We propagate the reward for both sides

The AI now properly choses the optimal path for the active player

src/policy/backprop/mod.rs

@@ -69,20 +69,16 @@ fn standard_backprop<S: GameState>(
     let mut current_id: usize = node_id;
     loop {
         let node = arena.get_node_mut(current_id);
-        let player = node.state.get_current_player().clone();
-        match rewards.get(&player) {
-            Some(reward) => {
         node.increment_visits();
-                node.record_player_reward(player, *reward);
+        for (player, reward) in rewards.iter() {
+            node.record_player_reward(player.clone(), *reward);
+        }
         if let Some(parent_id) = node.parent {
             current_id = parent_id;
         } else {
             break;
         }
     }
-            None => (),
-        }
-    }
 }
 
 fn weighted_backprop<S: GameState>(
@@ -94,21 +90,16 @@ fn weighted_backprop<S: GameState>(
     let mut current_id: usize = node_id;
     loop {
         let node = arena.get_node_mut(current_id);
-        let player = node.state.get_current_player().clone();
         let weight = weight_for_depth(depth_factor, node.depth);
-        match rewards.get(&player) {
+        for (player, reward) in rewards.iter() {
-            Some(reward) => {
+            node.record_player_reward(player.clone(), (*reward) * weight);
-                node.increment_visits();
+        }
-                node.record_player_reward(player, (*reward) * weight);
         if let Some(parent_id) = node.parent {
             current_id = parent_id;
         } else {
             break;
         }
     }
-            None => (),
-        }
-    }
 }
 
 /// Calculate the weight based on the current depth