Beating Stockfish 14 on time

[Henk]

O wait I did not read this:

"Now if your engine does need 10ms to wake up after "stop" + "position fen " + "go" then its another story."

[dangi12012]

We are all guessing here.
Maybe we can have a log for all the uci commands sent to the engine - and all output?
(Like Arena 3.1.1 show all output option)`

That would show if its a gui issue or an engine issue because we all can see where the milliseconds vanish.

[Henk]

Reality:

[pgn]
[Event "Computer Chess Game"]
[Site "LAPTOP-1FK7MTIP"]
[Date "2022.01.05"]
[Round "-"]
[White "Skipper_8_19_2_3"]
[Black "Stockfish 14"]
[Result "0-1"]
[TimeControl "60"]
[Annotator "1. +0.42 1... -0.35"]

1. e4 {+0.42/7} e5 {-0.35/23 4} 2. Nf3 {+0.01/6 1.0} Nc6 {-0.49/20 0.2} 3.
Nc3 {+0.26/6 1.0} Nf6 {-0.04/18 0.6} 4. d4 {+0.11/6 1.0} exd4
{-0.05/23 2.4} 5. Nxd4 {+0.49/7 0.9} Bb4 {-0.26/21 0.2} 6. f3 {+0.25/6 0.9}
d5 {+0.20/21 2.7} 7. Nxc6 {+0.13/7 0.9} bxc6 {+0.60/17 0.5} 8. e5
{-0.11/6 0.9} Ne4 {+1.75/20 1.2} 9. fxe4 {+1.90/7 0.9} Qh4+ {+2.00/20 0.2}
10. Kd2 {+1.21/7 0.9} d4 {+2.16/22 0.3} 11. a3 {+0.45/6 0.9} dxc3+
{+8.45/20 2.0} 12. Ke3 {-0.45/6 0.8} Bc5+ {+14.23/19 1.1} 13. Qd4
{-8.76/7 0.8} Qe1+ {+14.46/18 0.2} 14. Kd3 {-12.04/8 0.8} Ba6+
{+99.97/245 0.8} 15. Qc4 {-327.64/10 0.8} Rd8# {+99.99/245 0.1}
{Xboard adjudication: Checkmate} 0-1
[/pgn]

[pgn]
[Event "Computer Chess Game"]
[Site "LAPTOP-1FK7MTIP"]
[Date "2022.01.05"]
[Round "-"]
[White "Skipper_8_19_2_3"]
[Black "Stockfish 14"]
[Result "0-1"]
[TimeControl "60"]
[Annotator "1. +0.41 1... -0.21"]

1. e4 {+0.41/7} e5 {-0.21/22 3} 2. Nc3 {+0.17/6 1.0} Nf6 {-0.24/20 1.4} 3.
f4 {+0.20/6 1.0} d5 {+0.09/18 0.6} 4. exd5 {+0.58/6 1.0} Nxd5
{+0.40/16 0.6} 5. fxe5 {+0.15/7 0.9} Nxc3 {+0.53/20 1.5} 6. dxc3
{+0.03/6 0.9} Qh4+ {+0.71/18 0.7} 7. Ke2 {+0.27/6 0.9} Bg4+ {+2.71/17 1.0}
8. Nf3 {+0.68/6 0.9} Nd7 {+2.65/18 0.2} 9. Qd5 {+0.55/6 0.9} Bc5
{+4.70/18 1.2} 10. Qxb7 {+1.48/6 0.9} Qf2+ {+8.14/18 0.8} 11. Kd1
{+1.40/6 0.8} Rd8 {+8.83/20 0.1} 12. Bg5 {+1.55/6 0.9} Nf6+ {+9.19/19 1.2}
13. Kc1 {+1.57/6 0.8} Qe1+ {+99.97/245 0.8} 14. Nxe1 {-327.64/7 0.8} Rd1#
{+99.99/245 0.2}
{Xboard adjudication: Checkmate} 0-1
[/pgn]

[Henk]

Skipper9 losing from Stockfish 14 with 16:4 in 5 second games.
Can't play against fairy-max at this speed for fairy max loses on time in almost every game giving 20:0

By the way Skipper9 losing from Stockfish 14 with 18:2 in 8 second games.

[Henk]

Now also 18:2 in 15 second games. Maybe lucky run.

[Henk]

Video for hyperbullet game:


Only needed to play two games to get a win. I deleted the first one (of course)

[dangi12012]

Video for hyperbullet game:


Only needed to play two games to get a win. I deleted the first one (of course)

Can you somehow create millisecond accurate output? Also the strings that are sent to the engine is also important.
Maybe a simple exe that forwards + logs stdio in both directions to and from the engine with a timestamp.

My guess: The GUI does not subtract its own action time (or no fixed buffer time per move). Since Stockfish probably is given more ms than it can use this needs to be debugged.

[Jouni]

I can play SF in CuteChess GUI with 1ms for move. No time losses! How is this possible and what means 10 ms move overhead?

[mvanthoor]

Sometimes I wonder if you're ever doing something useful with your chess program; like trying to improve it or something.

[Jouni]

I just tried to figure fastest time control to detect progress fast. E.g. 5 + 0,05 gives similar result to SPCC or NCM. You get 5000 games fast.