Alexander - ShashChess without nnue.
Rating Chess Engines Diary CEDR=3558
Unfortunately, the Stockfish community has decided to shelve the classic evaluation function, despite the fact that the engine is far beyond any human being.
In addition, the classical evaluation function is the only one well understood by a human player.
In fact, even an IM/GM struggles to understand a neural network game.
Conversely, even for a GM, it makes no difference to have a 3100- or 3500-elo engine....
We believe that the future is precisely trying to decrypt the neural network to rewrite classic strategy books in a modern way.
The first step, in our opinion, is precisely to try to compare the two evaluation functions.
In addition, the classical evaluation function allows the implementation of a true handicap mode, based on a real thinking system and the concept of avatars to simulate not only the playing strength, but also the style of a player.
Alexander should be the only reference engine for an OTB player below the IM title.
An IM or GM can occasionally consult ShashChess to try to penetrate the secrets of NNUE.
This is also why we decided to separate the two worlds.
"Scientific research, if it does not improve people's lives, is a sterile fruit of hypertrophied egos"
In addition, the classical evaluation function is the only one well understood by a human player.
In fact, even an IM/GM struggles to understand a neural network game.
Conversely, even for a GM, it makes no difference to have a 3100- or 3500-elo engine....
We believe that the future is precisely trying to decrypt the neural network to rewrite classic strategy books in a modern way.
The first step, in our opinion, is precisely to try to compare the two evaluation functions.
In addition, the classical evaluation function allows the implementation of a true handicap mode, based on a real thinking system and the concept of avatars to simulate not only the playing strength, but also the style of a player.
Alexander should be the only reference engine for an OTB player below the IM title.
An IM or GM can occasionally consult ShashChess to try to penetrate the secrets of NNUE.
This is also why we decided to separate the two worlds.
"Scientific research, if it does not improve people's lives, is a sterile fruit of hypertrophied egos"
github:https://github.com/amchess/Alexander/releases/tag/4.1
Alexander 7.0 what's new?
Aligned with Stockfish patch
Oct 14, 2025
Use huge pages for worker data
A professional trace
Alexander 6.0 vs other engines:
| Lambergar 1.3 | 5.5/22 | -11 | 22 Games |
| YuliGM Pro 20 | 3/20 | -14 | 20 Games |
| Caissa 1.23 | 1.5/20 | -17 | 20 Games |
| ShashChess 39.1 | 1/20 | -18 | 20 Games |
| Claw 3.0-M | 0.5/16 | -15 | 16 Games |
| HypnoS ++ 1.03 | 0.5/16 | -15 | 16 Games |
| Stockfish dev-20250906 | 0.5/16 | -15 | 16 Games |
| PlentyChess 6.0.18 | 0.5/16 | -15 | 16 Games |
| Pawnocchio 1.8.1 | 0/16 | -16 | 16 Games |
| SF-POLY 200825 | 0/16 | -16 | 16 Games |
| Brainlearn 31 | 1/14 | -12 | 14 Games |
| DarkZoryn 2.0 | 2/12 | -8 | 12 Games |
| Spirit XXL | 2/12 | -8 | 12 Games |
| OmegaCor 280825 | 1.5/12 | -9 | 12 Games |
| Revolution 2.40 130925 | 1.5/12 | -9 | 12 Games |
| Zangdar 5 | 1/10 | -8 | 10 Games |
| Alexandria 8.1.2 | 0.5/10 | -9 | 10 Games |
| Wordfish 2.50 dev-210925 | 0.5/10 | -9 | 10 Games |
| Halogen 15.4.2 | 0.5/10 | -9 | 10 Games |
| Revolution 2.70 dev-210925 | 0.5/10 | -9 | 10 Games |
| Raid 4.0 | 0.5/10 | -9 | 10 Games |
Comments
Post a Comment