Should TCEC switch to Ponder = ON ?

[Martin Thoresen]

I am debating this with my self, and I know there are people that want Ponder or "Permanent brain" to be activated but I also know that there
are people that want it to stay the way it is now.

If Ponder = ON then the max number of cores or threads each engine can use is 3.

With Ponder = OFF (like it is now) then the max number of cores or threads each engine can use is 6.


What do you think/want? Arguments for either?

[Dann Corbit]

I am debating this with my self, and I know there are people that want Ponder or "Permanent brain" to be activated but I also know that there
are people that want it to stay the way it is now.

If Ponder = ON then the max number of cores or threads each engine can use is 3.

With Ponder = OFF (like it is now) then the max number of cores or threads each engine can use is 6.


What do you think/want? Arguments for either?

My thought is that you should do what you prefer.
The CPU time will come out approximately the same either way, so the quality of the chess will not suffer in either case.

[mhalstern]

I would leave ponder off to get the most out of the cores. The only situation I recommend putting ponder on would be if you were using 2 p's connected via an (do we still use the RS232?) interface. This way each engine could use all available cores all of the time.

[Houdini]

Ponder ON is the closest to a real tournament situation, so I'm in favor of that.

[Dann Corbit]

I would leave ponder off to get the most out of the cores. The only situation I recommend putting ponder on would be if you were using 2 p's connected via an (do we still use the RS232?) interface. This way each engine could use all available cores all of the time.

Half the time with twice the cores = twice the time with half the cores.

[mhalstern]

Thanks for the math lesson. I should have got that earlier.

To run an engine match with 2 pc's what is the current support connection standard? Also, what GUI's will work with this setup?

[Graham Banks]

I am debating this with my self, and I know there are people that want Ponder or "Permanent brain" to be activated but I also know that there
are people that want it to stay the way it is now.

If Ponder = ON then the max number of cores or threads each engine can use is 3.

With Ponder = OFF (like it is now) then the max number of cores or threads each engine can use is 6.


What do you think/want? Arguments for either?

Do what you prefer Martin. You won't please everybody either way.

[MikeB]

I am debating this with my self, and I know there are people that want Ponder or "Permanent brain" to be activated but I also know that there
are people that want it to stay the way it is now.

If Ponder = ON then the max number of cores or threads each engine can use is 3.

With Ponder = OFF (like it is now) then the max number of cores or threads each engine can use is 6.


What do you think/want? Arguments for either?

It's your choice - so there is no right or wrong answer.

But since you asked, I voted for "ponder=on". I believe most authors program their engines for ponder to be on and they are anticipating likewise from their opponent. The area that I believe would be impacted most is time usage. A chess author would anticpate that he would get a certain % of correct ponder hits and program his time usage accordingly and perhaps be more aggressive in time usage ( i.e., that use more time when the program does not ponder correctly.). With no pondering, there are no correct ponder hits and the program would spread his time usage more evenly, knowing that it is impossible to get a correct ponder hit.

[overtond]

I believe all tests should be done this way {2 PCs connected via a network with ponder ON} - I would like to see the UCI/GUI standard modified to include this method and for all test houses {CCRL/CEGT/etc.} to include this test method in their test service offering.

If anyone knows how to do this today using the GUI application {preferably ShredderClassic GUI} then please let me know.

[hgm]

WinBoard / XBoard have been able to run engines on remote machines since the stone age.