ClashRoyaleAi: an open-source, deterministic Clash Royale simulator for RL, with recurrent PPO, lookahead search and expert iteration [P]

ClashRoyaleAi is an open-source, deterministic Clash Royale simulator designed for reinforcement learning (RL), featuring recurrent PPO, lookahead search, and expert iteration. The system plans by simulating matches 10 seconds ahead to score candidate plays. A 1-ply lookahead significantly improved the win rate against a heuristic bot from 0.625 to 0.944, though distilling this back into the network only yielded a +0.045 improvement. The project was built with a friend, Ambash, and utilized AI coding tools.