Merge pull request #1145 from gaffney2010/master

Implemented Weiner, k41r from Axelrod's Second

Author: marcharper

axelrod/strategies/_strategies.py

@@ -9,7 +9,7 @@
     UnnamedStrategy, SteinAndRapoport, TidemanAndChieruzzi)
 from .axelrod_second import (
     Champion, Eatherley, Tester, Gladstein, Tranquilizer, MoreGrofman,
-    Kluepfel, Borufsen, Cave, WmAdams, GraaskampKatzen)
+    Kluepfel, Borufsen, Cave, WmAdams, GraaskampKatzen, Weiner)
 from .backstabber import BackStabber, DoubleCrosser
 from .better_and_better import BetterAndBetter
 from .bush_mosteller import BushMosteller
@@ -270,6 +270,7 @@
     Tullock,
     TwoTitsForTat,
     VeryBad,
+    Weiner,
     Willing,
     Winner12,
     Winner21,

axelrod/strategies/axelrod_second.py

@@ -762,6 +762,7 @@ def strategy(self, opponent: Player) -> Action:
 
         number_defects = opponent.defections
         perc_defects = number_defects / turn
+
         # Defect if the opponent has defected often or appears random.
         if turn > 39 and perc_defects > 0.39:
             return D
@@ -812,11 +813,13 @@ def strategy(self, opponent: Player) -> Action:
         if opponent.history[0] == D:
             number_defects -= 1
 
-        if number_defects in [4, 7, 9]: return D
+        if number_defects in [4, 7, 9]:
+            return D
         if number_defects > 9 and opponent.history[-1] == D:
             return random_choice((0.5) ** (number_defects - 9))
         return C
 
+
 class GraaskampKatzen(Player):
     """
     Strategy submitted to Axelrod's second tournament by Jim Graaskamp and Ken
@@ -880,3 +883,95 @@ def strategy(self, opponent: Player) -> Action:
             return D
 
         return opponent.history[-1] # Tit-for-Tat
+
+
+class Weiner(Player):
+    """
+    Strategy submitted to Axelrod's second tournament by Herb Weiner (K41R),
+    and came in seventh in that tournament.
+
+    Play Tit-for-Tat with a chance for forgiveness and a defective override.
+
+    The chance for forgiveness happens only if `forgive_flag` is raised
+    (flag discussed below).  If raised and `turn` is greater than `grudge`,
+    then override Tit-for-Tat with Cooperation.  `grudge` is a variable that
+    starts at 0 and increments 20 with each forgiven Defect (a Defect that is
+    overriden through the forgiveness logic).  `forgive_flag` is lower whether
+    logic is overriden or not.
+
+    The variable `defect_padding` increments with each opponent Defect, but
+    resets to zero with each opponent Cooperate (or `forgive_flag` lowering) so
+    that it roughly counts Defects between Cooperates.  Whenever the opponent
+    Cooperates, if `defect_padding` (before reseting) is odd, then we raise
+    `forgive_flag` for next turn.
+
+    Finally a defective override is assessed after forgiveness.  If five or
+    more of the opponent's last twelve actions are Defects, then Defect.  This
+    will overrule a forgiveness, but doesn't undo the lowering of
+    `forgiveness_flag`.  Note that "last twelve actions" doesn't count the most
+    recent action.  Actually the original code updates history after checking
+    for defect override.
+
+    Names:
+
+    - Weiner: [Axelrod1980b]_
+    """
+
+    name = "Weiner"
+    classifier = {
+        'memory_depth': float('inf'),
+        'stochastic': False,
+        'makes_use_of': set(),
+        'long_run_time': False,
+        'inspects_source': False,
+        'manipulates_source': False,
+        'manipulates_state': False
+    }
+
+    def __init__(self):
+        super().__init__()
+        self.forgive_flag = False
+        self.grudge = 0
+        self.defect_padding = 0
+        self.last_twelve = [0] * 12
+        self.lt_index = 0 # Circles around last_twelve
+
+    def try_return(self, to_return):
+        """
+        We put the logic here to check for the defective override.
+        """
+
+        if np.sum(self.last_twelve) >= 5:
+            return D
+        return to_return
+
+    def strategy(self, opponent: Player) -> Action:
+        if len(opponent.history) == 0:
+            return C
+
+        # Update history, lag 1.
+        if len(opponent.history) >= 2:
+            self.last_twelve[self.lt_index] = 0
+            if opponent.history[-2] == D:
+                self.last_twelve[self.lt_index] = 1
+            self.lt_index = (self.lt_index + 1) % 12
+
+        if self.forgive_flag:
+            self.forgive_flag = False
+            self.defect_padding = 0
+            if self.grudge < len(self.history) + 1 and opponent.history[-1] == D:
+                # Then override
+                self.grudge += 20
+                return self.try_return(C)
+            else:
+                return self.try_return(opponent.history[-1])
+        else:
+            # See if forgive_flag should be raised
+            if opponent.history[-1] == D:
+                self.defect_padding += 1
+            else:
+                if self.defect_padding % 2 == 1:
+                    self.forgive_flag = True
+                self.defect_padding = 0
+
+            return self.try_return(opponent.history[-1])

axelrod/tests/strategies/test_axelrod_second.py

@@ -407,6 +407,7 @@ def test_strategy(self):
             (C, D), (C, D), (C, D), (D, D), (D, C)]
         self.versus_test(opponent, expected_actions=actions)
 
+
 class TestKluepfel(TestPlayer):
     name = "Kluepfel"
     player = axelrod.Kluepfel
@@ -453,6 +454,7 @@ def test_strategy(self):
                    (D, C), (D, D)]
         self.versus_test(axelrod.Random(0.5), expected_actions=actions, seed=10)
 
+
 class TestBorufsen(TestPlayer):
     name = "Borufsen"
     player = axelrod.Borufsen
@@ -497,6 +499,7 @@ def test_strategy(self):
                     [(D, C)] * 25
         self.versus_test(axelrod.WinShiftLoseStay(D), expected_actions=actions)
 
+
 class TestCave(TestPlayer):
     name = "Cave"
     player = axelrod.Cave
@@ -546,6 +549,7 @@ def test_strategy(self):
         actions += [(D, D), (D, C)] * 100 # Random finally detected
         self.versus_test(axelrod.Alternator(), expected_actions=actions, seed=2)
 
+
 class TestWmAdams(TestPlayer):
     name = "WmAdams"
     player = axelrod.WmAdams
@@ -586,6 +590,7 @@ def test_strategy(self):
         actions += [(C, C)] * 99
         self.versus_test(changed_man, expected_actions=actions, seed=1)
 
+
 class TestGraaskampKatzen(TestPlayer):
     name = "GraaskampKatzen"
     player = axelrod.GraaskampKatzen
@@ -621,3 +626,71 @@ def test_strategy(self):
         actions += [(D, C)]
         actions += [(D, D), (D, D), (D, C)] * 20 # Defect here on
         self.versus_test(Delayed_GK_Foil, expected_actions=actions)
+
+
+class TestWeiner(TestPlayer):
+    name = "Weiner"
+    player = axelrod.Weiner
+    expected_classifier = {
+        'memory_depth': float('inf'),
+        'stochastic': False,
+        'makes_use_of': set(),
+        'long_run_time': False,
+        'inspects_source': False,
+        'manipulates_source': False,
+        'manipulates_state': False
+    }
+
+    def test_strategy(self):
+        actions = [(C, C)] * 100  # Cooperate forever
+        self.versus_test(axelrod.Cooperator(), expected_actions=actions)
+
+        actions = [(C, C)]
+        actions += [(C, D), (D, C)] # Tit-for-Tat
+        # Opponent's last move was a C with 1 D between
+        actions += [(C, D)] # Tit-for-Tat. Raise forgiveness flag.
+        actions += [(C, C)] # Tit-for-Tat. Use forgiveness flag.
+        # Opponent's last move was a C, but defect_padding counted as 0.
+        actions += [(C, D), (D, C)] # Tit-for-Tat
+        # Opponent's last move was a C with 1 D between
+        actions += [(C, D)] # Tit-for-Tat. Raise forgiveness flag.
+        actions += [(D, C)] # Tit-for-Tat. Try forgiveness flag.
+        # This time grudge=20, so the forgiveness flag doesn't work.
+        actions += [(C, D)] # Tit-for-Tat.
+        # This is the 5th opponent defect, won't be counted for 2 turns
+        actions += [(D, C)] # Tit-for-Tat.
+        actions += [(D, D), (D, C)] * 100 # Defect now on.
+        self.versus_test(axelrod.Alternator(), expected_actions=actions)
+
+        # Build an opponent that will cause a wasted flag.
+        opponent_actions = [C, D, C, C, C, C, D, D]
+        Flag_Waster_1 = axelrod.MockPlayer(actions=opponent_actions)
+        actions = [(C, C), (C, D), (D, C)]
+        actions += [(C, C)] # Raise flag, like in Alternator
+        actions += [(C, C)] # Use flag, but don't change outcome
+        actions += [(C, C)]
+        actions += [(C, D)] # Don't raise flag
+        actions += [(D, D)] # Don't use flag
+        self.versus_test(Flag_Waster_1, expected_actions=actions)
+
+        # Demonstrate that grudge is not incremented on wasted flag.
+        opponent_actions = [C, D, C, C, C, C, D, C, D, C]
+        Flag_Waster_2 = axelrod.MockPlayer(actions=opponent_actions)
+        actions = [(C, C), (C, D), (D, C)]
+        actions += [(C, C)] # Raise flag, like in Alternator
+        actions += [(C, C)] # Use flag, but don't change outcome
+        actions += [(C, C), (C, D), (D, C)]
+        actions += [(C, D)] # Raise flag
+        actions += [(C, C)] # Use flag to change outcome
+        self.versus_test(Flag_Waster_2, expected_actions=actions)
+
+        # Show grudge passing over time
+        opponent_actions = [C, D, C, D, C] + [C] * 11 + [C, D, C, D, C]
+        Time_Passer = axelrod.MockPlayer(actions=opponent_actions)
+        actions = [(C, C), (C, D), (D, C)]
+        actions += [(C, D)] # Raise flag
+        actions += [(C, C)] # Use flag to change outcome
+        actions += [(C, C)] * 11
+        actions += [(C, C), (C, D), (D, C)]
+        actions += [(C, D)] # Raise flag
+        actions += [(C, C)] # Use flag to change outcome

axelrod/tests/strategies/test_meta.py

@@ -335,7 +335,7 @@ def test_strategy(self):
         self.versus_test(opponent=axelrod.Alternator(),
                          expected_actions=actions, seed=0)
 
-        actions = [(C, C), (C, D), (C, C), (C, D), (D, C)]
+        actions = [(C, C), (C, D), (D, C), (C, D), (D, C)]
         self.versus_test(opponent=axelrod.Alternator(),
                          expected_actions=actions, seed=1)
 

docs/reference/overview_of_strategies.rst

@@ -111,7 +111,7 @@ repository.
    "K38R_", "Nelson Weiderman", "Not Implemented"
    "K39R_", "Tom Almy", "Not Implemented"
    "K40R_", "Robert Adams", "Not Implemented"
-   "K41R_", "Herb Weiner", "Not Implemented"
+   "K41R_", "Herb Weiner", ":class:`Weiner <axelrod.strategies.axelrod_second.Weiner>`"
    "K42R_", "Otto Borufsen", ":class:`Borufsen <axelrod.strategies.axelrod_second.Borufsen>`"
    "K43R_", "R D Anderson", "Not Implemented"
    "K44R_", "William Adams", ":class:`WmAdams <axelrod.strategies.axelrod_second.WmAdams>`"