Stockfish 18 losing on time

[Francois Vannier]

Hi,

I am using the last V4.2.2 on a Raspberry Pi 4b 8gb. Playing SF18 in Auto mode at 15+10 or 10+5 with white, very often SF lost on time. This could happen even in the early phase of the game. No issue with SF16 in Auto mode.

Is this a known issue ? Any log to collect ?

Thanks.

Francois

[Randy Reade]

Not known to me. It should not happen that often, though. The stock StockfishNN.uci file uses a Move Overhead of 2000, which should prevent time issues. The engine default is 10! You could try increasing it (max 5000) and play some games (shorter would be easier on you. :^)). Could be something else, though.

As for the log, you could add the following line to StockfishNN.uci:

Debug Log File = /home/pi/SF18_debug.log

Then play a game. If it loses on time, zip the SF18_debug.log file and attach it here.

If it doesn't lose on time, delete the log file (as it might get fairly large) and try again until you capture a game where SF flagged.

Don't forget to change the .uci file back (or comment out the Debug line using a semi-colon).

Maybe others have seen the same issue?

Randy

--
You received this message because you are subscribed to the Google Groups "PicoChess" group.
To unsubscribe from this group and stop receiving emails from it, send an email to picochess+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/picochess/CADa9EqUGHoVTeGJ8%3DLDO3%2BgdKiKRxp75t%3DaZ8zf2f_DeP3Bnxw%40mail.gmail.com.

[Francois Vannier]

Hi Randy,

Here is the log, one attempt was enough. I was playing white, 10+5 with SF18 in Auto mode. I switched back to the DGT board, to rule out the possibility the Chessnut Move board might be the cause.

Thanks.

Francois

To view this discussion visit https://groups.google.com/d/msgid/picochess/CAE5WduujJ5NqZAK1kGM7ioJS1beOaQ_g7TFgRwdi39RA%2Bq2Eeg%40mail.gmail.com.

[Randy Reade]

Hi Francois,

I guess you must have been using a Book setting since the first move SF processes is after e2e4 g8f6 e4e5. Some questionable moves by SF :^) but I don't think it's a SF issue although its time management is questionable but likely normal at low strength settings.

At the end, with SF down to 12 seconds in the clock, it is sent the UCI commands:

>> position startpos moves e2e4 g8f6 e4e5 f6e4 d2d3 e4f6 e5f6 g7f6 b1c3 d7d5 g1f3 e7e5 d3d4 e5e4 f3d2 b8c6 f1b5 a7a6 b5c6 b7c6 d2b3 f8d6 e1g1 h7h5 h2h3 d8e7 f1e1 c8e6 f2f3 h8g8 g1f1 f6f5 f3e4 f5e4 d1h5 e7f8 c1h6
>> go wtime 521000 btime 12000 winc 5000 binc 5000

(Note for anyone interested, this is fen r3kqr1/2p2p2/p1pbb2B/3p3Q/3Pp3/1NN4P/PPP3P1/R3RK2 b q - 2 19)

SF starts thinking and responds after 5.2 seconds with:

<< info depth 17 seldepth 46 multipv 1 score cp -177 nodes 606314 nps 114919 hashfull 61 tbhits 0 time 5276 pv f8e7 e1e2 g8g6 h6e3 e8c8 e2f2 d6g3 c3e2 g3f2 e3f2 e7f6 h5e5 e6h3 g2h3 f6f3 e5f4 f3h1 e2g1 e4e3 f2e3 h1g2 f1e1 d8e8 e1d1 g6g3 f4f5 c8b7 b3a5 b7a7
<< bestmove g8g2 ponder h6f8

But, then something happens as the next command from PicoCHess is:

>> position startpos
>> go depth 50

for the analysis. 22.8 seconds later PicoChess sends:

>> quit

Can you add 'Analysis = false' to StockfishNN and test some more?

Randy

[Francois Vannier]

Thanks Randy,

You nailed it, I guess : 3 games in a row without any issue with 'Analysis = false'.

The time management seems always questionable : most of the main time is consumed in a dozen moves and then it did manage the remaining moves with the Fischer increments.  Moreover, the Auto level is really weak, more than expected from the "Auto" mode ;-)

I removed the log line & I'll continue to play with this setting and let you know of any adversarial event.

Francois 

--

You received this message because you are subscribed to the Google Groups "PicoChess" group.
To unsubscribe from this group and stop receiving emails from it, send an email to picochess+...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/picochess/CAE5WduumvQ0%2Bz2%3DKfT4gA_aV%2BmOJFzfm6pfc-O6q9_W%3DoTTVPw%40mail.gmail.com.

[Johan Sjöblom]

Hello Francois.

Would it be possible for you to send the picochess log also: /opt/picochess/logs/picochess.log

BR, Johan

[Johan Sjöblom]

This is a bit technical, but as a result of your stockfish debug file I found a way to save CPU power so I made a hotfix.

Before this hotfix, after the engine returned a move, Picochess could start the normal “user-turn engine analysis” even though the internal game board had not yet been updated with that move (waiting for the physical eboard execution). That meant analysis could run on a stale pre-engine-move position.
After the hotfix, this user-turn engine analysis is blocked until the move is actually executed on the eboard and pushed to the internal game board. This reduces unnecessary CPU load and can explain some odd analysis output seen in that window.
It does not by itself explain Stockfish time losses.

-- Johan

[DJ Dekker]

Hi Francois,

I've also been seeing some funny Stockfish behaviour lately. Sometimes (not always) it loses on time in bullet games, but sometimes there are also error messages in the log about the analysis engine not being loaded shortly after playing 1.e4:

I'm trying to understand what's going on here.

Greetings,

DJ

Op donderdag 5 maart 2026 om 20:17:08 UTC+1 schreef messi...@gmail.com:

[DJ Dekker]

Hi Johan,

I applied your hotfix "fix startup ok timing and ordering" and there are no error messages anymore. Thanks for the quick fix!

Greetings,

DJ