I’m currently implementing a Tic-Tac-Toe (TTT) engine using Monte Carlo Tree Search (MCTS). However, I’ve encountered an issue with the function responsible for calculating the best moves. Initially, I chose the child node with the highest evaluation score, but this approach failed to block the opponent’s winning moves. On the other hand, when I used the highest Upper Confidence Bound (UCB) score with c = 0 , the AI performed perfectly. Can someone explain why the UCB score method works better in this scenario, or is it just a bug in my implementation?

I am working on a Tic-Tac-Toe engine using the Monte Carlo Tree Search (MCTS) algorithm. However, I’ve encountered a bug where the AI sometimes fails to block the opponent’s winning moves, resulting in losses. Additionally, the program occasionally crashes due to a segmentation fault.