How do yiu determine ELO by seconds?

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
Post Reply
Peeter
Member
Posts: 23
Joined: Fri Dec 01, 2017 4:46 pm

How do yiu determine ELO by seconds?

Post by Peeter »

Hi all
I have acquired a few boards now and having great fun. I have a Saitek Challenger, Master and Fidelity excellence. What i would like to know is how seconds computed relate to overall ELO? They go from 1 second to minutes. But how does that relate to ELO? Is there a way to roughly calculate this?

Thanks Peter
CompDad
Member
Posts: 30
Joined: Wed Feb 08, 2017 9:59 pm

Post by CompDad »

Weiki ELO is a 30s per movement or 30 min total. I supose if you give more time for a computer, the better it will compute, so a higher ELO, BUT I'm not so sure about it, maybe Nick can correct me.

For example, the Mephisto Master has at about 2100 ELO wiki.

The Mephisto Milano can be adjusted for any level between 0 to 2000 ELO.

I don't know for others computers, so the Wiki ELO is just a reference.
Peeter
Member
Posts: 23
Joined: Fri Dec 01, 2017 4:46 pm

Post by Peeter »

Cheers. I more or less know the max but on these three you cant adjust by elo. So the Master would be 2100 or so at 30 seconds thinking time?. But how do i know how many seconds is 1800(my level roughly)? Is it a sliding scale of some sort?. Thanks again.
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Peeter wrote:Cheers. I more or less know the max but on these three you cant adjust by elo. So the Master would be 2100 or so at 30 seconds thinking time?. But how do i know how many seconds is 1800(my level roughly)? Is it a sliding scale of some sort?. Thanks again.
Hi Peeter,

There is nothing scientific really. What happens is the more time you give the computer to think the deeper it searches. Therefore you might try and play it at 5 seconds per move first and see how you do. Master also thinks while you think (ponder) therefore if Master is still difficult to beat then turn Ponder off.

If you beat it at 5 seconds then increase it to 10 seconds for the next game and so on.

Regards
Nick
CompDad
Member
Posts: 30
Joined: Wed Feb 08, 2017 9:59 pm

Post by CompDad »

Peeter wrote:Cheers. I more or less know the max but on these three you cant adjust by elo. So the Master would be 2100 or so at 30 seconds thinking time?. But how do i know how many seconds is 1800(my level roughly)? Is it a sliding scale of some sort?. Thanks again.
Basically, yes.

Try 20s mov. The idea is with less time to think, the computer plays worse.

Another thing, the Master Chess has a power save mode. If I recall correctly, in PS 0 it has full power. I think the default is PS2, so it is a little bit handicapped, e.g., no ponder.

The best thing is try.

Good luck
Peeter
Member
Posts: 23
Joined: Fri Dec 01, 2017 4:46 pm

Post by Peeter »

Thanks guys. I will try the levels one by one till i find my level. I guess thats half the fun :).
kaseldop
Member
Posts: 10
Joined: Fri Dec 18, 2015 11:34 pm

Post by kaseldop »

It was often supposed that a doubling of speed gave program a 80elo increase. But I seem to remember Selective Search magazine brought this down to 50-60 as programs got stronger. Also the increases applied to computer verses computer, slightly less against humans.
There was a computer that was entered a Tournement that was stuck on 5 seconds a move
Instead of full time allocation and only played 250 elo lower than normal.
So most computers (dedicated any way) will probably play 250-300 below max when on blitz settings.
Never exact science as different programs would improve more than others with extra time or faster/More memory hardware.
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

The first column in the wiki is just a reference about how strong computers are compared to each other (at 30 secs/move average). But the elo listed, fits generally speaking the real elo. The real elo however, is only reached at tournament level, e.g. 40 moves in 2 hours or 3 minutes per move average.

So if kaseldop is right (it is generally accepted), one could also compute the other way round: half the thinking time = half the speed, so 90 secs average is about 50-60 elo less, and 45 secs average is another 50-60 elo-points less.
It has a lot to do with how deep the program computes, one halfmove deeper gives a clear raise in playing strength.

So could it be reasonable to subtract about 100-120 from the active elo to get an estimation of the real elo at 30-45 secs/move average? And halving further down you could subtract about 200-300 elo at 5 secs/move.
It also makes a difference if a program uses a brute force or a selective algorithm (and other special techniques), but I'm afraid that is another story... ;-)

Regards, Paul
2024 Special thread: viewtopic.php?f=3&t=12741
2024 Special results and standings: https://schaakcomputers.nl/paul_w/Tourn ... 25_06.html
If I am mistaken, it must be caused by a horizon effect...
Post Reply