How much time handicap can LCzero give Magnus Carlsen ?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Chessqueen
Posts: 622
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

How much time handicap can LCzero give Magnus Carlsen ?

Post by Chessqueen » Sat Aug 31, 2019 3:25 am

Is it possibly that LCZero could play against Magnus Carlsen with only 1%. It can play versus other top engines with only 30%.
https://tcec.chessdom.com/live.html?amp&amp&amp

lkaufman
Posts: 3722
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by lkaufman » Sat Aug 31, 2019 5:47 am

Chessqueen wrote:
Sat Aug 31, 2019 3:25 am
Is it possibly that LCZero could play against Magnus Carlsen with only 1%. It can play versus other top engines with only 30%.
https://tcec.chessdom.com/live.html?amp&amp&amp
I have some (slightly) relevant data. I have the Revelation chessboard, which has Komodo 12.1 built in, but running on a processor that typically gives it just 85,000 or so NPS. You can specify a percentage of CPU power to use. Apparently it doesn't work quite as stated, as the NPS numbers come out about double what they should, but you can allow for this. The lowest setting is 1%, which is really 2% or so, maybe 1700 nps on average, roughly 0.1% of what it would get on a typical single core of an i7. Although I am a GM, I find this setting to be roughly fair for games where we both average around 15" per move. LcZero on a 2080 with 10% time is probably at least as strong as Komodo 12.1 on single core i7, so at 1% it would probably be about the same strength as Revelation at about 170000 nps (twice its actual speed). It would be hard to find two GMs further apart in strength than I am from Magnus, but a hundred to 1 speed ratio is enormous, and I'm quite certain that Magnus would not win a match from Komodo 12.1 on the Revelation board at full strength at any plausible time limit, this is easily worked out from CCRL ratings with adjustment for speed. So the short answer to your question is yes, LcZero on a 2080 at 1% of normal speed would win a match from MC at any plausible time limit (let's say with at least 2" increment in case Lz0 needs .02 seconds or so to work properly).
Komodo rules!

User avatar
Nordlandia
Posts: 2439
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by Nordlandia » Sat Aug 31, 2019 6:26 am

Larry what about using 1 hyperthreaded thread instead of real core. As hyperthreaded is considered inferior. This is better for the human side. Crippling the engine further.

Raphexon
Posts: 91
Joined: Sun Mar 17, 2019 11:00 am
Full name: Henk Drost

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by Raphexon » Sat Aug 31, 2019 6:59 am

Probably like 2-5% at classical time control.

At something like Blitz or Bullet she needs less time but can become wonky at extremely short TC.

zullil
Posts: 5667
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by zullil » Sat Aug 31, 2019 9:56 am

Nordlandia wrote:
Sat Aug 31, 2019 6:26 am
Larry what about using 1 hyperthreaded thread instead of real core. As hyperthreaded is considered inferior. This is better for the human side. Crippling the engine further.
This suggestion doesn't make sense.

User avatar
Nordlandia
Posts: 2439
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by Nordlandia » Sat Aug 31, 2019 11:42 am

zullil wrote:
Sat Aug 31, 2019 9:56 am
Nordlandia wrote:
Sat Aug 31, 2019 6:26 am
Larry what about using 1 hyperthreaded thread instead of real core. As hyperthreaded is considered inferior. This is better for the human side. Crippling the engine further.
This suggestion doesn't make sense.
Based on what reasoning ?

A physical core is stronger than hyperthreaded.

zullil
Posts: 5667
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by zullil » Sat Aug 31, 2019 11:47 am

Nordlandia wrote:
Sat Aug 31, 2019 11:42 am
zullil wrote:
Sat Aug 31, 2019 9:56 am
Nordlandia wrote:
Sat Aug 31, 2019 6:26 am
Larry what about using 1 hyperthreaded thread instead of real core. As hyperthreaded is considered inferior. This is better for the human side. Crippling the engine further.
This suggestion doesn't make sense.
Based on what reasoning ?

A physical core is stronger than hyperthreaded.
All cores are physical. There are no "hyperthreaded cores". Unless you a using more threads than physical cores, each thread should---unless your OS is defective---run on its own core, with no need for "hyperthreading." So if you are using just 1 search thread, hyperthreading is irrelevant..

User avatar
Nordlandia
Posts: 2439
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by Nordlandia » Sat Aug 31, 2019 3:28 pm

There is program called Process Lasso that allow select affanity for background programs. So all you have to do is to let stockfish analyse a random position and open Process Lasso and let SF get hyperthreaded instead of real cores.

Active Processes -> double click on stockfish.exe and affanity then align from there.

zullil
Posts: 5667
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by zullil » Sat Aug 31, 2019 4:51 pm

Nordlandia wrote:
Sat Aug 31, 2019 3:28 pm
There is program called Process Lasso that allow select affanity for background programs. So all you have to do is to let stockfish analyse a random position and open Process Lasso and let SF get hyperthreaded instead of real cores.

Active Processes -> double click on stockfish.exe and affanity then align from there.
Yes, I suppose you could restrict Stockfish so that all threads are forced onto one or a small number of cores. But if you're using just one thread, as you suggested, at best this might strengthen Stockfish a bit.

There's really no such thing as a hyperthreaded core.

User avatar
Ovyron
Posts: 2693
Joined: Tue Jul 03, 2007 2:30 am

Re: How much time handicap can LCzero give Magnus Carlsen ?

Post by Ovyron » Mon Sep 02, 2019 8:08 am

Nordlandia wrote:
Sat Aug 31, 2019 6:26 am
Crippling the engine further.
What's the point of making the engine as strong as possible and then cripple it as much as you can? Wouldn't it be simpler to download some weak engine that already plays at around the level that you want? As hardware gets stronger you just need to get weaker and weaker engines in there.
Great spirits have always encountered violent opposition from mediocre minds.

Post Reply