This is my Chess RL Website - utilizes ideas from Bertsekas et al., "Superior Computer Chess with Model Predictive Control, Reinforcement Learning, and Rollout" (arXiv 2409.06477)
First page allows you to play chess with someone else on the same device or against a bot
Second page allows you to upload a .pth file to play against a chess bot either playing black or white
Get your local path to stockfish and change the stockfish path in game.py and rllearn.py
Install python version >= 13.3 and run pip install -r requirements.txt
Run s.ps1 to launch the website
- Invoke penalty for repetition of moves
- Game rules begin to be ignored after training for a high number of iterations
- Pretrain on existing dataset of chess positions before rl happens
- TODO