Checkers AI

Built this from scratch after my reinforcement learning class. See if you... read more

• Q-tables serialized with pickle so the trained policy loads instantly, no retraining
• Hand-tuned evaluation heuristics for the alpha-beta agent

RL agent trained across 10,000+ self-play games using reward shaping and epsilon decay. Benchmarked over 5,000 matches to verify policy convergence and tactical improvement over time.