New engine releases & news H1 2024

[Frank Quisinsky]

Hi Jonathan,

OK, but the fact is that I have had fun with your Winter a long time now. My wish that Winter is in the group of the best available was too high I think. But not important, you really generate a fun engine for self playing chess as well, just great. And when I see what Winter 2.0 can do in the still running 66+6 tournament, just fantastic.

So the strength goes up with more and more time (in comparison to other strong engines). These are very interesting chess programs, i.e. if I can see that with more time the engine is able to find the better move.

I look forward to your development.
Sure, you are right in that what you do ...
I like Winter in both cases ... the engine and season.



Best
Frank

[eduherminio]

Lynx 1.3.0
+101.3 (+/- 6.6) vs 1.2.0
Estimated 2651 CCRL blitz

[Gabor Szots]

Hi Jonathan,

I stopped the test after around 300 games ...
Not sure, but it seems the version is not stronger as 3200 Elo (my own rules).

I am using on Intel i9-10900k the
Winter_3.0_New.exe

Maybe I am wrong, but this compile is the fastest (I test the compiles before I start).

Best
Frank

Based on my ongoing tournament Winter 3.0 is significantly sronger than 2.0, maybe not by as much as claimed but not far. So I don't think there is reason for panic.
I am going to submit my games for this week's CCRL update, then we shall see.

[lithander]

Leorik 3.0
This is the first version that uses NNUE evaluation.I expect ~3270 CCRL Elo!

[Dariusz]

Leorik 3.0
This is the first version that uses NNUE evaluation.I expect ~3270 CCRL Elo!

Thanks for Leorik 3.0

Unfortunately, it doesn't work on my Macs. Instead, versions 2.3 and 2.4 work. And versions 2.5 and 3.0 do not work.
In a private message I will send you the messages from command line (terminal) after opening Leorik 3.0.
If there is a chance to check it, I will be grateful. Thanks!

[Dariusz]

Leorik 3.0
This is the first version that uses NNUE evaluation.I expect ~3270 CCRL Elo!

Thanks for Leorik 3.0

Unfortunately, it doesn't work on my Macs. Instead, versions 2.3 and 2.4 work. And versions 2.5 and 3.0 do not work.
In a private message I will send you the messages from command line (terminal) after opening Leorik 3.0.
If there is a chance to check it, I will be grateful. Thanks!

Thanks to the commitment of the author Leorik engine and Alex (the author of Altair), version 3.0.1 working on Mac has appeared on GitHub.
Gentlemen, many thanks

https://github.com/lithander/Leorik/releases/tag/3.0

[akanalytics]

Odonata 0.8.0 released.

Small increase in Elo. +30 maybe.

Available for download here: https://github.com/akanalytics/odonata/ ... /tag/0.8.0 (Linux, Windows and Mac)

Possibly one more hand-crafted-evaluation release after this one, before I migrate to NNUE and give up tweaking evaluation functions ...

Andy

[Frank Quisinsky]

Hi there,

1. PlentyChess 0.3.0 NN
by Patrick Leonhardt, GER
https://github.com/Yoshie2000/PlentyChess

is very strong
My test with 4+2 is still running
https://www.amateurschach.de/fling/test-44.html

2. Lizard 10.2 NN
by Liam McGuire, USA
https://github.com/liamt19/Lizard

Best
Frank

[mar]

Cheng 4.46 is out:

https://github.com/kmar/cheng4/releases/tag/4.46

- new NN, 800M selfplay positions, 10k nodes, new topo: 736-224*2-1
- augmented net eval with old HCE EG scalers/recognizers
- bench 9213631

scores +100 elo against 4.45 at 60+0.6" (5k games)

expected 50-100 against other programs

this release concludes the switch to NN eval, don't expect any new versions anytime soon

thanks to Lars for testing

have fun

[Dariusz]

Caissa-1.17
https://github.com/Witek902/Caissa/releases/tag/1.17

Compilations for Linux, Mac and Windows can be downloaded from my site (Files section).