update

src/mcts.jl

@@ -452,13 +452,14 @@ function runMCTS(
     if node.isterminal
       # MCTS arrive at the leaf node that is also a terminal state,
       # do nothing then go directly to backpropagation
+      backpropagate(leafNode, node.reward)
     else
       expand(a, node, decisionMaker, progressValueEstimator, isterminal, n=n)
       leafNode = UCTselect(node, w)
       simTrajectoryReward = simulate(a, leafNode, decisionMaker, progressValueEstimator,
                                     isterminal, maxDepth, n=n)
+      backpropagate(leafNode, simTrajectoryReward)
     end
-    backpropagate(leafNode, simTrajectoryReward)
   end
 
   best_child_state = argmax([child.total_reward / child.visits for child in values(root.children)])