Made rewards make more sense and removed extra punishment for death

entities/player.py

@@ -161,7 +161,7 @@ class Player(pygame.sprite.Sprite):
             self.old_state_features = self.state_features
 
             self.reward = self.stats.exp\
-                + self.stats.health/self.stats.stats['health'] - 1
+                + self.stats.health/self.stats.stats['health']
                 # - nearest_dist/np.sqrt(np.sum(self.map_edge))
 
         self.state_features = [

level.py

@@ -293,7 +293,6 @@ class Level:
         for player in self.player_sprites:
             if player.is_dead():
                 print(f"\nPlayer {player.player_id} is dead\n")
-                player.stats.exp = -10
                 self.dead_players[player.player_id] = player.is_dead()
 
         self.done = True if (self.dead_players.all() == 1

main.py

@@ -116,7 +116,7 @@ def main():
 
                     episode_reward[player.player_id][step] = player.reward
 
-                    if not no_training and ((step % horizon == 0 and step != 0) or player.is_dead()):
+                    if not no_training and ((step % horizon == 0 and step != 0):
 
                         player.agent.learn()