remove previous model action reset on model change, accidentally made…

… it delayed, might need to revisit on another selector run

CoyoteParser.py

@@ -705,9 +705,11 @@ def parse_actions(self, actions: Any, state: GameState) -> np.ndarray:
                 parsed_actions.append(action)
                 continue
             action = int(action[0])  # change ndarray [0.] to 0
-            # action changed, reset previous actions for stability of model swaps to keep in distribution
-            if action != self.prev_model_actions[i]:
-                self.prev_actions[i] = np.asarray([0] * 8)
+            # # action changed, reset previous actions for stability of model swaps to keep in distribution
+            # ****** THIS PROBABLY DIDN'T WORK AND MADE IT PLAY BADLY ********
+            # At least from testing in rlbo this appears to be the case due to the delay of one tick
+            # if action != self.prev_model_actions[i]:
+            #     self.prev_actions[i] = np.asarray([0] * 8)
             # action = 22  # TODO testing remove this
             # zero_boost = bool(action >= self.get_model_action_size())  # boost action 1 means no boost usage
             # if action >= self.get_model_action_size():