I was replaying a won game by Rebel against SF16 and noticed SF played almost instantly one move before the time control, move 40 and 80. I don't think it did matter in this game but could be disastrous.
Ghppn likes this post
RubiChess
Posts : 7 Join date : 2023-04-29
Subject: Re: SF time control weakness Sat Oct 14, 2023 7:35 am
Known problem with bad GUI / wrong setting of option "Move Overhead".
Posts : 2608 Join date : 2020-11-17 Location : Netherlands
Subject: Re: SF time control weakness Sat Oct 14, 2023 9:55 am
Seems to me as the programmer something you can fix in the code since you in what type of time control you are. Am I mistaken?
Mclane and Ghppn like this post
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: SF time control weakness Sat Oct 14, 2023 10:46 am
Its not a bug. Its a feature.
Ghppn likes this post
Admin Admin
Posts : 2608 Join date : 2020-11-17 Location : Netherlands
Subject: Re: SF time control weakness Sat Oct 14, 2023 10:51 am
Mclane wrote:
Its not a bug. Its a feature.
A feature with no future
Mclane and Ghppn like this post
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: SF time control weakness Sat Oct 14, 2023 11:40 am
This feature is in the engine not only in stockfish 16 but in 15 etc. And was the reason i had problems in arena because arena is a slow GUI.
With move overhead etc. you have to „fine tune“ on each hardware / GUI combination.
Ghppn likes this post
RubiChess
Posts : 7 Join date : 2023-04-29
Subject: Re: SF time control weakness Sat Oct 14, 2023 2:52 pm
Admin wrote:
Seems to me as the programmer something you can fix in the code since you in what type of time control you are. Am I mistaken?
No. The programmer has already done his job by offering an option "Move Overhead" for the user to make the engine interact with different GUIs (good ones like cutechess and buggy ones like Arena) in best possible way. The problem is not about the time control, it is just most visible in moves/time when user didn't set Move Overhead appropriate for his GUI/computer. RTFM helps again: https://github.com/official-stockfish/Stockfish/wiki/UCI-&-Commands#setoption has a section about Move Overhead.
One could argue if the default value 10ms is good but SF developers and most of the important tourneys and testers use cutechess which works perfectly with this value and you don't want to waste like 1000ms with a default value that makes even Arena users happy when you play tons of STC games with cutechess.
And Commander McLane doesn't need to fine-tune anything. Just use max. possible value of 5000 for Move Overhead should work on any computer and with any tool from the past.
Regards, Andreas
Ghppn likes this post
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: SF time control weakness Sat Oct 14, 2023 4:08 pm
„One could argue if the default value 10ms is good“ : indeed. One could argue. One could also argue why other engines do it in a much less complicated way.
It reminds me on tesla cars where you have to input strange commands on the touch screen to get normal features done where in normal cars you have a button .
E.g. to open the glove department you have to go on the touch screen in a submenu. While in normal cars you just open it with your hands.
Nezhman and Ghppn like this post
RubiChess
Posts : 7 Join date : 2023-04-29
Subject: Re: SF time control weakness Sat Oct 14, 2023 4:27 pm
Mclane wrote:
„One could argue if the default value 10ms is good“ : indeed. One could argue. One could also argue why other engines do it in a much less complicated way.
You mean less complicated like setting number of threads to an appropriate value in Rebel/CSTal via option that is hidden to the GUI?
Nothing more to say.
Ghppn likes this post
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: SF time control weakness Sat Oct 14, 2023 8:46 pm
Nice whataboutism but i would direct this feature to the programmers in charge.