Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF ratings versus USCF ratings

Author: Dann Corbit

Date: 12:21:50 07/02/01

Go up one level in this thread


On July 02, 2001 at 14:45:57, Mike S. wrote:
[snip]
>Rating inflation is a big issue I think... I do not understand much of it. I
>have no idea if the SSDF ratings may be effected by this too (the rating
>correction which was done, seems to indicate that).

It does not really matter much if there is ratings inflation.  You can't compare
old ratings [ten years ago, for instance] with new ones or ratings from one pool
of players against a different pool.  The difference in ELO will still be a good
guestimator of performance (in terms of broad averages).

If you want to create ratings inflation, it is very easy.  Just insert a bunch
of knuckleheads into the pool.  If you add a million players of 500 ELO, just
watch the meteoric rise of the good players.  And the better "500 ELO" players
will no longer be 500 ELO.  They will have a much higher number.  It would be a
funny, skewed distribution, and so I think the ELO model might even break down
because it probably assumed a gaussian distribution or some near approximation
of it. (IOW, it could no longer be used to predict even approximation by large
numbers accurately).




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.