Stockfish time management

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

Moderators: hgm, Rebel, chrisw

Viz
Posts: 219
Joined: Tue Apr 09, 2024 6:24 am
Full name: Michael Chaly

Re: Stockfish time management

Post by Viz »

Yeah, it's all cool, but imagine stockfish playing without access to opening theory like it always does unlike your GM.
Look at GM time management in FRC then where they (almost) don't have prep, and surprisingly you will see them spending half of their time on first 10 moves.
So what stockfish does is perfectly "human-like" just that it's like human who completely forgot all his opening preparation.
mosfel24
Posts: 12
Joined: Thu May 02, 2024 4:24 pm
Full name: Moshe Felman

Re: Stockfish time management

Post by mosfel24 »

Viz wrote: Thu Aug 01, 2024 4:37 pm Yeah, it's all cool, but imagine stockfish playing without access to opening theory like it always does unlike your GM.
Look at GM time management in FRC then where they (almost) don't have prep, and surprisingly you will see them spending half of their time on first 10 moves.
So what stockfish does is perfectly "human-like" just that it's like human who completely forgot all his opening preparation.
Are you telling me that Stockfish is like a GM without preparation in terms of opening? In that case, I believe that Stockfish can do better; indeed, that it can gain a significant chunk of elo by hardcoding in optimal opening moves.

FRC is a whole different story, as we are talking about standard chess here. I do not understand why you derail the conversation.
Ras
Posts: 2582
Joined: Tue Aug 30, 2016 8:19 pm
Full name: Rasmus Althoff

Re: Stockfish time management

Post by Ras »

mosfel24 wrote: Thu Aug 01, 2024 5:09 pmthat it can gain a significant chunk of elo by hardcoding in optimal opening moves.
That's called "opening book". If you let Stockfish play with opening book, you'll see that it takes no time at all for moves like 1. e4 because it's in the book. Now, putting an opening book into Stockfish to bypass the user setting for the opening book... that's called "cheating". You can gain Elo in engine matches by cheating? Who would have thought.
Rasmus Althoff
https://www.ct800.net
Peter Berger
Posts: 693
Joined: Thu Mar 09, 2006 2:56 pm

Re: Stockfish time management

Post by Peter Berger »

Viz wrote: Thu Aug 01, 2024 4:37 pm Yeah, it's all cool, but imagine stockfish playing without access to opening theory like it always does unlike your GM.
Look at GM time management in FRC then where they (almost) don't have prep, and surprisingly you will see them spending half of their time on first 10 moves.
So what stockfish does is perfectly "human-like" just that it's like human who completely forgot all his opening preparation.
But is this really true? I'd expect the net to have learned quite a lot about the classical starting position by now. :D
chesskobra
Posts: 271
Joined: Thu Jul 21, 2022 12:30 am
Full name: Chesskobra

Re: Stockfish time management

Post by chesskobra »

mosfel24 wrote: Thu Aug 01, 2024 4:04 pm
I have hardcoded this graph, taken from CPW, into all of my engines. Of course it is scaled to the time control my engine is playing at, and it works perfectly. When I watch the games, it seems so human-like. I have also expanded this to cyclic time controls with my good judgement.
Which is your engine? It is an interesting idea. But I think the time management of engines and GMs would be different. Especially in engine-engine matches the games would be much longer on average (I don't know the numbers, but is the average number of moves like ~90 for matches between strong engines?) So for me a question of interest is how many moves an engine should aim to make before it starts playing only on increment? Or even better, what game state (in terms of the number of pieces left) an engine should aim before it starts to play only on increment? I have observed that a position with +2 or +3 advantage and 14 pieces on the board can easily take Stockfish another 30-40 moves to convert against another engine. Also, many games between strong engines would reach endgame phase. So should an engine play an entire 14 piece endgame only on increment?