UCI compliant chess engine

Play online

Challenge me on Lichess: https://lichess.org/@/FataliiBot

Download

Binaries for Linux and Windows are available on the release page.

Building from source

To build the engine from source, you need a Rust compiler. Clone the repo:

# Using SSH
git clone git@github.com:FitzOReilly/fatalii.git
# Alternatively, using HTTPS
git clone https://github.com/FitzOReilly/fatalii.git

and build the engine:

cd fatalii
cargo build --profile release-lto --package fatalii --bin fatalii

The binary will be in target/release-lto/fatalii.

Usage

Fatalii supports the UCI protocol (universal chess interface), so it can be used with a UCI compatible GUI. Some choices are Cute Chess, Arena or Lucas Chess.

UCI options

• Hash
  The size of the hash table in MB
• Move Overhead
  Subtract this value from the movetime to compensate for network delays or GUI overheads
• UCI_Chess960
  Enable Chess960 if this is set to true

Supported variants

Fatalii supports both standard chess and Chess960 (a.k.a. Fischer Random Chess).

Features

• Bitboards using file rank mapping
• Move generator using kindergarten bitboards for sliding pieces
• Evaluation
  • Piece square tables (symmetrical)
  • Pawn structure: passed, isolated, backward and doubled pawns
  • Mobility
  • Bishop pair
  • Tempo
  • King tropism
  • Tapered evaluation for all parameters
  • Tuned with training positions from the Zurichess dataset quiet-labeled.v7
• Search
  • Iterative deepening
  • Principal variation search
  • Aspiration windows
  • Quiescence search
  • Move ordering
    • Root move ordering based on the previous iteration and on subtree size
    • Principal variation move
    • Hash move from the transposition table (if there are multiple TT entries for the position, all of them will be used)
    • Queen promotions
    • Winning and equal captures (estimated by static exchange evaluation (SEE))
    • Killer heuristic
    • Countermove heuristic
    • History heuristic
    • Losing captures (negative SEE)
    • Underpromotions last
  • Pruning
    • Fail-soft alpha-beta pruning
    • Null move pruning
    • Futility pruning
    • Reverse futility pruning
    • Late move reductions
    • Late move pruning
    • SEE pruning
    • Delta pruning in quiescence search
  • Check extensions
• Transposition table
  • Zobrist hashing
  • 4 entries per bucket
  • Replacement scheme based on entry age and depth
• Draw detection
  • 3-fold repetition
  • 50 move rule
  • Insufficient material

Thanks to

• The Chess Programming Wiki. It has been extremely helpful during development. A lot of ideas have been taken from it and I've also learned a lot from it.
• Other open source chess engines.
• Lichess for being an awesome chess site and for letting bots play there. And https://github.com/lichess-bot-devs/lichess-bot for making it easy to create such a bot.
• The folks at CCRL for rating the engine.
• Cute Chess. The CLI was extensively used for self-play testing.

More helpful resources

• The UCI specification: https://www.shredderchess.com/download/div/uci.zip
• http://talkchess.com/forum3/index.php
• https://andrewra.dev/2019/08/05/testing-in-rust-writing-to-stdout/
• Evaluation tuning
  • Andrew Grant's paper: https://github.com/AndyGrant/Ethereal/blob/master/Tuning.pdf
  • Good explanation of gradient descent optimization algorithms: https://www.ruder.io/optimizing-gradient-descent/
  • The Zurichess datasets: https://bitbucket.org/zurichess/tuner/downloads/
• Tuning search parameters
  • https://chess-tuning-tools.readthedocs.io/en/latest/