---
license: mit
tags:
  - chess
  - pytorch
  - reinforcement-learning
  - behavioral-cloning
---

# Chess Expert

A deep-learning chess model that learns from **grandmaster games** (behavioral
cloning, à la Maia). A residual CNN with a **policy head** (which move) and a
**value head** (how good the position is), plus an optional shallow look-ahead
search so it avoids hanging pieces.

- Code: https://github.com/AhPro7/chess-expert
- `chess_expert.pt` — load with the repo's `src.play.ChessEngine`.
- `logs/` — TensorBoard training curves (`tensorboard --logdir logs`).

Trained on CPU-friendly inference; see the repo README for how to play (terminal
or a clickable Colab board).