magog.1.0
https://github.com/macsmol/magog/releases/tag/v1.0
First release of my engine written in Go. I guess I did reasonable amount of testing but please let me know if you find some obvious bugs/crashes. Thanks and I hope you'll enjoy it!
New engine releases & news H1 2024
Moderators: hgm, chrisw, Rebel
-
- Posts: 5
- Joined: Sat Jun 15, 2024 4:43 pm
- Full name: Maciej Smolczewski
-
- Posts: 6
- Joined: Sun Jan 15, 2023 1:21 am
- Full name: Krisztian Peocz
Re: New engine releases & news H1 2024
Renegade 1.1.0
https://github.com/pkrisz99/Renegade/re ... tag/v1.1.0
This release features an improved neural network, FRC support, some search patches, and faster move generation.
https://github.com/pkrisz99/Renegade/re ... tag/v1.1.0
This release features an improved neural network, FRC support, some search patches, and faster move generation.
Code: Select all
Score of Renegade 1.1.0 vs Renegade 1.0.0: 520 - 87 - 393 [0.717] 1000
Elo difference: 161.1 +/- 17.2, LOS: 100.0 %, DrawRatio: 39.3 %
Renegade - https://github.com/pkrisz99/Renegade
-
- Posts: 373
- Joined: Thu Jan 22, 2015 3:21 pm
- Location: Zurich, Switzerland
- Full name: Jonathan Rosenthal
Re: New engine releases & news H1 2024
Winter v4.0 has been released! A big thank you to Jim Ablett for the compiles this time around!
https://github.com/rosenthj/Winter/releases/tag/v4.0
This version of Winter has a GNN with the NNUE style perspective feature. It has roughly twice the weights of the v3.0 net and a full evaluation requires roughly 4x the amount of floating point operations. It has been trained on just over 200M positions, over 4x as many samples as the prior net.
Winter's search has also undergone some changes, including a decent amount of tuning which was made possible thanks to OpenBench and everyone contributing there. Winter would not be the same without you guys! The release also contains a key patch contribution by Alexander Litov (Equisetum and Drofa).
Currently running self play regression test vs v3.0:
Elo | 149.21 +- 9.54 (95%)
Conf | 40.0+0.40s Threads=1 Hash=64MB
Games | N: 2418 W: 1103 L: 124 D: 1191
Penta | [1, 51, 370, 542, 245]
http://chess.grantnet.us/test/37432/
https://github.com/rosenthj/Winter/releases/tag/v4.0
This version of Winter has a GNN with the NNUE style perspective feature. It has roughly twice the weights of the v3.0 net and a full evaluation requires roughly 4x the amount of floating point operations. It has been trained on just over 200M positions, over 4x as many samples as the prior net.
Winter's search has also undergone some changes, including a decent amount of tuning which was made possible thanks to OpenBench and everyone contributing there. Winter would not be the same without you guys! The release also contains a key patch contribution by Alexander Litov (Equisetum and Drofa).
Currently running self play regression test vs v3.0:
Elo | 149.21 +- 9.54 (95%)
Conf | 40.0+0.40s Threads=1 Hash=64MB
Games | N: 2418 W: 1103 L: 124 D: 1191
Penta | [1, 51, 370, 542, 245]
http://chess.grantnet.us/test/37432/
-Jonathan
-
- Posts: 19
- Joined: Fri Mar 10, 2023 9:33 am
- Full name: Martin Novák
Re: New engine releases & news H1 2024
Motor 0.6.0 https://github.com/martinnovaak/motor/r ... tag/v0.6.0
approximately +70 elo in selfplay
approximately +70 elo in selfplay
-
- Posts: 84
- Joined: Fri Jun 28, 2024 9:24 am
- Full name: Shawn Xu
-
- Posts: 949
- Joined: Sat May 13, 2006 1:08 am
Re: New engine releases & news H1 2024
Please use this thread for anouncements in second half of 2024
-
- Posts: 1783
- Joined: Fri Jul 14, 2006 7:56 am
- Location: London, England
- Full name: Jim Ablett
Re: New engine releases & news H1 2024
Update:Jim Ablett wrote: ↑Mon May 13, 2024 1:33 am
Tucano v11.19 by Alcides Schulz
https://github.com/alcides-schulz/Tucano
Tucano v11.19 64 ja
Windows/Linux/Android clang profiled builds
https://drive.proton.me/urls/3H358PSYQ0#qsBGH6ieOcBN
Jim.
Tucano 11.24 dev.9ce5a2c 64 JA
Windows/Linux & Android profiled builds with embedded net
https://drive.proton.me/urls/SFCTZFFZHC#F5edsest5Uqf
Jim.