Pandinha-engine is a chess engine that use MinMax Algorithm and Alpha–beta pruning to search moves.
The engine follows the UCI Protocol, so it works with any GUI that accepts custom engines especified with the UCI Protocol.
The following are some design/implementation decisions made:
The search can be done sequential or can be parallel, in the parallel approach, is created a pool of processes.
To generate moves is created a function that return a python generator, where the moves are computed on demand, this is useful because when some nodes are cutted-off the remaining moves aren't calculated.
To represent the board is used a square-centric approach, where the board is an array with 64 integer values, where each integer represent a colored piece or an empty square.
The evaluation function is purely material-based, which use the standard weights:
- Pawn: 1
- Bishop: 3
- Knight: 3
- Rook: 5
- Queen: 9
To use the engine you need to install an GUI's that supports the UCI Protocol to communicate with the engine.
After installing a GUI, access the src/settings.py file and change the values to your preference. Also, you need figure out how to add a custom engine in your GUI, by doing that add the engine main file path: main.py.
The recommended GUI is jerry because the engines were tested using this GUI. However, soon all the UCI commands will be implemented to allow the engine to work with all UCI GUI's.
The project was developed as a learning proposal, so some features to turn this software useful are needed to be implemented. The following list some of them:
- Add support to castle movement.
- Add support to En Passant movement.
- Currently is supported pawn promotion to queen and knight. Is needed to add support to rook and bishop promotion.
- The depth of the search is a fixed constant. however, the value must be set dynamically by communicating with the GUI interface, this is useful for supporting diferent time formats: classical, blitz, bullet and so on.
- Implement all commands/responses used in the UCI protocol.
- NORVIG, P.; RUSSELL, S. Inteligência Artificial. Elsevier, 3a ed., 2013.
- Board representation - Chess Programming
- Move generation - Chess Programming
- Python Generators
As shown in TODO section, there is a lot of work to complete, I am open to accepting pull requests implementing some of these features.
Released under the MIT License. See LICENSE.md for details.