Stockfish 14.10.2025 is a big improvement

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

Moderator: Ras

Jouni
Posts: 3687
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Stockfish 14.10.2025 is a big improvement

Post by Jouni »

My short UHO Test:

Code: Select all

                         
1   SFdev           +24  +59/=96/-45 53.50%  107.0/200
2   stockfish17.1   -24  +45/=96/-59 46.50%   93.0/200

RT has currently +28 elo 1 and 8 cores against SF17. Changes bigger PAWN_HISTORY_SIZE, Remove condition in qsearch and Use huge pages for worker data.
Jouni
Jouni
Posts: 3687
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Stockfish 14.10.2025 is a big improvement

Post by Jouni »

After 400 games

Code: Select all


                         
1   SFdev           +10  +110/=192/-98 51.50%  206.0/400
2   stockfish17.1   -10  +98/=192/-110 48.50%  194.0/400

My previous test was +3 elo.
Jouni
Jouni
Posts: 3687
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Stockfish 14.10.2025 is a big improvement

Post by Jouni »

NCM has tested hundreds of useless patches. But where is it now when there is progress :cry: ?
Jouni
Jouni
Posts: 3687
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Stockfish 14.10.2025 is a big improvement

Post by Jouni »

Testing starts now, but no all time high score :roll: . https://nextchessmove.com/dev-builds
Jouni
kiroje
Posts: 73
Joined: Wed Jul 25, 2012 10:12 am
Location: Copenhagen, Denmark
Full name: Kim Jensen

Re: Stockfish 14.10.2025 is a big improvement

Post by kiroje »

Unfortunately it still seems to have a search explosion problem which now is even worse than version 17.1. It seems that the developers have struggled with this since version 16 as far as I can read(i may be wrong of course)
“Modern chess is too much concerned with things like pawn structure. Forget it, checkmate ends the game.” – Nigel Short
Uri Blass
Posts: 10905
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Stockfish 14.10.2025 is a big improvement

Post by Uri Blass »

kiroje wrote: Tue Oct 21, 2025 2:07 pm Unfortunately it still seems to have a search explosion problem which now is even worse than version 17.1. It seems that the developers have struggled with this since version 16 as far as I can read(i may be wrong of course)
What do you mean by search explosion problem?

Are there positions when it cannot go deep because of extensions?
kiroje
Posts: 73
Joined: Wed Jul 25, 2012 10:12 am
Location: Copenhagen, Denmark
Full name: Kim Jensen

Re: Stockfish 14.10.2025 is a big improvement

Post by kiroje »

Uri Blass wrote: Tue Oct 21, 2025 5:51 pm
kiroje wrote: Tue Oct 21, 2025 2:07 pm Unfortunately it still seems to have a search explosion problem which now is even worse than version 17.1. It seems that the developers have struggled with this since version 16 as far as I can read(i may be wrong of course)
What do you mean by search explosion problem?

Are there positions when it cannot go deep because of extensions?
I tested this and look at the node counts program reports from UCI

[d]7k/8/5KBP/8/8/8/8/8 w - - 0 1

Code: Select all

								
						10	11	12	13	14	15	16	17		18
Stockfish 17.1					135.609	136.545	151.696	211.766	375.027	446.812	553.487	674.836		682.161
Stockfish dev-20251014-75edbee0			182.025	182.119	182.190	217.999	222.246	504.981	509.870	1.290.296.086
PlentyChess 7.0.0				3.664	5.006	6.323	7.402	13.212	15.672	22.126	27.019		32.354
Berserk 13					2.737	4.931	16.184	23.655	56.664	150.905	236.398	240.306		263.463
I simply stopped after 2 hours search for stockfish newest development at depth 18 so it isnt there
“Modern chess is too much concerned with things like pawn structure. Forget it, checkmate ends the game.” – Nigel Short