Elo or Clarke national grading list?

[Alex Holowczak]

Yeah, that's why I said ELO, rather than FIDE. But the Clarke system and the ELO system do different things, so why ask people to replace one with the other?

Clarke was good at working out performance over a season. We seem to be moving away from that with semi-annual grading lists. So maybe a changing methodology is required?

[Sean Hewitt]

I do struggle with the concept that if 70% of people want to do A whilst 30% prefer B then some people think we should do B.

[Gavin Strachan]

Personally i don't care too greatly, but if there was one system then it would have more of a global significance (you can claim to only x number of points away from Anand) and also stops any disparities from occurring where people can have an ELO grade quite different to their ECF one. So therefore it would be great to just have one global system.

[Alex Holowczak]

Personally i don't care too greatly, but if there was one system then it would have more of a global significance (you can claim to only x number of points away from Anand) and also stops any disparities from occurring where people can have an ELO grade quite different to their ECF one. So therefore it would be great to just have one global system.

It would be two different systems. A national one, and an international one.

The international one will be done using the Elo system.

My question is how should the ECF do its national one: in the same way as Elo, or in the way it does now (Clarke)?

[Paolo Casaschi]

It would be two different systems. A national one, and an international one.

The international one will be done using the Elo system.

My question is how should the ECF do its national one: in the same way as Elo, or in the way it does now (Clarke)?

What some people here seem to think is that your national rating system rulebook should be: "go to the FIDE website and read the FIDE Elo for the player and use this as the national rating"

[Brian Valentine]

The poll's a bit misleading. Elo is a calculation method, mentioning FIDE muddies the waters. You would have both a national Elo and an international Elo which would probably differ.

Not necessarily, you could just drop the ECF rating and promote FIDE Elo rating. If the FIDE Elo rating covers the vast majority of chess players in the country, you dont need a national rating anymore. So the ECF rating could actually be replaced the the FIDE Elo rating.

Presumably it would be possible to predict whether grades would inflate or deflate if English chess was solely internationally rated. Anyone have a steer as to which way it would go?

The average grade on the ECF list has been declining slowly since the big adjustment. However its not clear this is the best measure of inflation. The evolution is very delicate depending on the proportions increasing their grade games (generally improving), those playing less game (generally declining) and new entrants (generally lower than average grades). Other things being equal I'd expect the average grade to slowly improve over the next few year.

An Elo system inflation result would depend on the treatment of juniors. The FIDE treatment of juniors is weak when their initial rating is set too soon. It's difficult telling what would happen as it depends on a large number of factors and the natural thing is to pick a few random ideas and ignore others. Using the FIDE algorithm we know the junior treatment would tend to be deflationary on entrance, but inflationary afterwards (low k factor fast improving from high k factor on average steady). People playing more games will improve more in a calendar period than declining activity players will reduce ratings. I would expect a FIDE implementation to be inflationary.

This all assumes the grading/rating team don't do something precipitious to solve the suspected re-emerging stretch.

[Roger de Coverly]

I do struggle with the concept that if 70% of people want to do A whilst 30% prefer B then some people think we should do B.

If B is a long standing arrangement and both the current and historic position, you don't overthrow it on a 51% vote.

Given that the regrading wrecked the historic continuity of the Clarke grades, the case for their retention is not in fact that great.

[Paul Cooksey]

I do struggle with the concept that if 70% of people want to do A whilst 30% prefer B then some people think we should do B.

If B is a long standing arrangement and both the current and historic position, you don't overthrow it on a 51% vote.

I agree with both Sean and Roger There is a big difference between 50.01% and 70%. Andrew Farthing's comment in another context he wanted broad consensus on major changes reassured me. I don't see 75% as a magic number, big changes need greater support even if 50% is enough constitutionally.

The Clarke system seems anachronistic to me. It was a brilliantly simple way to come up with a grade with the minimum of calculation. But given modern computer power, that doesn't seem a big advantage any more. I'm not familiar with any other arguments in its favour; is there a relevant thread?

[Roger de Coverly]

I'm not familiar with any other arguments in its favour; is there a relevant thread?

Many and varied.

From over three years ago

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

[Paul Cooksey]

I'm not familiar with any other arguments in its favour; is there a relevant thread?

Many and varied.

From over three years ago

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

Thanks Roger. I was familiar with that thread. I'll rephrase my question; leaving aside the practicalities of change and respect for tradition, is there any theoretical reason to prefer ECF? If you were a completely new federation for example.

[Alex Holowczak]

I agree with both Sean and Roger

Has that ever happened before?

[Roger de Coverly]

is there any theoretical reason to prefer ECF? If you were a completely new federation for example.

You would probably boot both systems into life the same way, by running recursions based on the algorithm of 400 and its ECF equivalent. So the start grades/ratings could be the same. Beyond that you have to decide to what extent you reflect past performances in the current rating. As Elo was designed in and for an environment in which most events were tournaments, then you use that if that's your system. Clarke systems perhaps perform better when you have a distinct league season with a gap between years and there's an interest in measuring performance over the season. A league only federation is unlikely, so some form of Elo style system is most probable.

If you are familiar with the international system there are some odd looking features in the national Welsh and Scots systems, presumably left overs from their conversion from a Clarke system

[Paul Cooksey]

As Elo was designed in and for an environment in which most events were tournaments, then you use that if that's your system. Clarke systems perhaps perform better when you have a distinct league season with a gap between years and there's an interest in measuring performance over the season.

Good points. So I am going to annoy Alex by asking why he didn't include a voting option for both?

Annual ECF grade to show overall performance last year. Monthly Elo to show current strength. Maintaining the ECF alongside the Elo would be minimal extra effort, since they can run off the same database. Of course asking the graders to submit monthly (as a standard format csv?) so I can have the Elo, is considerable extra effort.

[Alex Holowczak]

So I am going to annoy Alex by asking why he didn't include a voting option for both?

Because I didn't think of it.

[E Michael White]

It is sometimes suggested that switching to an Elo basis and increasing junior k-factors would solve the problems, as used by other chess federations. This is unlikely to be totally successful as in individual games the k-factor for the opponent, if a non-junior, should be reduced for that game. If this is not done a deflationary strand is left active. The old ECF system tackled this by giving age related increments but the implementation was somewhat inept and its efficiency not monitored.