|
| 1 | +"""Tests for the Bayesian Forgiver strategy.""" |
| 2 | + |
| 3 | +import axelrod as axl |
| 4 | +from axelrod.tests.property import strategy_lists |
| 5 | + |
| 6 | +from .test_player import TestPlayer |
| 7 | + |
| 8 | +C, D = axl.Action.C, axl.Action.D |
| 9 | + |
| 10 | + |
| 11 | +class TestBayesianForgiver(TestPlayer): |
| 12 | + """Test suite for BayesianForgiver strategy.""" |
| 13 | + |
| 14 | + name = "Bayesian Forgiver: 1.0, 1.0, 0.45, 2.5" |
| 15 | + player = axl.BayesianForgiver |
| 16 | + expected_classifier = { |
| 17 | + "memory_depth": float("inf"), |
| 18 | + "stochastic": False, |
| 19 | + "makes_use_of": set(), |
| 20 | + "inspects_source": False, |
| 21 | + "manipulates_source": False, |
| 22 | + "manipulates_state": False, |
| 23 | + } |
| 24 | + |
| 25 | + def test_initial_strategy(self): |
| 26 | + """Test that the strategy starts by cooperating.""" |
| 27 | + actions = [(C, C)] |
| 28 | + self.versus_test(axl.Cooperator(), expected_actions=actions) |
| 29 | + |
| 30 | + def test_vs_cooperator(self): |
| 31 | + """Test behavior against always cooperate - should cooperate.""" |
| 32 | + actions = [(C, C), (C, C), (C, C), (C, C), (C, C)] |
| 33 | + self.versus_test(axl.Cooperator(), expected_actions=actions) |
| 34 | + |
| 35 | + def test_vs_defector(self): |
| 36 | + """Test behavior against always defect - should eventually defect back.""" |
| 37 | + # First move: cooperate |
| 38 | + # After seeing D, Bayesian model updates: alpha=1, beta=2, mean=0.33 |
| 39 | + # Even with high uncertainty, mean is below threshold, so defect |
| 40 | + actions = [(C, D), (D, D), (D, D), (D, D), (D, D)] |
| 41 | + self.versus_test(axl.Defector(), expected_actions=actions) |
| 42 | + |
| 43 | + def test_vs_alternator(self): |
| 44 | + """Test behavior against alternating strategy.""" |
| 45 | + # With parameters (1.0, 1.0, 0.45, 2.5): |
| 46 | + # Early uncertainty is high, so threshold is high, but mean cooperation ~0.5 |
| 47 | + # When opponent defects, we evaluate: mean < threshold → defect |
| 48 | + # This creates an alternating pattern |
| 49 | + actions = [(C, C), (C, D), (D, C), (C, D), (D, C), (C, D)] |
| 50 | + self.versus_test(axl.Alternator(), expected_actions=actions) |
| 51 | + |
| 52 | + def test_vs_tit_for_tat(self): |
| 53 | + """Test behavior against Tit For Tat - should achieve mutual cooperation.""" |
| 54 | + actions = [(C, C), (C, C), (C, C), (C, C), (C, C)] |
| 55 | + self.versus_test(axl.TitForTat(), expected_actions=actions) |
| 56 | + |
| 57 | + def test_vs_suspicious_tit_for_tat(self): |
| 58 | + """Test against Suspicious Tit For Tat (starts with D).""" |
| 59 | + # STF starts with D, then mirrors |
| 60 | + # BF starts with C, sees D → defects back |
| 61 | + # This creates an alternating pattern |
| 62 | + actions = [(C, D), (D, C), (C, D), (D, C), (C, D)] |
| 63 | + self.versus_test(axl.SuspiciousTitForTat(), expected_actions=actions) |
| 64 | + |
| 65 | + def test_vs_grudger(self): |
| 66 | + """Test behavior against Grudger.""" |
| 67 | + # Grudger cooperates until any defection, then defects forever |
| 68 | + # BayesianForgiver should cooperate, so mutual cooperation |
| 69 | + actions = [(C, C), (C, C), (C, C), (C, C), (C, C)] |
| 70 | + self.versus_test(axl.Grudger(), expected_actions=actions) |
| 71 | + |
| 72 | + def test_vs_random(self): |
| 73 | + """Test behavior against Random strategy with specific seed.""" |
| 74 | + # Random behavior - BayesianForgiver should adapt |
| 75 | + # With seed=3, Random plays: C, C, C, C, D |
| 76 | + actions = [(C, C), (C, C), (C, C), (C, C), (C, D)] |
| 77 | + self.versus_test(axl.Random(), expected_actions=actions, seed=3) |
| 78 | + |
| 79 | + def test_vs_mock_single_defection(self): |
| 80 | + """Test response to a single defection.""" |
| 81 | + # Opponent cooperates then defects once then cooperates |
| 82 | + # With new parameters: after seeing one D among many Cs, still punishes |
| 83 | + # but quickly returns to cooperation |
| 84 | + opponent = axl.MockPlayer(actions=[C, C, C, D, C, C, C]) |
| 85 | + actions = [ |
| 86 | + (C, C), # Both cooperate |
| 87 | + (C, C), # Both cooperate |
| 88 | + (C, C), # Both cooperate |
| 89 | + (C, D), # Opponent defects - alpha=4, beta=2 |
| 90 | + (D, C), # Punish the defection |
| 91 | + (C, C), # Resume cooperation |
| 92 | + (C, C), # Continue cooperation |
| 93 | + ] |
| 94 | + self.versus_test(opponent, expected_actions=actions) |
| 95 | + |
| 96 | + def test_vs_mock_consistent_defector(self): |
| 97 | + """Test punishment of consistent defector.""" |
| 98 | + # Opponent defects consistently |
| 99 | + opponent = axl.MockPlayer(actions=[D, D, D, D, D, D]) |
| 100 | + actions = [ |
| 101 | + (C, D), # Start optimistic, see D - alpha=1, beta=2 |
| 102 | + (D, D), # Mean = 0.33, below threshold even with uncertainty |
| 103 | + (D, D), # Continue defecting |
| 104 | + (D, D), # Continue defecting |
| 105 | + (D, D), # Continue defecting |
| 106 | + (D, D), # Continue defecting |
| 107 | + ] |
| 108 | + self.versus_test(opponent, expected_actions=actions) |
| 109 | + |
| 110 | + def test_vs_mock_mixed_behavior(self): |
| 111 | + """Test adaptation to mixed behavior.""" |
| 112 | + # Opponent with mixed C and D: [C, C, D, C, D, C, C, D] |
| 113 | + # With default parameters, responds more reactively to defections |
| 114 | + opponent = axl.MockPlayer(actions=[C, C, D, C, D, C, C, D]) |
| 115 | + actions = [ |
| 116 | + (C, C), # Start cooperating |
| 117 | + (C, C), # Continue cooperating |
| 118 | + (C, D), # Opponent defects |
| 119 | + (D, C), # Punish defection |
| 120 | + (C, D), # Opponent cooperated, we cooperate, they defect |
| 121 | + (D, C), # Punish again |
| 122 | + (C, C), # Resume cooperation |
| 123 | + (C, D), # Cooperate, opponent defects |
| 124 | + ] |
| 125 | + self.versus_test(opponent, expected_actions=actions) |
| 126 | + |
| 127 | + def test_parameter_changes(self): |
| 128 | + """Test that different parameters affect behavior.""" |
| 129 | + # Test with different prior - more pessimistic |
| 130 | + player = self.player(prior_alpha=1.0, prior_beta=2.0) |
| 131 | + opponent = axl.Defector() |
| 132 | + match = axl.Match([player, opponent], turns=3) |
| 133 | + result = match.play() |
| 134 | + # With pessimistic prior and defector opponent, should defect quickly |
| 135 | + self.assertEqual(result[0], (C, D)) # First move still C |
| 136 | + self.assertEqual(result[1][0], D) # Should defect after seeing D |
| 137 | + |
| 138 | + def test_reset(self): |
| 139 | + """Test that reset properly reinitializes the strategy.""" |
| 140 | + player = self.player() |
| 141 | + opponent = axl.Cooperator() |
| 142 | + |
| 143 | + # Play some rounds - opponent cooperates so alpha should increase |
| 144 | + for _ in range(5): |
| 145 | + player.strategy(opponent) |
| 146 | + opponent.strategy(player) |
| 147 | + player.update_history(C, C) |
| 148 | + opponent.update_history(C, C) |
| 149 | + |
| 150 | + # Alpha should have changed (beta stays the same since opponent always cooperates) |
| 151 | + self.assertNotEqual(player.alpha, player.prior_alpha) |
| 152 | + |
| 153 | + # Reset |
| 154 | + player.reset() |
| 155 | + |
| 156 | + # Should be back to initial values |
| 157 | + self.assertEqual(player.alpha, player.prior_alpha) |
| 158 | + self.assertEqual(player.beta, player.prior_beta) |
| 159 | + |
| 160 | + def test_clone(self): |
| 161 | + """Test that cloning preserves parameters.""" |
| 162 | + player = self.player( |
| 163 | + prior_alpha=3.0, |
| 164 | + prior_beta=2.0, |
| 165 | + base_forgiveness_threshold=0.4, |
| 166 | + uncertainty_factor=2.0, |
| 167 | + ) |
| 168 | + clone = player.clone() |
| 169 | + |
| 170 | + self.assertEqual(clone.prior_alpha, 3.0) |
| 171 | + self.assertEqual(clone.prior_beta, 2.0) |
| 172 | + self.assertEqual(clone.base_forgiveness_threshold, 0.4) |
| 173 | + self.assertEqual(clone.uncertainty_factor, 2.0) |
| 174 | + self.assertEqual(clone.alpha, 3.0) |
| 175 | + self.assertEqual(clone.beta, 2.0) |
0 commit comments