Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rebel's Phony Elo settings

Author: Dave Gomboc

Date: 23:12:17 04/03/99

Go up one level in this thread


On April 03, 1999 at 21:57:06, odell hall wrote:

>Hi CCC
>
>
> I set rebel 10's elo at my uscf rating (1810) to see just how realistic the
>rating are.  I played a six game match at game/5.  I was soundly crushed every
>game!!. Do not believe for a minute that the elo settings for rebel are
>accurate. Rebel plays much stronger than the elo might suggest. I believe with
>programs getting as strong as they are that a reasonable goal for programmers
>would be to develop realistic elo settings for their programs.

When I beta-tested Rebel 8 (and 9) I found that the elo function worked quite
well!  And I tested using 5 minute blitz games.  Some others have already
mentioned that USCF ratings are seriously inflated compared to other parts of
the world, maybe as much as 200 points.  (They're not alone: Canada's are
getting to be just as bad.)  Also, you probably should take it off of the
tournament book, so you're likely to have a better chance to get an edge in the
opening if you're the theory type.

There is a certain limit below which it doesn't go... it might be 1400, it might
be 1600, I don't know exactly.  The lower limit might even depend on the
hardware you're running on.  I propose that you do an experiment: set it to
1400, see if you start beating it up, or if it is still crushing you.  If you
are still getting crushed then it serves you right for having such a fast
computer. <grin>  (Actually, that was just a joke: I'd expect that the program
compensates for the machine speed.)

Maybe I will go play Rebel 9 a couple of games here with the elo function
dropped and see what happens...

Dave Gomboc



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.