Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: ELO and size of the population

Author: Robert Hyatt

Date: 09:59:25 10/08/03

Go up one level in this thread


On October 08, 2003 at 12:24:23, Francesco Di Tolla wrote:

>On October 08, 2003 at 09:40:07, Robert Hyatt wrote:
>>
>>Size isn't so important.  Content is.  You won't have the same players,
>>so the ratings won't be comparable.
>
>I understand what you mean, but I was not referring to SSDF vs. FIDE.
>
>I know that there is some overlap beteen say FICS and ICC, and that people have
>a lower FICS ELO on average, similarly happend in Italian ELO vs. FIDE ELO.
>
>Has this any relation to the size of the population? (Please note the here the
>assumpion that the ELO system are the same is not necessarely true).
>
>Was just an arguing about the math behind elo system. (I did some simulations
>with a smll code, but the result depend strongly on the rule that determines who
>wins the game in the simulation, and it is far to easy to "write" the result you
>want with this rule).
>
>regards
>Franz


I think a population of "2" works fine.  The Elo of the two players will
definitely predict the outcome of the next game with reasonable accuracy.

When you have a large population, you end up with lots of "sub-populations"
with slight rating differences.  IE most of the USCF players in Birmingham
AL will never compete with players from Los Angeles.  There might be some
crossover in that LA players might play Houston TX players and they might
the Birmingham players.  But the larger the population, the more likely
this skewing happens.  With small populations it is likely everyone plays
everyone (my 2 for example).



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.