From: Larry Kaufman <lkaufman <at> COMCAST <dot> NET>
Date: 26 oct 2002
Subject: Re: Elo system

----- Original Message -----
From: "Albrecht Heeffer" <albrecht.heeffer <at> PANDORA <dot> BE>
To: <SHOGI-L <at> TECHUNIX <dot> TECHNION <dot> AC <dot> IL>
Sent: Saturday, October 26, 2002 10:31 AM
Subject: Elo system


> Hello all,
>
> Chessbase.com features an article by statistician Jef Sonas with
> a proposal to improve the ELO rating system. He advocates a simpler linear
> formula to calculate the expectancies and argues that this provides
> a better predication than the current ELO system. I find his article
> quite convincing. The shogi world could be first to adopt the Sonas
> system.
> http://www.chessbase.com/newsdetail.asp?newsid=562
>
> Greetings,
>
> Albrecht Heeffer

     I am familiar with Sonas and his work, and while I do agree with some
of his ideas, I think it is obvious that a linear formula cannot be as good
as a curve like the logistic one we use.  He may obtain a better fit, simply
because the coefficients he chose make the linear formula fit better than
the logistic with its current coefficients.  But if the logistic curve were
also modified to pick the coefficient that best fits his data, it would
probably fit better.  Simply put, it is ridiculous to think that the
probability of defeating an opponent increases with increased rating
difference, and then suddenly stops increasing (either at 100% or at some
lower figure).  I'm sure Sonas knows this, but prefers the linear model
because it is mathematically much easier to work with and because his data
involves relatively few pairings with large rating differences.  That does
not apply to us.

     Larry Kaufman