Stockfish disappointing result against stockfish

[Damir]

My dissapointment with the SF team is as follows... they should work harder in improving big hardware win over a smaller hardware.
All I get is draws with my 64 cores when playing smaller hardware. I remember a time when having a big hardware was a bonus over smaller hardware. Today that is not the case anymore, as nnue has evened the gap between small and big hardware. It seems to me SF team are only interested in improving Stockfish for small machines and not big machines.

[Sopel]

So you want Stockfish to play suboptimal moves and hope the opponent makes a mistake. That's not how good engines work. Improving an engine will inevitably make it more drawish in objectively drawn positions.

[Uri Blass]

So you want Stockfish to play suboptimal moves and hope the opponent makes a mistake. That's not how good engines work. Improving an engine will inevitably make it more drawish in objectively drawn positions.

I do not want stockfish to play suboptimal moves.

I want stockfish to play optimal moves that set traps to the opponent.

If there are 2 drawing moves when the first is a simple draw and the second set a good trap for the opponent(but is also a draw if the opponent is strong enough) then I want stockfish to play the move that set the good trap.

[tcusr]

So you want Stockfish to play suboptimal moves and hope the opponent makes a mistake. That's not how good engines work. Improving an engine will inevitably make it more drawish in objectively drawn positions.

I do not want stockfish to play suboptimal moves.

I want stockfish to play optimal moves that set traps to the opponent.

If there are 2 drawing moves when the first is a simple draw and the second set a good trap for the opponent(but is also a draw if the opponent is strong enough) then I want stockfish to play the move that set the good trap.

alphabeta expects optimal play by both parts.

[Javier Ros]

So you want Stockfish to play suboptimal moves and hope the opponent makes a mistake. That's not how good engines work. Improving an engine will inevitably make it more drawish in objectively drawn positions.

I do not want stockfish to play suboptimal moves.

I want stockfish to play optimal moves that set traps to the opponent.

If there are 2 drawing moves when the first is a simple draw and the second set a good trap for the opponent(but is also a draw if the opponent is strong enough) then I want stockfish to play the move that set the good trap.

There was a chess engine called TrappyBeowulf that identify and set traps, see

https://github.com/vollmerm/TrappyBeowulf
https://ieeexplore.ieee.org/document/41 ... er=4100129

it would be very interesting to implement these ideas in Stockfish and get a more interesting style of play.

[Ovyron]

If there are 2 drawing moves when the first is a simple draw and the second set a good trap for the opponent(but is also a draw if the opponent is strong enough) then I want stockfish to play the move that set the good trap.

What you want is called "swindle", and it's an incredibly difficult thing to implement without hurting elo.

What is happening here is that Stockfish 15 is really damn good at defending, and I'm the kind of person that has never used the word damn lightly. So it'll find the best moves that can draw the game really fast without needing high node count or big hardware.

Kayra is a Stockfish derivative, it was made famous because someone created a really wide book that played all sorts of openings including gambits and then it went undefeated for more than 1000 games at playchess...

You can play as black gambits that lead to some +1.80 scores and Stockfish can defend them, so what do you expect?

If you want this swindle concept to work, you need to make an opening book that beats that stockfish that you think should be losing, you show us the moves that beat it (the traps), then how Stockfish 15 doesn't play the traps. I'm sure then someone will look into it and possibly implement swindling (playing those traps without the book.)

Otherwise, all we see is a brick wall and bombs that can't break it, we need to see the moves you want Stockfish 15 to play to try to win.

[Werewolf]

SF Dev been quiet since June 7th, is it just holiday period or something else?

[Andrew]

They often go for a week or few between new versions. Probably just a few tests in a row not
getting a decent result.

I'd also really like them to have a second page (or a filter option?) that shows just releases from the major
updates (eg V15, 14.1 14 etc) or updates to the network files.

Andrew

[Werewolf]

They often go for a week or few between new versions. Probably just a few tests in a row not
getting a decent result.

I'd also really like them to have a second page (or a filter option?) that shows just releases from the major
updates (eg V15, 14.1 14 etc) or updates to the network files.

Andrew

Thanks Andrew.

[Lazy_Frank]

My dissapointment with the SF team is as follows... they should work harder in improving big hardware win over a smaller hardware.
All I get is draws with my 64 cores when playing smaller hardware. I remember a time when having a big hardware was a bonus over smaller hardware. Today that is not the case anymore, as nnue has evened the gap between small and big hardware. It seems to me SF team are only interested in improving Stockfish for small machines and not big machines.

I am quite happy Stockfish team choose the way improve algorithm instead of begging on big hardware resources, like Leela project currently (due big networks).
Finding a amazing strong moves on average home user PC makes Stockfish very friendly to users (i guess).