Exploring Tactics in Fighter Duel Lite using Monte Carlo Tree Search

This report explores tactics and game balance in Fighter Duel Lite, a two-player aerial combat simulation, using Monte Carlo Tree Search (MCTS). Our analysis revealed that move order and pilot expertise significantly influenced outcomes, with first-move advantage and specific opening maneuvers correlating with higher success rates. The analysis was done by developing an MCTS-based agent that employs some Fighter Duel Lite specific heuristics to guide the search, combined with a reward function to evaluate game states. The MCTS-agent achieved a win rate exceeding 90\% against a heuristically filtered random opponent, maintaining strong performance across varying starting positions and pilot skill configurations.