Main
Date: 08 Jan 2009 04:34:02
From: milivella
Subject: Elo: multiple iterations?
Hi. I'm playing a bit applying the Elo method to football club teams:
http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/2fce2aeb5d0ecdbf/
Now, in this context there are many matches between teams of the same
nation, few matches between teams of different nations. Computing Elo
ratings in the normal way rewards teams that are strong in their
national league, even if they are not good at international level. So
I've tried to iterate the computation over and over (i.e. when you
end, re-start form the first match, but using the last ratings), and
the results were quite interesting: the ratings converge, and they
converge to a reasonable rating (Celtic Glasgow is no more in European
top 10!).

So, it seems that multiple iterations produce better ratings, in this
context. (OTOH, multiple iterations don't skew the recent-form nature
of Elo: e.g. if you invert the order of the matches - the first
becoming the last and vice versa -, you have quite different ratings.)

Now, my question is: what about multiple iterations in Chess Elo
ranking? Has it ever be tried or at least discussed?

Maybe the different nature of chess tournaments doesn't require to re-
iterate the computation.

--
Cheers
milivella




 
Date: 09 Jan 2009 01:17:20
From: Offramp
Subject: Re: Elo: multiple iterations?
On Jan 8, 2:16=A0pm, milivella <[email protected] > wrote:
> milivella:
>
> > Hi. I'm playing a bit applying the Elo method to football club teams:
>
> In case you're interested in the results of such experiments... here
> are the relevant links:http://groups.google.com/group/rec.sport.soccer/br=
owse_thread/thread/...http://groups.google.com/group/rec.sport.soccer/brows=
e_thread/thread/...
>
> --
> Cheers
> milivella

Wow! I am stronger than Chelsea!


 
Date: 08 Jan 2009 13:23:33
From: milivella
Subject: Re: Elo: multiple iterations?
[email protected]:

> I am unsure how to answer your question above - but I note XPR ratings
> offer more than a number - they also give black and white results
> [home and away?] and other stats actually useful to the chessplayer.
>
> They are pretty friendly, google them and ask their opinion? Use my
> name as a reference if you like - we joke around a bit, including such
> stuff as the effect of altitude on draw rates in chess.

Thanks. You are referring to
http://www.cxrchess.com/
don't you? I'll ask them.

--
Cheers
milivella


 
Date: 08 Jan 2009 09:04:09
From:
Subject: Re: Elo: multiple iterations?
On Jan 8, 9:16=A0am, milivella <[email protected] > wrote:
> milivella:
>
> > Hi. I'm playing a bit applying the Elo method to football club teams:
>
> In case you're interested in the results of such experiments... here
> are the relevant links:http://groups.google.com/group/rec.sport.soccer/br=
owse_thread/thread/...http://groups.google.com/group/rec.sport.soccer/brows=
e_thread/thread/...
>
> --
> Cheers
> milivella

I think the first use of Elo's idea outside of chess was in horse
racing in the US. The ex-Pres of USCF Don Shultz knows quite a bit
about that. More recently I see that a wrestler had a rating!

I am unsure how to answer your question above - but I note XPR ratings
offer more than a number - they also give black and white results
[home and away?] and other stats actually useful to the chessplayer.

They are pretty friendly, google them and ask their opinion? Use my
name as a reference if you like - we joke around a bit, including such
stuff as the effect of altitude on draw rates in chess.

Cordially, and good luck, Phil Innes


 
Date: 08 Jan 2009 06:16:35
From: milivella
Subject: Re: Elo: multiple iterations?
milivella:

> Hi. I'm playing a bit applying the Elo method to football club teams:

In case you're interested in the results of such experiments... here
are the relevant links:
http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/1ebb94e4931ddbcb/
http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/51b6ed10ed572317/

--
Cheers
milivella