Chess-AI

Play it on lichess.com

A fixed-depth version of the engine by forking the lichess-bot API. Runs in Oracle Cloud Infrastructure.

Download the macOS Preview

Research from chessprogramming.org.

Description

Third version of a personal project to create a chess AI from scratch in Swift. Without evaluations, the engine can reaches ~2.5m nodes per second (NPS) on the start position without bulk counting.

Details

The engine currently uses:

• Hybrid board representation – Bitboards and Piece List
• Move generation dependent on bitwise operations and bitmasks
• Negamax search algorithm with Alpha-Beta pruning
• Move ordering to search moves with higher likelihood of benefits
• Simple evaluation function counting material and piece-square tables
• FEN and PGN parser and serializer for importing and exporting games
• Small file containing chess openings in PGN format

Future Goals Include:

• Complete Zobrist hashing for transposition table
• Multithreading
• Complete iterative deepening
• Neural network evaluation option
• UCI protocol for CLI
• Ditch this project completely and start all over again in C++ because C++ is faster