Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Gravy for the brain that supports a 2500+ elo standard for computer GM's

Author: Robert Hyatt

Date: 05:57:11 06/19/01

Go up one level in this thread


On June 19, 2001 at 05:44:07, Mark Young wrote:

>On June 19, 2001 at 05:15:30, Vine Smith wrote:
>
>>On June 19, 2001 at 03:49:07, Mark Young wrote:
>>
>>>On June 19, 2001 at 00:31:40, Robert Hyatt wrote:
>>>
>>>>On June 18, 2001 at 18:01:49, Chris Carson wrote:
>>>>
>>>>>Nice summary Mark.  Calculate the overall average for GM's, I did it last year
>>>>>and it was around 2525.   Programs are performaing higher than that and
>>>>>performance is the only thing that matters in my opinion.  :)
>>>>>
>>>>>I for one do not care about a title, but the performance is the key.  :)
>>>>>
>>>>>Best Regards,
>>>>>Chris Carson
>>>>
>>>>
>>>>Several points:
>>>>
>>>>1.  anyone check the _age_ of each GM being 'averaged' into this mess?  IE
>>>>a 75 year old GM is still a GM, but won't have much of a rating.
>>>>
>>>>2.  anyhone check to make sure ratings were current?  IE Fischer's rating is
>>>>not particularly interesting since it is 30 years old and from a different
>>>>rating era.
>>>>
>>>>The main problem is probably age.  Just like the "baby-boomers" are threatening
>>>>retirement systems around the world, they are also moving up into the "older-GM"
>>>>group as well.  If you have more old GM players, then you will have a lower
>>>>overall GM average.  Which means exactly nothing of course.
>>>
>>>This is data from the latest Fide list. If you think the data is flawed....Old
>>>Grandmasters, Out of date Grandmasters, mistakes in calculations, or just fraud
>>>etc. Then make your case with data not theory. It's easy to bluster, it much
>>>harder to do the work to prove the case as I and others are doing.
>>Gee, I thought it was "harder to do the work" of programming computers to play
>>chess. But now I see that Prof. Hyatt should just give up improving Crafty's
>>code, and instead compile masses of statistics to prove that its rating is 2500,
>>2600 or whatever. New versions of Crafty will consist of countless pages of ever
>>more refined calculations always resulting in higher Elo ratings.
>
>Then I suggest he keep working on crafty, and stop making false and misleading
>claims about what a grandmasters is, Since Dr. Hyatt does not have the time to
>find out and to check his facts.


At least "he" doesn't try to first define the result he wants, then go about
finding ways to massage the data to produce that result.  "A statistician can
prove anything he wants, given enough time."

Use all the old GM ratings.  Use 1970 ratings mixed with 2000 ratings.  Use
players that are 25 along with players that are 80.  Use the ones that play
multiple events a year and the ones that play one event every several years.
Use the average IQ of a WWF wrestler.  And the weight of the average Japanese
wrestler.

Should be able to prove your point quite handily with all of that.

Meanwhile, for the last 25 years, a GM had to play at the 2600 level for 20-30
games in order to get the GM title.  Nothing else to be said there.  A few
"honorary" exceptions, but most of the exceptions, when you look at them, would
have been unnecessary since to win an interzonal (for example) you had to win
a zonal too, and both would probably give you that 2600+ performance _anyway_
at least the interzonal certainly would.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.