Drofa - UCI chess engine (written in C++11), Author: Alexander Litov.
Drofa started as fork of the Shallow Blue chess engine. My initial intention was to take weak, but stable and working chess engine and try to improve it, learning c++ along the way. The base-goal is to reach strenght of play comparable to the VICE chess engine. The mid-goal - get to ~2500 rating on the CCRL Far goal (probably will never be reached) - get into top 100 of the CCRL.
Chages from Shallow Blue
For now Drofa can be considered as highly advanced fork of the Shallow Blue, with 3 features being completely rewritten by me:
Hashtable
Pawn Hashtable
King Safety Evaluation
Evaluation function in general was refactored with an elo-gain in self-play, although its still reqire much work, because for now tapered eval in fact is used much less than in original Shallow Blue, which is a shame.
Some features were added during the work:
QSearch move generator (althought i think current implementation is lazy and it can be done better)
Null-move pruning
Delta pruning (again, very lazy implementation)
Various bugfixes
But most of the elo-gainers was some code restructuring and optimizations. It allowed Drofa to be ~8x times faster than Shallow Blue.
Drofa 1.0 download
All files (games, tables and engines) - download from page
Comments
Post a Comment