ShashChess

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, chrisw, Rebel

Viz
Posts: 223
Joined: Tue Apr 09, 2024 6:24 am
Full name: Michael Chaly

Re: ShashChess

Post by Viz »

Aligned with Stockfish 17.
Why are you using stockfish 17 patches if they clearly use flawed methodology of testing at bullet?
criko
Posts: 141
Joined: Wed Feb 28, 2018 2:50 pm

Re: ShashChess

Post by criko »

Jouni
Posts: 3484
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: ShashChess

Post by Jouni »

As expected bigger net and other changes make ShashChess worse solver (High Tal). Just one example. 110 positions HTC test. ShashChess 35.1 solved 93 and used 26 minutes. Version 36 solved 84 and used 36 minutes. Also questionable "hide current move" patch included.
Jouni
criko
Posts: 141
Joined: Wed Feb 28, 2018 2:50 pm

Re: ShashChess

Post by criko »

Jouni wrote: Fri Sep 20, 2024 5:55 pm As expected bigger net and other changes make ShashChess worse solver (High Tal). Just one example. 110 positions HTC test. ShashChess 35.1 solved 93 and used 26 minutes. Version 36 solved 84 and used 36 minutes. Also questionable "hide current move" patch included.
What does this mean for practical game play? Will it play worse chess?
Hai
Posts: 629
Joined: Sun Aug 04, 2013 1:19 pm

Re: ShashChess

Post by Hai »

Jouni wrote: Fri Sep 20, 2024 5:55 pm As expected bigger net and other changes make ShashChess worse solver (High Tal). Just one example. 110 positions HTC test. ShashChess 35.1 solved 93 and used 26 minutes. Version 36 solved 84 and used 36 minutes. Also questionable "hide current move" patch included.
Bigger net means: 20000 kn/s -> 10000 kn/s
Solution: Use 16 cores CPU instead of 8 cores. 10000 kn/s -> 20000 kn/s
Hai
Posts: 629
Joined: Sun Aug 04, 2013 1:19 pm

Re: ShashChess

Post by Hai »

criko wrote: Sat Sep 21, 2024 10:00 am
Jouni wrote: Fri Sep 20, 2024 5:55 pm As expected bigger net and other changes make ShashChess worse solver (High Tal). Just one example. 110 positions HTC test. ShashChess 35.1 solved 93 and used 26 minutes. Version 36 solved 84 and used 36 minutes. Also questionable "hide current move" patch included.
What does this mean for practical game play? Will it play worse chess?
It's tactically weaker but positionally and strategically stronger (Stockfish 17 +46 elo).
To get the same tactical strength, which you had before, you need to BUY better hardware. Instead of old AMD Ryzen 8 cores CPU, buy AMD 9950X or 9950X3D 16 cores CPU.
If you have a 7950X 16 cores CPU, go for the 7970X 32 cores CPU.
Of course the positional and strategical strength will increase too.
Hai
Posts: 629
Joined: Sun Aug 04, 2013 1:19 pm

Re: ShashChess

Post by Hai »

@criko

If you use ShashChess, use also High Tal and MultiPV=4 and MCTS on and MCTSThreads=2 or more.
Example:
CPU with 8 cores.
Threads=6
MCTSThreads=2
You should get around +10 solved positions more.
User avatar
Rebel
Posts: 7254
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: ShashChess

Post by Rebel »

Hai wrote: Sat Sep 21, 2024 4:59 pm
criko wrote: Sat Sep 21, 2024 10:00 am
Jouni wrote: Fri Sep 20, 2024 5:55 pm As expected bigger net and other changes make ShashChess worse solver (High Tal). Just one example. 110 positions HTC test. ShashChess 35.1 solved 93 and used 26 minutes. Version 36 solved 84 and used 36 minutes. Also questionable "hide current move" patch included.
What does this mean for practical game play? Will it play worse chess?
It's tactically weaker but positionally and strategically stronger (Stockfish 17 +46 elo).
Laughable.
90% of coding is debugging, the other 10% is writing bugs.