This is an engine + analysis interface for duck chess that runs entirely in the browser, with the networks trained with tabula rasa self-play, like AlphaZero.
Live version: https://peter.website/duck-chess
The engine uses TensorFlow.js to perform inference, and a WASM backend written in Rust for MCTS.
Currently everything is a big mess, but I can clean up the code at some point if folks want to modify it. But I'm going to leave it as a massive mess until someone expresses interest.
Feel free to email me at [email protected]
All contents of this repo are public domain, except for the chess piece images, which are CC BY-SA 3 created by Cburnett (except except for the duck, which I drew). I additionally make it all available under CC0.