New engine releases & news 2021

[Brunetti]

New engine: Mantissa https://github.com/jtheardw/mantissa

The author claims 2000/2100 Lichess Elo strength; my 1'+1" preliminary test shows around 2700.

Alex

[Guenther]

Mantissa 2.1.3 released
https://github.com/jtheardw/mantissa

Stockfish 21092119 (dev) released
https://abrok.eu/stockfish/

Blunders 0.1.0 (New) released (haven't compiled it yet)
So now we have a Blunder and a Blunders chess engine :)
https://github.com/paulolemus/blunders

author: Paulo Lemus
country: USA

[Gabor Szots]

Blunders 0.1.0 (New) released (haven't compiled it yet)
So now we have a Blunder and a Blunders chess engine
https://github.com/paulolemus/blunders

author: Paulo Lemus
country: USA

I haven't found indication of strength on its home page but based upon a very quick test it seems it's about 1600.

[mvanthoor]

I haven't found indication of strength on its home page but based upon a very quick test it seems it's about 1600.

That's not a high rating for an engine that's looking quite complicated code-wise, IMHO. However, given the feature set, it seems OK. (It does some things more in "The Rust Way" than my own engine, at least at the moment.)

[Brunetti]

New engine: Pawn https://github.com/ruicoelhopedro/pawn.

"A UCI alpha-beta chess engine, largely inspired by Stockfish." It's far simpler than SF, I wouldn't call it a derivative; anyway source code is there.
I made a Windows binary, available here: https://e4e6.com/download/. 1'+1" Elo ~ 2330.


Alex

[msarchet]

New Engine - Panzer

Github Repo Here https://github.com/msarchet/panzer

Bitboard based C++ engine, currently compiles on linux, working on completing bits for Windows builds. Just got it to actually playing chess with a basic AlphaBetaMin/Max evaluation. Currently using the PieceSquare Evaluations from the CPW-Engine. I have quite a few todos left. But I'm pretty excited that I got it working. Up next is finishing my FEN string interpretation and adding Quiescence and Transposition Tables. On my Ryzen 9 5950x my multi-threaded perft evaluation gets around 150M-250M NPS depending on the number of nodes searched. I don't remember the single core numbers off the top of my head.

I previously wrote two move generators one in Python and another in C++ (inside of Panzer's src as old version), both using the 0x88 representations, they were error prone and slow in various cases which made debugging beyond perft(5) painfully slow.

[Graham Banks]

Shallow 14

http://dimock.anabar.ru/samples/shallow_v4_x64.7z

[Brunetti]

Shallow 14

Shallow 4, 1'+1" Elo ~2820.

Alex

[Gabor Szots]

Shallow 14

Shallow 4, 1'+1" Elo ~2820.

Alex

How do you know the version number is 4? At the site there is only version 3 and the engine does not display any version number.

[Graham Banks]

Shallow 4

Shallow 4, 1'+1" Elo ~2820.

Alex

How do you know the version number is 4? At the site there is only version 3 and the engine does not display any version number.

Sent: Sat Sep 25, 2021 8:18 am
From: dimock
Recipient: Graham Banks

Graham good day,
I've published new version of Shallow
it is available here:http://dimock.anabar.ru/samples/shallow_v4_x64.7z
I've named it as "version 4"
There are some bugs fixed and also performance is optimized.
___
Thank you,
Dmitry