Show HN: Tiny chess model, 392 KiB, ~1523 elo

Category: ai-ml

Tags: chess, neural-network, game-ai, numpy, alpha-beta

Score: 5.5/10 (Innovation: 5, Technical: 6, Documentation: 7, Utility: 4)

chesslm is a tiny neural network (100k parameters, 392 KiB) that evaluates chess positions, combined with alpha-beta search and a Tkinter/web interface. It trains on Stockfish labels and achieves a nominal Elo of ~1523 against limited Stockfish, showcasing efficient inference with NumPy. The project is interesting for its compact model and educational value in chess AI.

Target audience: AI enthusiasts, chess programmers, hobbyists

Repository: https://github.com/skorotkiewicz/chesslm · Python · 2 stars

View on Hacker News