In this project we present a deep learning solution based on AlphaZero for playing "Bughouse" , a chess variant also known as "tandem chess". We use supervised learning in the form of a Deep Convolutional Recurrent Neural Network (DCRNN). The main point of our approach is that we use an asynchronous Monte Carlo tree search algorithm to support the neural network. In addition, a Bughouse server environment is implemented on which real games can be played. Our resulting bughouse engine is then tested and evaluated against other engines in the bughouse environment.
Attached is the bughouse server of Moritz Willig. It is required to install NodeJS and npm. Afterthat, please follow the instructions in the README.md. The server will be optimized. Git repository:: https://github.com/MoritzWillig/tinyChessServer.git