New engine releases & news H1 2023

[bastiball]

Winter v2.0 is released!

Please be patient with regards to Windows binaries, they should be out this weekend.

Search is mostly unchanged. Most of the improvement is from more self play training data (all DFRC without standard starting position), improved architecture and tablebase rescoring. In older data I did not do tablebase rescoring as it for some reason surprisingly resulted in a weaker net.

DFRC self play bullet testing between Winter v2.0 and Winter v1.0:

ELO | 100.40 +- 7.99 (95%)
CONF | 40.0+0.40s Threads=1 Hash=64MB
GAMES | N: 5000 W: 2344 L: 938 D: 1718

I don't have non-DFRC results, as I am out of principle focusing on DFRC for now That being said, I would expect the regular chess improvement to be less than this.

https://github.com/rosenthj/Winter/releases/tag/v2.0

Looking forward to test in FRC

[Gabor Szots]

Viridithas 7.0.0

https://github.com/cosmobobak/viriditha ... tag/v7.0.0

[Damir]

Wrong link. It says the page does not exist...

[Gabor Szots]

Wrong link. It says the page does not exist...

Copy the whole text. For some reason not the whole line got underlined. I correct it below:

https://github.com/cosmobobak/viriditha ... tag/v7.0.0

[Dariusz]

Wrong link. It says the page does not exist...

Copy the whole text. For some reason not the whole line got underlined. I correct it below:

https://github.com/cosmobobak/viriditha ... tag/v7.0.0

Viridithas 7.0.0 for Linux, Mac, etc. here: https://www.chessengeria.com/files

[Brunetti]

Protej 0.6.4
http://protej.info
NOPOP version included for my friend Guenther
Gain +90 in bullet rating.

Alex

[MOBMAT]

Wrong link. It says the page does not exist...

Copy the whole text. For some reason not the whole line got underlined. I correct it below:

https://github.com/cosmobobak/viriditha ... tag/v7.0.0

Viridithas 7.0.0 for Linux, Mac, etc. here: https://www.chessengeria.com/files

I DL'ed it from:
https://chessengines.blogspot.com/2023/ ... -nnue.html

Loaded it up in Arena and found an issue.
The configuration indicates the Hash size can be between 1 and 1048576.
I set it to 16384.
Tried to analyze a position and it hung. I checked the UCI output and during startup it output this:

Code: Select all

[2023-01-29 16:07:16.606<--1:thread 'main' panicked at 'Hash value must be between 1 and 8192', src\uci.rs:275:13
2023-01-29 16:07:16.606<--1:note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

I then set the Hash to 8192 and did not get the error, but when I attempted to analyze a position, the engine output "best move: Bb2-d4" after a few seconds, which is a valid move, but far from the correct move.
Does this engine search? does it support analysis mode (go infinite) ?

[voffka]

Igel 3.4.0 at https://github.com/vshcherbyna/igel/releases/tag/3.4.0 (official executable binaries for Windows and IGN-3 net).

Hello there,

I am really sorry to disrupt the testing community having released Igel 3.3.0 just 2 weeks ago, but ... When debugging some unrelated issue I discovered a really serious bug in Igel's handling of nnue accumulator for null moves. This bug is present since August 2020 and results in Igel having wrong evaluations which is especially noticible in long time control in early and mid games when nnue eval is dominant compared to lazy (HCE) eval.

When the fix was applied to master branch (Igel 3.3.0) I saw a massive ~25 elo increase at LTC: http://chess.grantnet.us/test/31014/.

When I applied this fix to my dev version (upcoming Igel 3.4.0) which features fully NNUE eval I saw huge elo gains ranging from 75 to 90 elo in LTC against Igel 3.3.0: http://chess.grantnet.us/test/31024/

I obviously see a valid reason to interrupt my dev cycle and release whatever Igel state I currently have even though it may not feature the strongest Igel network released in 3.4.0 (HalfKav2_hm).

Disclaimer: Keep in mind that these are indicative numbers and elo gain against other engines may vary.

Igel 3.4.0 is yet another milestone in Igel's development as I removed all HCE code and the eval is purely NNUE.

What's new:

- fixing major bug in accumulator handling for null moves
- avoid heap objects (std::vector) in search function to reduce heap fragementation
- switching to pure nnue evaluation
- new network architecture HalfKav2_hm
- new network ign-3-debc71a4

Regression run against Igel 3.3.0 on Long Time Control (60+0.6)

Code: Select all

ELO   | 75.10 +- 5.38 (95%)
CONF  | 60.0+0.60s Threads=1 Hash=64MB
GAMES | N: 8480 W: 3133 L: 1328 D: 4019

In case of any questions please ask them in the main thread for Igel at https://talkchess.com/forum3/viewtopic. ... 76#p942376

[Wolfgang]

SmallBrain 7.0

[Szil]

BlackCore v5.1: https://github.com/SzilBalazs/BlackCore ... es/tag/5.1

Code: Select all

ELO   | 126.18 +- 17.78 (95%)
CONF  | 8.0+0.08s Threads=1 Hash=16MB
GAMES | N: 1000 W: 489 L: 141 D: 370