Here you can find some code[1] about implementation of chess engine via reinforcement learning of AlphaZero methods.
Supervised learning could be done over PGN games files of professional chess players (with high elo rating that means). This data was taken from the chess database called FICS.
Here some nice results (over 3000 games of SL, elo >= 2000). I'm (white) playing against the program (black):
To start SL put PGN files in data/play_data dir and run the command:
python src/chess_zero/run.py sl
For reinforcement learning use: self, opt and eval commands:
| Command | Description |
|---|---|
self |
is Self-Play to generate training data by self-play using BestModel |
opt |
is Trainer to train model and generate next-generation models |
eval |
is Evaluator to evaluate whether the next-generation model is better than BestModel. If better, replace BestModel |
All in all it's just about generating PGN data which contains self-played games by the chess engine.