my new omega improvement

code/nekiwo/ShirtlessMangos.py

@@ -0,0 +1,104 @@
+import numpy as np
+
+# Licensed under MIT in https://github.com/Prisoners-Dilemma-Enjoyers/PrisonersDilemmaTournament
+# Member(owner) of the Unofficial Prisoner's Dilemma Discord server  https://discord.gg/6UBBKxyk
+# Big thanks to everyone who participated in our server, it was a very fun week
+
+# My shamelessly stolen version of OTFT
+# from https://arxiv.org/ftp/cs/papers/0609/0609017.pdf
+
+# Includes small improvements
+# (It's really hard to improve this strat without ruining it)
+
+def DetectJoss(history, window):
+    AlignMe1 = history[0, -window:]
+    AlignEnemy1 = history[1, -window - 1:-1]
+
+    AlignMe2 = history[0, -window - 1:-1]
+    AlignEnemy2 = history[1, -window:]
+
+    if np.count_nonzero(np.absolute(AlignMe1 - AlignEnemy1) == 1) < np.floor(window / 3) or \
+        np.count_nonzero(np.absolute(AlignMe2 - AlignEnemy2) == 1) < np.floor(window / 3):
+        # Oof, it's TFT/Joss
+        return True
+
+    return False
+
+
+def tft(history):
+    choice = 1
+
+    if history.shape[1] > 0 and history[1, -1] == 0:
+        choice = 0
+
+    return choice
+
+
+def strategy(history, memory):
+
+    # Standard OTFT stuff
+
+    RandomThreshold = 5
+    DeadlockThreshold = 1
+
+    randomness = 0
+    deadlock = 0
+
+    if memory == None:
+        memory = [randomness, deadlock]
+    else:
+        randomness = memory[0]
+        deadlock = memory[1]
+
+    # I tried many, many alternative but regular TFT works the best anyway
+    choice = tft(history)
+
+    # One of the improvemnts which checks in case the first move was defect
+    # Random has a 50% chance of doing it, this is very big red flag
+    if history.shape[1] == 1 and history[1, -1] == 0:
+        randomness += 2
+
+    if history.shape[1] > 1:
+        if deadlock >= DeadlockThreshold:
+            choice = 1
+
+            if deadlock == DeadlockThreshold:
+                deadlock += 1
+            else:
+                deadlock = 0
+
+        else:
+            if history[1, -1] == history[1, -2] == 1:
+                # Adjusting those values improved random detection by 0.003 - 0.010 points on average
+                # Makes it easier to find actual random from Joss-likes
+                randomness -= 1.25
+
+            if history[1, -1] != history[1, -2]:
+                randomness += 1
+
+            if history[0, -1] != history[1, -1]:
+                randomness += 1.25
+
+            # Checks if patterns are similar to Joss-likes
+            # Random has a chance of doing this accidentally so I'm not giving it a lot of points
+            if history.shape[1] > 8 and DetectJoss(history, 8):
+                randomness -= 0.25
+
+
+            if randomness >= RandomThreshold:
+                choice = 0
+            else:
+
+                if history[1, -1] != history[1, -2]:
+                    deadlock += 1
+                else:
+                    deadlock = 0
+
+
+
+
+    memory = [randomness, deadlock]
+
+    return choice, memory
+
+    # Thanks Cary!!