Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: New Poll - Now Taking Place..

Author: Howard Exner

Date: 23:19:27 04/22/98

Go up one level in this thread


On April 22, 1998 at 13:57:15, Djordje Vidanovic wrote:

>On April 22, 1998 at 13:01:35, Bruce Moreland wrote:
>
>>Points   Score
>>0           50%
>>100       60%
>>200       75%
>>300       85%
>>
>>I've rounded these a little, but they're within a percent or so.
>>
>>So, if you think DB is 300 points better than micro program X, you'd
>>think it would score at least +8 -1 =1 or +7 -0 =3.

>DB may beat the best micro by 8.5-1.5, or even, 9-1, or, 10-0, but this
>does NOT mean that its results against humans will be just as good, and
>300 rating points better.

I support this view also that comp vs comp strength is uniquely
different
than comp vs human strength. It is risky to extrapolate one to the
other.


>I'd venture to say that its hypothetical
>results vs humans should not exceed the best micro's results by more
>than 100-150 rating points.  Thus, we talk of two different things -- DB
>may exploit a micro (or chess software)'s weaknesses quite consistently,
>but this does not apply to humans who are always prone to mistakes,
>errors, etc. Thus, another 'guestimate': DB (approximately) 2700, best
>micro (computer software) on a Pentium II / AMD K6 400 Mhz 2550-2600.

There is so little data out there on comp vs human play at 40/2 (elo is
based on 40/2). Are there clues to estimating strength on seeing "X" ply
further than other programs? How many points is 10 ply vs 9 and is this
the same as 16 ply vs 15?




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.