How many elo do ponder equals to

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.

Which engine do you believe got the upper hand in 1 minute bullet without increment

Engine A | 4-core with ponder
0
No votes
Engine B | 8-core ponder off
0
No votes
 
Total votes: 0

MikeB
Posts: 3393
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: How many elo do ponder equals to

Post by MikeB » Sun Jun 23, 2019 6:50 am

Nordlandia wrote:
Sun Jun 23, 2019 6:04 am
MikeB: how did you let 3-core utilize ponder?

I don't know of GUI that allow one side use ponder and the other not. Like a sort of handicap.

One solution is to disable ponder in the stockfish source code if that is a possible.
I use engine match software that was written by Bob Hyatt - it's very flexible. Each engine can have it's own parameters, something most GUI's do not allow you to do. It's not even a GUI, it's all terminal based.

Uri Blass
Posts: 8557
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: How many elo do ponder equals to

Post by Uri Blass » Sun Jun 23, 2019 7:14 am

MikeB wrote:
Sat Jun 22, 2019 10:57 pm
zullil wrote:
Sat Jun 22, 2019 12:44 pm
Nordlandia wrote:
Sat Jun 22, 2019 12:21 pm
Engine A ponder while engine B doesn't. Advantage +/- 50 elo. Doubling is worth more. Now the question is if it's likely engine A equalize engine B in fast time formats or who got the upper hand. The effectiveness of ponder is probably higher in shorter time controls than in longer.

Do you put your bet on 4-core PB=on versus 8-core PB=off in 1 minute bullet games?

Hardware is assumed to be equal.

I'm referring to to this old thread.
http://www.talkchess.com/forum3/viewtop ... =7&t=48864
If A and B are the same engine, it seems that A's pondering will very often be successful in pre-searching the move eventually made by B. If so, it's roughly like giving A twice the search time that B has. So I'm voting for A.
You are forgetting B is using twice as many cores. So A has to ponder CORRECTLY , just to stay even in productive CPU time . Every ponder that was incorrect , was inefficient use of CPU time.
pondering give less than being 2 times faster but doubling the number of cores also give less than being 2 times faster.

Uri Blass
Posts: 8557
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: How many elo do ponder equals to

Post by Uri Blass » Sun Jun 23, 2019 7:20 am

Nordlandia wrote:
Sat Jun 22, 2019 12:21 pm
Engine A ponder while engine B doesn't. Advantage +/- 50 elo. Doubling is worth more. Now the question is if it's likely engine A equalize engine B in fast time formats or who got the upper hand. The effectiveness of ponder is probably higher in shorter time controls than in longer.

Do you put your bet on 4-core PB=on versus 8-core PB=off in 1 minute bullet games?

Hardware is assumed to be equal.

I'm referring to to this old thread.
http://www.talkchess.com/forum3/viewtop ... =7&t=48864
I think that results can be different for different engines.

not all engines have the same parallel search and some engines may have time management that is only optimized for ponder on games and not for ponder off games or the opposite.

I can add that in 1 minute game the question is also how much time the side that ponder need to react to the move of the opponent(the side that does not ponder does not need to stop caculating when it get a move so it can respond faster).

if the engine that ponder check for the moves of the opponent only once per 0.1 seconds it may be no significant disadvantage for long time control games but a significant disadvantage for 1 minute games.

User avatar
Nordlandia
Posts: 2410
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: How many elo do ponder equals to

Post by Nordlandia » Sun Jun 23, 2019 8:19 am

Another workaround is to disable the ponder fuction internally in stockfish, and rename it as "SF_Ponder_OFF" inside GUI. Any help will be appreciated.

HGM: which winboard version do you recommend?

User avatar
hgm
Posts: 23629
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: How many elo do ponder equals to

Post by hgm » Sun Jun 23, 2019 6:15 pm

Nordlandia wrote:
Sun Jun 23, 2019 8:19 am
Another workaround is to disable the ponder fuction internally in stockfish, and rename it as "SF_Ponder_OFF" inside GUI. Any help will be appreciated.
Sure, but that requires programming and compiling.
HGM: which winboard version do you recommend?
4.8.0.

Just install the engine you don't want to ponder with 'Special WinBoard options':

/firstComputerString="easy\n"

That would overrule any ponder setting of the GUI. For what you want you would have to overrule the CPUs setting as well, as this is also treated as a common setting in WinBoard, without special provision for 'core odds'.

/firstComputerString="easy\ncores 6\n"

and then run the match with max CPUs = 3 and ponder on.

User avatar
Nordlandia
Posts: 2410
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: How many elo do ponder equals to

Post by Nordlandia » Sun Jun 23, 2019 7:03 pm

hgm wrote:
Sun Jun 23, 2019 6:15 pm
Nordlandia wrote:
Sun Jun 23, 2019 8:19 am
Another workaround is to disable the ponder fuction internally in stockfish, and rename it as "SF_Ponder_OFF" inside GUI. Any help will be appreciated.
Sure, but that requires programming and compiling.
HGM: which winboard version do you recommend?
4.8.0.

Just install the engine you don't want to ponder with 'Special WinBoard options':

/firstComputerString="easy\n"

That would overrule any ponder setting of the GUI. For what you want you would have to overrule the CPUs setting as well, as this is also treated as a common setting in WinBoard, without special provision for 'core odds'.

/firstComputerString="easy\ncores 6\n"

and then run the match with max CPUs = 3 and ponder on.
[ /firstcomputerstring "" ] worked to disable ponder for one side. I got 8-cores so so it seems i can test 2 versus 4.

MikeB
Posts: 3393
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: How many elo do ponder equals to

Post by MikeB » Sun Jun 23, 2019 7:17 pm

Uri Blass wrote:
Sun Jun 23, 2019 7:14 am
MikeB wrote:
Sat Jun 22, 2019 10:57 pm
zullil wrote:
Sat Jun 22, 2019 12:44 pm
Nordlandia wrote:
Sat Jun 22, 2019 12:21 pm
Engine A ponder while engine B doesn't. Advantage +/- 50 elo. Doubling is worth more. Now the question is if it's likely engine A equalize engine B in fast time formats or who got the upper hand. The effectiveness of ponder is probably higher in shorter time controls than in longer.

Do you put your bet on 4-core PB=on versus 8-core PB=off in 1 minute bullet games?

Hardware is assumed to be equal.

I'm referring to to this old thread.
http://www.talkchess.com/forum3/viewtop ... =7&t=48864
If A and B are the same engine, it seems that A's pondering will very often be successful in pre-searching the move eventually made by B. If so, it's roughly like giving A twice the search time that B has. So I'm voting for A.
You are forgetting B is using twice as many cores. So A has to ponder CORRECTLY , just to stay even in productive CPU time . Every ponder that was incorrect , was inefficient use of CPU time.
pondering give less than being 2 times faster but doubling the number of cores also give less than being 2 times faster.
Of course - but doubling of cores is guaranteed to search the correct move , pondering is hit and miss. So doubling is virtually guaranteed to be stronger than 1/2 the cores with pondering.

Uri Blass
Posts: 8557
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: How many elo do ponder equals to

Post by Uri Blass » Mon Jun 24, 2019 8:04 pm

MikeB wrote:
Sun Jun 23, 2019 7:17 pm
Uri Blass wrote:
Sun Jun 23, 2019 7:14 am
MikeB wrote:
Sat Jun 22, 2019 10:57 pm
zullil wrote:
Sat Jun 22, 2019 12:44 pm
Nordlandia wrote:
Sat Jun 22, 2019 12:21 pm
Engine A ponder while engine B doesn't. Advantage +/- 50 elo. Doubling is worth more. Now the question is if it's likely engine A equalize engine B in fast time formats or who got the upper hand. The effectiveness of ponder is probably higher in shorter time controls than in longer.

Do you put your bet on 4-core PB=on versus 8-core PB=off in 1 minute bullet games?

Hardware is assumed to be equal.

I'm referring to to this old thread.
http://www.talkchess.com/forum3/viewtop ... =7&t=48864
If A and B are the same engine, it seems that A's pondering will very often be successful in pre-searching the move eventually made by B. If so, it's roughly like giving A twice the search time that B has. So I'm voting for A.
You are forgetting B is using twice as many cores. So A has to ponder CORRECTLY , just to stay even in productive CPU time . Every ponder that was incorrect , was inefficient use of CPU time.
pondering give less than being 2 times faster but doubling the number of cores also give less than being 2 times faster.
Of course - but doubling of cores is guaranteed to search the correct move , pondering is hit and miss. So doubling is virtually guaranteed to be stronger than 1/2 the cores with pondering.
I do not see a reason that doubling the number of cores has to be stronger and I think that it is dependent on the number of cores that you start from and the smp implementation.

I totally agree that pondering is less than being twice faster in average but it is not clear that doubling the number of cores even gives an advantage.

There may be some n when 2^(n+1) cores are not better than 2^n cores when I cannot see how 2^n cores with pondering are not better than 2^n cores without pondering for some n.

Post Reply