English Chess Forum

Just a new thread splitting from the discussion approximately here.

There seem to me to be many good reasons in favour of ELO systems: It brings the UK into line with the rest of the world and it is statistically closer to reality, for a start.

At the risk of being a pain; maybe this should be in the "Grading Debate" section?

Richard James wrote:http://www.ecforum.org.uk/viewtopic.php?f=4&t=249

This is like one of those horrible items on Slashdot that links to a link to a link to a link ...

There's been stuff about this on the Grading section.

For example
http://www.ecforum.org.uk/viewtopic.php ... 7&start=21
is an outline of how you might get from here to there.

Is this form of filibustering on an internet forum by repeating a topic from the past?!

Alex Holowczak wrote:At the risk of being a pain; maybe this should be in the "Grading Debate" section?

At least i managed to move it out of ECF matters...

And i guess moving an entire thread is probably easier than splitting one?

Rob Thompson wrote:Just a new thread splitting from the discussion approximately here.

There seem to me to be many good reasons in favour of ELO systems: It brings the UK into line with the rest of the world and it is statistically closer to reality, for a start.

This is easy to say, and probably easy to prove - but it ain't always that easy. My experience is that other nations that use Elo-based ratings can be inflated. USCF and Spanish Elo, for instance.

What is this 'reality' you speak of?! Can you define it a little more specifically?

Yes a lot of nation's seem to be inflated in ELO terms. Just get Bogdan Lalic's view on British players Fide ratings, and that will answer your problem.

matt_ward wrote:Yes a lot of nation's seem to be inflated in ELO terms. Just get Bogdan Lalic's view on British players Fide ratings, and that will answer your problem.

Perhaps his problems may be at least partially solved if he ever tested players 200+ points below him.

Alex Holowczak wrote:At the risk of being a pain; maybe this should be in the "Grading Debate" section?

It now is

Rob Thompson wrote:There seem to me to be many good reasons in favour of ELO systems: It brings the UK into line with the rest of the world and it is statistically closer to reality, for a start.

The underlying distributions are the same really, the ECF approach is just a linear approximation to the Elo approach. There are two ways of doing performance ratings in Elo. One is to back solve for the rating which would give the actual score according to the tables, the other is to use the "rule of 400" which is to add the total ratings, add 400 times the excess of wins over losses and divide by the game count.

So the relationship between the systems is that the ECF uses the rule of 400 to obtain a performance rating which is then used as an estimate of strength for the next period. The Elo approach is that once the initial estimate of strength is established, you then adjust it using the k factor applied to the difference between the actual and expected scores.

An Elo method works much better if you want fairly frequent updates of the published ratings.

I don't think either works that well as a result predictor especially when you get into areas where one player is expected to score 85% or more against the standard of player represented by the opponent. The underlying theory may well only tell you that if A has a higher rating than B, that means that against the same opposition we expect A to score more points than B and this is accurate to about 8 ECF points or 60 Elo points.

I think the best argument is, if we had an ELO rating system, we would not have to join the ECF.

Roger de Coverly wrote:An Elo method works much better if you want fairly frequent updates of the published ratings.y

I agree with Roger. I think Elo's adaptability to more frequent lists is a major plus.

I also think it handles the issue of juniors better. k can be adjusted for young players so their rating catches up faster. (FIDE Elo relatively weak in this area. But only a minor tweak). The ECF system's arbitrary age bonus is a bit dubious, Most juniors improve rapidly for 2 or 3 years rather than steadily.

Simon Dixon wrote:I think the best argument is, if we had an ELO rating system, we would not have to join the ECF.

The international Elo system is controlled by FIDE, but access to it is always via the national federation, in this case the ECF. So you are subject to whatever rules the ECF cares to invent. In theory you can overturn these rules through the ECF Council but is practice it's difficult. There's also a quality test whereby FIDE don't, in general, rate games where the longest possible game length is under four hours.

Anyone with the necessary data and processing skills can set up a private grading or rating system. This can be ECF style, Elo style or any experimental system. The ratings produced will or should be internally consistent but may or may not bear any resemblance to more established systems.