Page 5 of 9

Re: SF PB (Stockfish)

Posted: Sun Dec 31, 2023 2:46 am
by Skynet
SF-PB-301223_x64 "Fix scores from reverse futility pruning." :beer:
Note: I didn't remove the "Slow Mover Option" because it can be useful for online players, besides it is a non functional patch and it doesn't affect the engine's strength. But be careful of how you use it.
From SF team ->
"The SlowMover option allows users to modify the timeLeft variant, impacting the
engine's time management. However, this feature, while theoretically flexible,
doesn't offer substantial benefits. Instead, it introduces the risk of
non-experienced users altering values without a clear understanding of the
effects, potentially leading to a weaker engine.

The vast majority of SF users don't use it anyway, and based on tests conducted
by fauzi several months ago suggest that changing it would only lose Elo."

https://github.com/official-stockfish/S ... ac73fa7074

ENGINE https://pixeldrain.com/u/S5cM12ad
Have a happiest New Year!

Re: SF PB (Stockfish)

Posted: Sun Dec 31, 2023 8:22 am
by ZamChess
Thank you Skynet ,

I wish you and your family a lot of health and happiness in the coming year.

Re: SF PB (Stockfish)

Posted: Sun Dec 31, 2023 1:22 pm
by Solista
Yesterday I tested my engine with Dimension 3072 on Lichess. One of the tests was Bullet with no bonus time and only 3 book moves for Black and only 1 book move for White. I had to set the slow mover significantly higher than 100. If I hadn't done that, the engine would only have been used to its maximum 50% of the time. Anyone can carry out this test themselves and see that you need a different value online to optimize the engine. Thanks for implementing the slow mover.

Result of test
https://solistachess.jimdosite.com/solista-news/

Happy New Year!

Re: SF PB (Stockfish)

Posted: Sun Dec 31, 2023 3:40 pm
by Adrian Oliva
The same good wishes to Alex and to all the friends of the forum! Have a very Happy New Year!
Greetings

Re: SF PB (Stockfish)

Posted: Mon Jan 01, 2024 2:28 pm
by ZamChess
Thank you Eduard ,

I wish you and your family a lot of health and happiness in the new year.

Re: SF PB (Stockfish)

Posted: Tue Jan 02, 2024 2:32 am
by Skynet
SF-PB-010124 "Fix formatting in search.cpp". :beer:
https://pixeldrain.com/u/EAS85bq3

Re: SF PB (Stockfish)

Posted: Mon Jan 08, 2024 2:55 am
by Skynet
SF-PB-070124 "Prefix abs with std::", with dual network, keep both nets in same folder with engine.
Many thanks to Marco for the helping me with code! :thank_you:

https://pixeldrain.com/u/VhbAn7UC
https://easyupload.io/dfyi8l

Re: SF PB (Stockfish)

Posted: Mon Jan 08, 2024 7:15 am
by Solista
Is the dual network now default in Stockfish?

Re: SF PB (Stockfish)

Posted: Mon Jan 08, 2024 5:11 pm
by IbaiBuR
Solista wrote: Mon Jan 08, 2024 7:15 am Is the dual network now default in Stockfish?
Yes, it was merged yesterday into the master branch of Stockfish, so it is the default now.

Re: SF PB (Stockfish)

Posted: Tue Jan 09, 2024 8:51 am
by Solista
IbaiBuR wrote: Mon Jan 08, 2024 5:11 pm
Solista wrote: Mon Jan 08, 2024 7:15 am Is the dual network now default in Stockfish?
Yes, it was merged yesterday into the master branch of Stockfish, so it is the default now.
In my opinion this is nonsense.

Because of some Elo points in bullet games where the small network is faster, you sacrifice the quality of the engine, especially for analysis.

Example:
Image

r1n2N1k/2n2K1p/3pp3/5Pp1/b5R1/8/1PPP4/8 w - - 0 1

Analysis by Stockfish 080124:

1.Nxe6 Nxe6 2.Kxe6 Bxc2 3.b4 Bb3+ 4.Kd7 Nb6+ 5.Kc6 Nc4 6.Rxg5 Ne5+ 7.Kc7 Ra7+ 8.Kb6 Rf7 9.f6 Bd5 10.Ka6 Rxf6 11.d4 h6 12.Rh5 Bf3 13.Rh3 Nd7 14.b5 h5 15.d5 Kg7 16.Ka5 Rf5 17.Rxf3 Rxf3
Depth: 29/46 00:00:10 95711kN, tb=188300
Black is clearly winning. :dunno:

With just one network, Ng6+ = is found immediately. The little stupid network took 30s on 6 cores.