I think u should focus more in evaluation by adding heuristics or neural net or optimized minimax for example negamax with alpha beta because mcts is prone to tactical over site especially in chess. You can watch YouTube video over here for more explanation in this link https://youtu.be/Ne40a5LkK6A?si=vxEv9iJ-RYYcQ5_j at 28:11 for mcts probably without simulation phase (I think it uses heuristics evaluation instead of random simulation based evaluation)