Empirical Comparison of Negamax with Alpha-Beta Pruning and Monte Carlo Tree Search for Mobile Chess AI on Flutter

Abstract

This paper implements and compares two chess artificial intelligence algorithms, Negamax with Alpha-Beta Pruning and Monte Carlo Tree Search (MCTS), within a cross-platform Flutter-based mobile chess application. The system employs advanced heuristic evaluation techniques incorporating material balance, positional analysis, king safety, pawn structure, and piece-square tables. Negamax utilises deterministic search with iterative deepening, Alpha-Beta pruning, quiescence search, and transposition tables, while MCTS performs probabilistic exploration using Upper Confidence Bound for Trees (UCT), progressive bias, and heuristic guided simulations. To evaluate performance under mobile hardware constraints, a benchmark framework executed 40 automated AI vs AI games across multiple predefined opening systems under identical runtime conditions on a CMF Phone 1 device using a fixed 5 second computation budget per move. Experimental results indicate that Negamax achieved substantially lower average move response times (0.94 ± 0.21 s) compared to MCTS (2.81 ± 0.52 s), while also reaching deeper average search depths and higher win rates. The findings suggest that Alpha-Beta-based search remains highly effective for resource-constrained mobile chess applications requiring efficient real-time decision-making, whereas MCTS provides broader probabilistic exploration at higher computational cost.