From: "Jeroen J.-W. Tiggelman" <CRMBJTI <at> HDETUD2 <dot> TUDELFT <dot> NL>
Date: 15 jul 1996
Subject: Re: "Theoretical" Elo ratings

> To further add on - It would be natural to assume that the average Elo would
> be lower than a 'rank standard' because players spend more time in a decline
> phase than a learning phase.  It may take 5 years for an amateur, 15 years

In addition, the respective lengths of the learning and decline phase
is irrelvant, since during the learning phase we may assume people to
be bjust about _at_ their grade, so that the presence of a decline
phase of any significant length would have this effect. :-)

> fully.  I'm sure many masters play well below their actual ranking in their
> later years.

However, Albrecht pointed out that it was a systematical error, and I
find it doubful that people who never reached a high grade would be
still playing tournaments in a decline phase..

> Are Elo ratings actually used to determine these rankings?  If promotions are
> based mainly on performance against similar or higher grades, do the
> committees separate the 'strong' 5 Dan from the 'weakening' 5 Dan?

If the population is large enough, the "behaviour" of playing on in a
decline phase statistically stable, and the minimum number of games
necessary to base a promotion on large enough, it is statistically
unlikely that this would make any difference as long as the requirements
would be such that the system was converging at about top speed. :-)

> title.  But if a Yokozuna fades, I have heard that he would be pressured to
> retire, rather than have the title become devalued by poor performances.  Is

I think the key here is that this _has been_ custom, therefore a
change would affect the population make-up and _hence_ influence the
grades (where in this case a devaluation could be expected). ;-)

Jeroen Tiggelman
crmbjti <at> hdetud2 <dot> tudelft <dot> nl