Since you’re here...

We hope you will consider supporting us today. We need your support to continue to exist, because good entries are more and more work time. Every reader contribution, however big or small, is so valuable. Support "Chess Engines Diary" even a small amount– and it only takes a minute. Thank you.
============================== My email: jotes@go2.pl



Tarnished 3.0 (Warrior) - new version chess engine

on


Tarnished - UCI chess engine
Author: Anik Patel (Bobingstern)

What has changed in the new version  Tarnished 3.0 (Warrior)?

This version comes with a bigger and stronger NNUE, several search patches, and fixed SMP!

Elo   | 207.46 +- 13.63 (95%)
Conf  | 40.0+0.40s Threads=1 Hash=128MB
Games | N: 1000 W: 586 L: 51 D: 363
Penta | [0, 2, 75, 309, 114]
https://chess.n9x.co/test/2367/

NNUE
Now using a (768->1024)x2->1x8 architecture trained on ~1.7b positions
Networks now stored at tarnished-nets
Search
Fixed SMP Scaling
Factorized LMR formulation (see code)
Conthist bugfixes
Static evaluation correction histories
Razoring
TT improvements
Staged move picker
FRC support

Tarnished 2.1 JA - tactical test


Tarnished 2.1 JA was tested against the position shown in the diagram, where the non-obvious move 1.Rg4+ wins.


Tarnished 2.1JA compiled by Jim Ablett and it solved the problem without any problems. Interestingly, the "original" 2.1 compiled failed and after 10 minutes of analysis it proposed 1.Bd5.



Comments

Post a Comment