Eleanor uses a fully neural network-based evaluation function, entirely trained on self-play data from an initially simple HCE using the bullet trainer.

Architecture: (768hm -> 1536)x2 -> 1x8
A 1536-hidden-layer perspective net with 8 output buckets and horizontal mirroring.

Prerequisites:

• clang++ compiler
• Access to make command

1. Clone the repository
2. Run make in the root directory
3. Enjoy your executable 🎉

You can interact with the engine via these commands:

• position — Set the board state using a FEN:

  • position fen [FEN] or
  • position startpos

• ucinewgame — Resets the position and game state

• quit — Terminates the program

• go — Begins search and returns the best move

  • Parameters (all in milliseconds):
    • wtime, btime — Remaining time for white/black
    • winc, binc — Increment time for white/black

• stop — Stops the current search

• bench — Runs a speed benchmark on a set of positions

The engine can be used in any GUI with UCI support.

Testing site

• Stockfish Discord Community — The best possible source for an engine dev.
• Chess Programming Wiki — Contains some good articles. Used it mostly when started the project.
• Maksim Korzh's Chess Engine in C (YouTube Series) — it was a good series helping with some fundamental stuff early on.

The project is a work in progress and actively developed. Contributions are highly appreciated.