Release of 4 new Rebel's with neural nets created by Chris Whittington playing the famous CSTAL style.
http://rebel13.nl/windows/rebel-14.html#other
https://github.com/egh-s/Rebel-14.1.02- ... stem-TAL-1
Release of 4 new Rebel's
Moderators: hgm, Rebel, chrisw
-
- Posts: 7025
- Joined: Thu Aug 18, 2011 12:04 pm
- Full name: Ed Schröder
Release of 4 new Rebel's
90% of coding is debugging, the other 10% is writing bugs.
-
- Posts: 1222
- Joined: Wed Mar 08, 2006 8:28 pm
- Location: Florida, USA
Re: Release of 4 new Rebel's
Brilliant — thanks Ed and Chris!
How has the aggressive CSTal style been embedded into the NNs? Is it trained on positions evaluated using the (old) CSTal evaluation function?
— Steve
How has the aggressive CSTal style been embedded into the NNs? Is it trained on positions evaluated using the (old) CSTal evaluation function?
— Steve
http://www.chessprogramming.net - Maverick Chess Engine
-
- Posts: 18764
- Joined: Thu Mar 09, 2006 6:40 pm
- Location: US of Europe, germany
- Full name: Thorsten Czub
Re: Release of 4 new Rebel's
I tried it out against several engines, all using same CPU threads of course (only 1) and it plays like Tal. It tries and tries to pull you into the dark Forrest, into the fog of no return.
Of course the opponents should be in the same range of ELO, otherwise it's not working.
Now we have Chess System Tal 3 NNUE in
4 nets.
I found 5' very good.
Now maybe the next release will give SMP and szyzgy tablebases and a big gambit book and it kills even the "big ones".
Of course the opponents should be in the same range of ELO, otherwise it's not working.
Now we have Chess System Tal 3 NNUE in
4 nets.
I found 5' very good.
Now maybe the next release will give SMP and szyzgy tablebases and a big gambit book and it kills even the "big ones".
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
Here we have a fairy tale of the day after tomorrow....
-
- Posts: 11660
- Joined: Thu Mar 09, 2006 12:57 am
- Location: Birmingham UK
Re: Release of 4 new Rebel's
mclane wrote: ↑Tue Feb 01, 2022 3:35 pm I tried it out against several engines, all using same CPU threads of course (only 1) and it plays like Tal. It tries and tries to pull you into the dark Forrest, into the fog of no return.
Of course the opponents should be in the same range of ELO, otherwise it's not working.
Now we have Chess System Tal 3 NNUE in
4 nets.
I found 5' very good.
Now maybe the next release will give SMP and szyzgy tablebases and a big gambit book and it kills even the "big ones".
Good, informative review - thanks!
Writing is the antidote to confusion.
It's not "how smart you are", it's "how are you smart".
Your brain doesn't work the way you want, so train it!
It's not "how smart you are", it's "how are you smart".
Your brain doesn't work the way you want, so train it!
-
- Posts: 18764
- Joined: Thu Mar 09, 2006 6:40 pm
- Location: US of Europe, germany
- Full name: Thorsten Czub
Re: Release of 4 new Rebel's
Wait until you see the tal games…
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
Here we have a fairy tale of the day after tomorrow....
-
- Posts: 2460
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
Re: Release of 4 new Rebel's
What about changing Stockfish-code, so that Stockfish can use these interesting new nets? This will not lead to a stronger Stockfish, of course, but perhaps to a very interesting playing (and somewhat weaker) Stockfish - which would be still much, much stronger than Rebel 14.1 (and uses multithreading etc.).Rebel wrote: ↑Tue Feb 01, 2022 11:21 am Release of 4 new Rebel's with neural nets created by Chris Whittington playing the famous CSTAL style.
http://rebel13.nl/windows/rebel-14.html#other
https://github.com/egh-s/Rebel-14.1.02- ... stem-TAL-1
I would love this and testing it for lower draw-rates and shorter games than the default-Stockfish!
Regards - Stefan (SPCC)
-
- Posts: 1632
- Joined: Tue Aug 21, 2018 7:52 pm
- Full name: Dietrich Kappe
Re: Release of 4 new Rebel's
Maybe if stockfish makes it’s nets available for Rebel that might be a fair trade. I’m thinking of making some of my nets available for Rebel 14.1.pohl4711 wrote: ↑Tue Feb 01, 2022 8:18 pmWhat about changing Stockfish-code, so that Stockfish can use these interesting new nets? This will not lead to a stronger Stockfish, of course, but perhaps to a very interesting playing (and somewhat weaker) Stockfish - which would be still much, much stronger than Rebel 14.1 (and uses multithreading etc.).Rebel wrote: ↑Tue Feb 01, 2022 11:21 am Release of 4 new Rebel's with neural nets created by Chris Whittington playing the famous CSTAL style.
http://rebel13.nl/windows/rebel-14.html#other
https://github.com/egh-s/Rebel-14.1.02- ... stem-TAL-1
I would love this and testing it for lower draw-rates and shorter games than the default-Stockfish!
Regards - Stefan (SPCC)
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".
-
- Posts: 18764
- Joined: Thu Mar 09, 2006 6:40 pm
- Location: US of Europe, germany
- Full name: Thorsten Czub
Re: Release of 4 new Rebel's
And how do you want to find out that stockfish tal plays unsound ?! By testing it against Komodo?pohl4711 wrote: ↑Tue Feb 01, 2022 8:18 pmWhat about changing Stockfish-code, so that Stockfish can use these interesting new nets? This will not lead to a stronger Stockfish, of course, but perhaps to a very interesting playing (and somewhat weaker) Stockfish - which would be still much, much stronger than Rebel 14.1 (and uses multithreading etc.).Rebel wrote: ↑Tue Feb 01, 2022 11:21 am Release of 4 new Rebel's with neural nets created by Chris Whittington playing the famous CSTAL style.
http://rebel13.nl/windows/rebel-14.html#other
https://github.com/egh-s/Rebel-14.1.02- ... stem-TAL-1
I would love this and testing it for lower draw-rates and shorter games than the default-Stockfish!
Regards - Stefan (SPCC)
The engine creates art, like a painter does, on fruit level.
Imo on the level of stockfish, this is difficult to measure.
So far chess system tal 3 NNUE plays only one core and only on fruit level.
It is still on a high level and much higher then chess system tal 2 was doing, because search has improved a lot from 1999 to 2004.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
Here we have a fairy tale of the day after tomorrow....
-
- Posts: 7025
- Joined: Thu Aug 18, 2011 12:04 pm
- Full name: Ed Schröder
Re: Release of 4 new Rebel's
Sure, the more the better and I can start a new rating list, the RRL, just kidding.dkappe wrote: ↑Tue Feb 01, 2022 8:26 pmMaybe if stockfish makes it’s nets available for Rebel that might be a fair trade. I’m thinking of making some of my nets available for Rebel 14.1.pohl4711 wrote: ↑Tue Feb 01, 2022 8:18 pmWhat about changing Stockfish-code, so that Stockfish can use these interesting new nets? This will not lead to a stronger Stockfish, of course, but perhaps to a very interesting playing (and somewhat weaker) Stockfish - which would be still much, much stronger than Rebel 14.1 (and uses multithreading etc.).Rebel wrote: ↑Tue Feb 01, 2022 11:21 am Release of 4 new Rebel's with neural nets created by Chris Whittington playing the famous CSTAL style.
http://rebel13.nl/windows/rebel-14.html#other
https://github.com/egh-s/Rebel-14.1.02- ... stem-TAL-1
I would love this and testing it for lower draw-rates and shorter games than the default-Stockfish!
Regards - Stefan (SPCC)
90% of coding is debugging, the other 10% is writing bugs.
-
- Posts: 11660
- Joined: Thu Mar 09, 2006 12:57 am
- Location: Birmingham UK
Re: Release of 4 new Rebel's
Both the new Rebel and Stockfish are open source, so Rebel could use the Stockfish net if it chose to (giving acknowledgement) - or am I missing something?
Writing is the antidote to confusion.
It's not "how smart you are", it's "how are you smart".
Your brain doesn't work the way you want, so train it!
It's not "how smart you are", it's "how are you smart".
Your brain doesn't work the way you want, so train it!