Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is there a rating inflation?

Author: Dann Corbit

Date: 12:02:18 06/03/02

Go up one level in this thread


On June 03, 2002 at 14:51:23, Sune Fischer wrote:

>On June 03, 2002 at 14:39:22, Sune Fischer wrote:
>
>
>>You have
>>Elo_1970(strength)=F(T_1970(strength)) and
>>Elo_2002(strength)=G(T_2002(strength)), now F(T(..)) and G(T(..)) are known
>>distributions, namely the ratinglists.
>>But we want to find how the strength evolved in time, how do we do that?
>>
>>If you treat F,G and T as unknowns (as I do), then you will get nowhere in you
>>analysis, you need to make assumtions or approksimations, that is unless I'm
>>overlooking something ;)
>
>Perhaps
>Elo_1970(strength)=F(S(1970),1970) and
>Elo_2002(strength)=G(S(2002),2002), would have been a better nameing
>convention;)
>So S is strength as function of time, and F and G are inflations, also functions
>of time.
>It seems hard to seperate S(time) from the other unknown functions, but perhaps
>statistics has methods?

For a method to be considered scientific, it must be repeatable and verifiable.
Can this be achieved if some of our data points are "dead"?

The experiment is much more complex than most people think (IMO).

The data points are not static -- the ability goes up and down over time.  Is it
a parabolic arc?  Does it have the same shape for everyone or is it radically
different for some?  Is there an accurate model that can predict the frequencey
of:
Wins White/Wins Black/Draws White/Draws Black
for every possible different ability difference?

I think we must admit that the ELO system is a crude but workable approximation
and the model that underlies it is a very coarse and highly oversimplified
approximation of reality.

All that having been said, I think Roy Jones' model is a little better but not a
lot.  And it is doubtful it will be adopted for the simple reason that the
current ELO system has so much interial mass and tradition behind it that it
would take a huge impact to move anything.

One thing (of my opinion) is that ELO calculations should not be made using
integer math.  This causes serious problems in accuracy at even small ELO
differences.  Yet some systems do this.



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.