Stockfish's blunders in the SSDF games

[Modern Times]

These blunders are in the last move of the time control and reduce the rating of stockfish13,14,15(there may be more games and I searched only for losses and not for cases when the blunder caused stockfish to draw instead of win).

I am not sure if it was a problem of previous versions of stockfish in the ssdf rating list.

It is a well known issue, it happens in CCRL 40/15 games as well.

Current SF still loses with X/Y time control. Issue never fixed! https://github.com/official-stockfish/S ... ssues/4000.

[syzygy]

These blunders are in the last move of the time control and reduce the rating of stockfish13,14,15(there may be more games and I searched only for losses and not for cases when the blunder caused stockfish to draw instead of win).

I am not sure if it was a problem of previous versions of stockfish in the ssdf rating list.

It is a well known issue, it happens in CCRL 40/15 games as well.

Current SF still loses with X/Y time control. Issue never fixed! https://github.com/official-stockfish/S ... ssues/4000.

According to the issue, it is a GUI issue. Apparently the GUI used introduces delays of up to 1.5 seconds.

[Modern Times]

In the CCRL case maybe so. But SSDF use a different GUI.

SSDF do not to my knowledge use ChessGUI that Graham uses for 40/15. Is it 2 GUIs that are at fault ? Possibly.

[Graham Banks]

In the CCRL case maybe so. But SSDF use a different GUI.

SSDF do not to my knowledge use ChessGUI that Graham uses for 40/15. Is it 2 GUIs that are at fault ? Possibly.

All GUIs except Cutechess I think.

[Jouni]

If I set Move Overhead 1000ms no time losses anymore. But SF time use is weird. E.g. with 60 moves in 5 minutes it usually has used only 1 minute after 30 moves! I think engine should use 50% after 30 moves .

[Ciekce]

Garry Kasparov still plays much better chess both strategically and positionally.

then why is he worse at the game than any superhuman engine

[Jouni]

Nice joke

Champions Showdown: Chess 9LX Field

GM Garry Kasparov 2552
GM Hikaru Nakamura 2808
GM Fabiano Caruana 2789
GM Wesley So 2781
GM Levon Aronian 2771
GM Leinier Dominguez 2727
GM Jeffery Xiong 2700
GM Ray Robson 2691
GM Sam Sevian 2689
GM Sam Shankland 2685

[syzygy]

In the CCRL case maybe so. But SSDF use a different GUI.

SSDF do not to my knowledge use ChessGUI that Graham uses for 40/15. Is it 2 GUIs that are at fault ? Possibly.

Do all engines suffer from this problem now and then, or is it only Stockfish?

And does it still happen if MoveOverhead is increased as suggested?

[syzygy]

If I set Move Overhead 1000ms no time losses anymore. But SF time use is weird. E.g. with 60 moves in 5 minutes it usually has used only 1 minute after 30 moves! I think engine should use 50% after 30 moves .

I guess Stockfish expects every move to incur a 1000ms delay, and in reality the GUI might only lag occasionally?

[Eelco de Groot]

In Thorsten's tournament Stockfish could lose on time because he simply used the default Move Overhead, which is only suited for Fishtest very fast games. Default is 10 ms I believe. With Cutechess and fast games that is possible. It would be strange if SSDF also used the default but it would explain timelosses at these long SSDFgames? Time control roughly comparable to TCEC where Stockfish also had the same issue once.