Walleye is a UCI-compatible chess engine written using the classic alpha-beta style AI.
Challenge me on Lichess! https://lichess.org/@/Walleye_Bot
The engine should also work in any chess program that supports UCI, at this time however it has only been tested with Cute Chess.
By default, the engine launches in UCI mode and expects to be loaded into a chess GUI.
However, you can run some commands from the terminal, such as -P
to watch the engine play against itself or -T
to benchmark move generation and evaluation.
# helpful when profiling, will accept a FEN string
./walleye -T --depth=5
# start a game from a FEN string and have the engine play against itself
./walleye --fen="r3k2r/p1ppqpb1/bn2pnp1/3PN3/1p2P3/2N2Q1p/PPPBBPPP/R3K2R w KQkq - 0 1" -P
Use ./walleye --help
for a complete list of commands.
It is strongly recommended you compile the engine with --release
for the best performance.
cargo build --release
In the cargo/config file the target CPU is set to native
. During testing I saw modest performance improvements using this setting. If however you plan to compile Walleye and use the binary on multiple systems you should delete these lines before compiling.
- Square Centric 12x12 Array
Extra board squares are sentinel squares to make boundary checking easier.
- Alpha-Beta Pruning
- Iterative Deepening
- Capture/Check Extension
- Killer Moves
- MVV-LVA
- PV Search
- Piece Square Table
- Terminal based games with unicode chess boards
- Robust logging
Walleye comes with a suite of unit tests and perft tests. It has been verified on a variety of positions to around depth 5.
# run perft tests
cargo test perft
# run all tests
cargo test
Walleye was ranked by CCRL, you can find its rating between versions here
Some resources and tools I found helpful when creating this engine.
If you find an issue with the engine please include the walleye_{PID}.log
file along with the report, you can enable this by setting DebugLogLevel
to Info
in the UCI options.
Walleye is under the MIT license.