)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
round_state = round_state.proceed(RaiseAction(2))
print("SB bet 2")
infoset = make_infoset(round_state, 0, False)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
def test_exceed_action_limit(self):
# P2 is the small blind.
sb_index = 1
round_state = create_new_round(sb_index)
# SB calls.
round_state = round_state.proceed(CallAction())
# BB checks.
round_state = round_state.proceed(CheckAction())
# Do 8 bets/raises to exceed the max 6 actions.
round_state = round_state.proceed(RaiseAction(2))
round_state = round_state.proceed(RaiseAction(4))
round_state = round_state.proceed(RaiseAction(6))
round_state = round_state.proceed(RaiseAction(8))
infoset = make_infoset(round_state, 0, False)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
round_state = round_state.proceed(RaiseAction(10))
infoset = make_infoset(round_state, 1, True)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
round_state = round_state.proceed(RaiseAction(12))
round_state = round_state.proceed(RaiseAction(14))
infoset = make_infoset(round_state, 1, True)
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
round_state = round_state.proceed(RaiseAction(16))
round_state = round_state.proceed(CallAction())
infoset = make_infoset(round_state, 0, False)
expected = torch.Tensor(
[1, 2, 1, 0, 0, 0, 2, 4, 14, 12, 0, 0, 0, 0, 0, 0, 0, 0])
self.assertTrue((infoset.bet_history_vec == expected).all())
bucket = bucket_small(infoset)
print(bucket_small_join(bucket))
def get_strategy(self, infoset, valid_mask):
"""
Does regret matching to return a probabilistic strategy.
"""
bucket = bucket_small(infoset)
bstring = bucket_small_join(bucket)
if bstring not in self._regrets:
self._regrets[bstring] = torch.zeros(Constants.NUM_ACTIONS)
total_regret = self._regrets[bstring]
with torch.no_grad():
r_plus = torch.clamp(total_regret, min=0)
if r_plus.sum() < 1e-3:
return torch.ones(Constants.NUM_ACTIONS) / Constants.NUM_ACTIONS
else:
return r_plus / r_plus.sum()
def add_regret(self, infoset, r):
"""
Adds an instantaneous regret to total regret.
"""
assert (len(r) == Constants.NUM_ACTIONS)
# t0 = time.time()
bucket = bucket_small(infoset)
bstring = bucket_small_join(bucket)
# elapsed = time.time() - t0
# print("Bucket=", elapsed)
if bstring not in self._regrets:
self._regrets[bstring] = torch.zeros(Constants.NUM_ACTIONS)
# CFR+ regret matching.
# https://arxiv.org/pdf/1407.5042.pdf
self._regrets[bstring] = torch.max(torch.zeros(Constants.NUM_ACTIONS),
self._regrets[bstring] + r)
def test_bucket_small_03(self):
# BB.T.H2|x.R.R.x|R.R.x.x|2P.x.x.x': tensor([168., 160., 142., 181., 152., 144.])
random.seed(123)
sb_index = 0
round_state = create_new_round(sb_index)
round_state = round_state.proceed(RaiseAction(4)) # SB raises.
infoset = make_infoset(round_state, 1, False)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(8)) # BB raises.
infoset = make_infoset(round_state, 0, True)
print("SB raises, BB raises")
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(12)) # SB raises.
infoset = make_infoset(round_state, 1, False)
print("SB raises, BB raises, SB raises")
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(16)) # BB raises.
round_state = round_state.proceed(RaiseAction(20)) # SB raises.
print("SB raise, BB raise, SB raise, BB raise, SB raise")
infoset = make_infoset(round_state, 1, False)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(CallAction()) # BB calls.
infoset = make_infoset(round_state, 1, False)
print("flop, no actions yet")
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(CheckAction()) # BB checks.
print("flop, BB checked")
infoset = make_infoset(round_state, 0, True)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(CheckAction()) # SB checks.
print("SB checked, now on turn")
infoset = make_infoset(round_state, 1, False)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(10)) # BB raises.
round_state = round_state.proceed(CallAction()) # SB calls.
print("turn, BB raised, SB called, now on river")
infoset = make_infoset(round_state, 1, False)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(CheckAction()) # BB checks.
print("river, BB checks")
infoset = make_infoset(round_state, 0, True)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(10)) # SB raises.
print("river, BB checks, SB raises")
infoset = make_infoset(round_state, 1, False)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(40)) # BB raises.
print("river, BB checks, SB raises, BB raises")
infoset = make_infoset(round_state, 1, False)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(60)) # SB raises.
round_state = round_state.proceed(RaiseAction(100)) # BB raises.
print("river, BB check, SB raise, BB raise, SB raise, BB raise")
infoset = make_infoset(round_state, 0, True)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)
round_state = round_state.proceed(RaiseAction(120)) # SB raises.
infoset = make_infoset(round_state, 1, False)
bucket = bucket_small_join(bucket_small(infoset))
print(bucket)