Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The data isn't really very skewed

Author: Bruce Moreland

Date: 12:39:34 03/02/99

Go up one level in this thread



On March 02, 1999 at 12:09:51, KarinsDad wrote:

>On March 02, 1999 at 01:05:48, Bruce Moreland wrote:
>
>>I don't know what this does to what you say in your post, but the top 6 humans
>>on ICC have (combined) played a total of 11 blitz games against my program.
>>These guys are rated 2950 and up.
>
>What are the results of those 11 games?

They got 2.5 points.

The games were all 5 3,  5 2, or 4 4, with one 3 3.  The players were GM Dlugy,
"GMAlex", and IM Banikas Hristos.  Dlugy got 2.5 out of 7 games, 5 of which were
with white.  The other three in the top six, all GM's, haven't played as far as
I can tell.

What's my point?  I'm not sure.  I am pretty certain they'd all get killed if
they played any of the strong computers at the time controls they played against
humans.  Playing the slower time controls that GM's play against computers would
be better for them, but I am still dead certain they'd lose points.

If the contention here is that the rating system is accurate against the general
pool, that's totally not true if the computers have anything to do with the
general pool.

>Kasparov is considered to be the best chess player in the world. However, there
>are corresponence players who could look at his games and find mistakes. Deep
>Blue was able to take him apart. But that does not matter. His (mostly human)
>opponents cannot find his mistakes over the board, so he has the best standard
>over the board rating in the world. The same applies to the best ICC players at
>blitz. It does not matter that a computer can play the game better at these
>speeds. What does matter is that their rating reflects the results of their
>games (just like Kasparov's rating does).

I don't understand what is being contended here.  Of course the rating reflects
the results of their games, that's an a priori statement I think, it's like
saying that you are the height you are.

If you go onto ICC and type "best c" you'll get 9 computers and 11 humans in the
blitz list (my program isn't there, it's off the bottom these days, what's on
there is mainly a bunch of multiprocessor crafties, and a few people running
commercials manually).  If you do a round-robin of 10-game matches between
everyone on that list, at any rational blitz time control, the list would
scramble completely, and I bet the computers would go *up*.

bruce



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.