-
Notifications
You must be signed in to change notification settings - Fork 3
/
test_game.py
124 lines (101 loc) · 3.95 KB
/
test_game.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
import numpy as np
from XFP import XFP, LeducRLEnv
seed = 100
np.random.seed(seed)
env = XFP()
env.compute_p2_best_response()
env.compute_p1_best_response()
p1 = 0.0
p2 = 0.0
for i in range(100):
cards = env.possible_cards_list[np.random.randint(24)]
game_state = ""
while True:
if game_state in env.ending:
v1, v2 = env.compute_payoff(cards, game_state)
p1 += v1
p2 += v2
# print cards, game_state, v1, v2
break
if game_state in env.round1_states_set:
rd = 1
else:
rd = 2
if game_state in env.player1_states_set:
# action = env.choose_action_p1(game_state, cards, rd)
if np.random.randint(0, 100) < 37:
action = 'C'
else:
action = 'B'
game_state = game_state + action
else:
action = env.choose_action_p2(game_state, cards, rd)
# if np.random.randint(0, 100) < 37:
# action = 'C'
# else:
# action = 'B'
game_state = game_state + action
print p1, p2
assert seed != 100 or p2 == 60 and seed == 100
# test realization
policy_p1 = {}
policy_p2 = {}
for key, _ in env.q_value1_final.iteritems():
p = np.random.rand()
policy_p1[key] = [p, 1.0 - p]
for key, _ in env.q_value2_final.iteritems():
p = np.random.rand()
policy_p2[key] = [p, 1.0 - p]
realization = XFP.compute_realization(policy_p1, policy_p2)
print XFP.tournament(seed, 10000, policy_p1, policy_p2)
print XFP.compute_payoff_given_realization(realization)
p1, p2 = XFP.compute_realization2policy(realization)
for key in policy_p1:
assert np.allclose(policy_p1[key], p1[key])
for key in policy_p2:
assert np.allclose(policy_p2[key], p2[key])
# test mix policy
env.finish()
env.opponent_policy_p2 = policy_p2
env.opponent_policy_p1 = policy_p1
env.compute_p1_best_response()
env.compute_p2_best_response()
player1_best_response_policy = XFP.convert_q_s_a2greedy_policy(env.q_value1_final)
player2_best_response_policy = XFP.convert_q_s_a2greedy_policy(env.q_value2_final)
env.finish()
# test best response with realization
env.opponent_realization_p1 = realization
env.opponent_realization_p2 = realization
env.opponent_realization_enable = True
env.compute_p1_best_response()
env.compute_p2_best_response()
player1_best_response_policy_given_realization = XFP.convert_q_s_a2greedy_policy(env.q_value1_final)
player2_best_response_policy_given_realization = XFP.convert_q_s_a2greedy_policy(env.q_value2_final)
for key, item in player1_best_response_policy.iteritems():
assert np.allclose(item, player1_best_response_policy_given_realization[key])
for key, item in player2_best_response_policy.iteritems():
assert np.allclose(item, player2_best_response_policy_given_realization[key])
env.finish()
env.opponent_realization_enable = False
print XFP.tournament(seed, 1000, player1_best_response_policy, policy_p2)
print XFP.tournament(seed, 1000, policy_p1, player2_best_response_policy)
realization_old = XFP.compute_realization(policy_p1, policy_p2)
realization_br1 = XFP.compute_realization(player1_best_response_policy, policy_p2)
realization_br2 = XFP.compute_realization(policy_p1, player2_best_response_policy)
mix_realization_p1 = XFP.mix_realization(realization_br1, realization_old, 0.5)
mix_realization_p2 = XFP.mix_realization(realization_br2, realization_old, 0.5)
new_policy_p1, p2_extra = XFP.compute_realization2policy(mix_realization_p1)
p1_extra, new_policy_p2 = XFP.compute_realization2policy(mix_realization_p2)
for key in policy_p1:
assert np.allclose(policy_p1[key], p1_extra[key])
for key in policy_p2:
assert np.allclose(policy_p2[key], p2_extra[key])
print XFP.tournament(seed, 1000, new_policy_p1, policy_p2)
print XFP.tournament(seed, 1000, policy_p1, new_policy_p2)
# test LeducRLEnv
leduc = LeducRLEnv(seed=seed)
print leduc.reset()
print leduc.act(1)
print leduc.act(1)
print leduc.act(1)
print leduc.act(1)