Doubling of time control

[fastgm]

Doubling of time control

Conditions:
1 Core, 128 MB Hash, 3000 games, 1500 opening positions, no ponder, no learning, no tablebases
GUI: Cutechess-Cli

Hardware:
Intel i5-750 @ 3.5 GHz

Engine:
Komodo 9.3, default settings

Download PDF:
http://fastgm.de/K93-Doubling-TC.pdf

Time in seconds

[Dann Corbit]

This testing clearly shows a fundamental flaw in the Elo model.
The engine does not get weaker at long time control but stronger.
When given an hour to think, compared to one second, the move chosen with more time allowed will clearly be a much better move.

The increase in draws probably just shows that more careful chess is played by both sides at slower time control.

[Laskos]

Important post, thank you, I will look at it more carefully today. This is the reference now to the doubling of time control.

[Sven]

This testing clearly shows a fundamental flaw in the Elo model.
The engine does not get weaker at long time control but stronger.
When given an hour to think, compared to one second, the move chosen with more time allowed will clearly be a much better move.

The increase in draws probably just shows that more careful chess is played by both sides at slower time control.

The reported Elo ratings are the rating differences between the engine with double time and the engine with normal time. It is just the differences that become smaller.

Based on the table above you get the following ratings if you transform "double time" into "same time but double speed" (@Kai: is the following correct?):

Code: Select all

Speed Elo
   1    0
   2  144
   4  277
   8  389
  16  490
  32  583
  64  656
 128  715
 256  766

Of course the strength of the "Speed = 1" engine is already quite high so this can't be used for weaker engines the same way.

[Guenther]

Important post, thank you, I will look at it more carefully today. This is the reference now to the doubling of time control.

Average ply depth would be interesting for this table too, if it could be extracted.

[fastgm]

The ply depth for each move is saved in the pgn-file, for example: 14. Bh3 {+0.33/27 191s}

Is there a tool that extracts the average ply depth from a pgn?

[Laskos]

The ply depth for each move is saved in the pgn-file, for example: 14. Bh3 {+0.33/27 191s}

Is there a tool that extracts the average ply depth from a pgn?

If there are no such tools, you could ask via personal message Ferdinand Mosca, he helped me many times with similar quests. He writes python scripts, and also builds executables. Very easy to use. You might explain to him the importance of your work by pointing to this thread.

[Laskos]

This testing clearly shows a fundamental flaw in the Elo model.
The engine does not get weaker at long time control but stronger.
When given an hour to think, compared to one second, the move chosen with more time allowed will clearly be a much better move.

The increase in draws probably just shows that more careful chess is played by both sides at slower time control.

The reported Elo ratings are the rating differences between the engine with double time and the engine with normal time. It is just the differences that become smaller.

Based on the table above you get the following ratings if you transform "double time" into "same time but double speed" (@Kai: is the following correct?):

Code: Select all

Speed Elo
   1    0
   2  144
   4  277
   8  389
  16  490
  32  583
  64  656
 128  715
 256  766

Of course the strength of the "Speed = 1" engine is already quite high so this can't be used for weaker engines the same way.

Yes, numbers seem correct, and it's good to have the total ELO. Also, to have relation to CCRL 40/40' rating, I wrote it next to the last data point, where it stands on time control:

Code: Select all

Speed Elo
   1    0
   2  144
   4  277
   8  389
  16  490
  32  583
  64  656
 128  715
 256  766    CCRL 40/40'   3260

[shrapnel]

This testing clearly shows a fundamental flaw in the Elo model.
The engine does not get weaker at long time control but stronger.
When given an hour to think, compared to one second, the move chosen with more time allowed will clearly be a much better move.

The increase in draws probably just shows that more careful chess is played by both sides at slower time control.

I've been crying myself hoarse for the last few years that testing through Blitz/STC games is NO SUBSTITUTE for testing through LTC games.
Glad to be vindicated.
Its not just a matter of engines being "more careful"....LTC games ruthlessly expose any weakness in the Evaluation and Search patterns of the Engine.
That is why in the TCEC, Houdini winning the Rapids is no guarantee that it will beat SF in the super-final ( though I hope it does, just to discomfit Adam ).
Adam has a point though, when he says that the Super-Final will be a different kind of ballgame.
Houdini can win only if it is REALLY the better Engine.

[Uri Blass]

This testing clearly shows a fundamental flaw in the Elo model.
The engine does not get weaker at long time control but stronger.
When given an hour to think, compared to one second, the move chosen with more time allowed will clearly be a much better move.

The increase in draws probably just shows that more careful chess is played by both sides at slower time control.

I've been crying myself hoarse for the last few years that testing through Blitz/STC games is NO SUBSTITUTE for testing through LTC games.
Glad to be vindicated.
Its not just a matter of engines being "more careful"....LTC games ruthlessly expose any weakness in the Evaluation and Search patterns of the Engine.
That is why in the TCEC, Houdini winning the Rapids is no guarantee that it will beat SF in the super-final ( though I hope it does, just to discomfit Adam ).
Adam has a point though, when he says that the Super-Final will be a different kind of ballgame.
Houdini can win only if it is REALLY the better Engine.

I believe that difference between rapid TCEC and slow time control TCEC is small and we do not have enough games.

Conditions are also not the same in long and rapid time control because in rapid time control there are many weak opponents.

Rapid TCEC is clearly slower time control than blitz espacially when you consider the number of cores and is clearly slower than the time controls in this thread.