Chess Grading Destruction

[Maurice Lawson]

Has anyone realised what a total mess has been created with the grading changes?
The new system has completely ruined what was a reasonably good system.
There was nothing much wrong with it - now there's nothing much right with it.
Gradually people will start realising. The system is now entirely illogical.
Says Mr Valentine...."There has also been some limited feedback on the "abolition" of X grades. (The category became defunct because of the change to category A). The issue is to whether 30 games over six months is picking up reliable changes in strength or extra "noise". The theory on stretch suggests that the system works better using twelve months worth of games."
.... In my opinion rubbish.
Grading should be based on a number of games played, not on a period of time. But anyway the system introduced simply creates multiple distortions.
There is vastly more "noise" in the new system than there ever was before. And it now most certainly does not give information on changes in strength. IT does exactly the opposite, namely hides changes in strength. Will be loved by all those who are declining, whilst it holds back those who are moving up.
The system used to be preferable in many ways to FIDE - now it has been vastly corrupted and is unfit for purpose.

[Joey Stewart]

I didnt realise they are planning a big change to the grades - what is going to happen to them?

[Roger de Coverly]

I didnt realise they are planning a big change to the grades - what is going to happen to them?

It's already happened.

Two changes.

If you play more than 30 games in a half year, you no longer get a grade based on six months only (category X). Instead it's always based on a whole year.

If you don't play 30 games in a year, the balance to make up 30 will be based on the average from the previous periods rather than counting games backwards.

The changes should stabilise the instability introduced by six monthly grading. When grades were published yearly, you got a 190 grade even if you played the first half of the season at 195 and the second half at 185. When they had the X grades, you could bounce from 195 to 185 and back again. You now get 190 again in both six month lists.

As with all grading and rating systems, what you publish and use in calculations is a compromise between reflecting recent results and underlying strength and experience.


If the grading system was wrecked it was during the totally unnecessary regrading exercise moving the average player up 20 to 25 points from 115-120 to 140.

[MartinCarpenter]

The effects of that regarding have, I suspect, finally fed through to all levels and a rough stability has returned. No idea if it achieved its formal aims or not

The other changes are of course just reverting to very long proven methods from before. The strict 30 game grading boundaries really did make things quite random.

[Brian Towers]

Roger, do you have a link to an online discussion / paper considering the relative strengths and weaknesses of BCF/ECF v Elo grading systems?

I'm curious as to whether the ECF has sound reasons other NIH and tradition for not moving towards its own version of Elo as many other countries have done. If they do exist then perhaps they should be publicized so that other federations and perhaps even FIDE itself could seriously consider moving towards an ECF grading system.

This next may be a rathole but I'm also curious as to whether in this modern, digital, online age the ECF has considered the exclusive online submission of results by league secretaries, club secretaries and event organizers. This might reduce administrative effort and costs. I see the occasional mention of the sterling work done by grading officers around the regions / counties and I presume this involves more work than just pressing the button to start the grading program running.

[Maurice Lawson]

Roger's example of what he calls bouncing is exactly why the smoothing is so bad. I want to know at what level I have been achieving over a reasonable number of (say for argument sake 30) games. (And I want to know the same for my opponents recent results!) If I have played at 195 then that's what I want to see, because that has been my playing strength during that set of games in the recent six months. The idea that I have actually been 190 all through that period is ridiculous! It is a statistician's response to the results rather than a competitor's. A year is far too long, and for an active player it is far too many games. It also uses vastly different numbers of games player by player, depending simply on how much they choose to play or not play. It is a mess, and an arbitrary mess!
This smoothing business is a severe red herring in the grading context. It has removed incentive from the system, because it over-emphasises smoothing. Grading is absolutely not about demonstrating consistency, and the change is a very bad one for chess and for incentive. The grading system should be about performance, measurement and incentive!

[JustinHorton]

How important is the difference between 190 and 195?

[Michael Farthing]

Maurice you are repeating these comments in many places without adding any significant reasoning - perhaps because at present you are feeling very angry about it. Brian Valentine consulted on this forum about the changes he was making and a number of constructive comments were made here that Brian took into account. The changes he made at that time were, by and large, received warmly and I for one felt they were a big improvement. With any system there will be differing points of view but if you want to make progress in changing points of view some detailed arguments and examples will be needed.

Your principle concern is regard to 'smoothing'. I agree with you that players who are rising or falling may have their changes delayed by the new system, but some quantification of the problem would be useful. I'm not convinced that for adult players (other, perhaps than those very newly in the system where the main problem is anyway likely to be the small number of games) the trend is being disguised by smoothing that eliminates noise variation. In the case of junior players the disguise of trend is a recognised problem that is explicitly allowed for in the current system by completely regrading juniors each time. It is possible that perhaps juniors should only have more recent games included but it would be useful to have some research to show at what number of games the removal of 'noise' by smoothing begins to disguise 'trend'. So far, you have not discussed that as an issue and if you were able to provide some evidence I think it very likely that Brian would listen attentively.

There are other serious anomalies that the recent changes have tried to address and you should not neglect these. One of them is that the actual dates that games are played are often not reported - for example, internal club games and some league results are simply submitted with a blanket date for the whole season. This makes the selection of games to include quite impossible. It might be that the grading team should insist on explicit dates, but this of course would mean a lot of chasing around by the local graders attempting to get dates out of clueless players saying, "I don't know, I didn't write the date on the scoresheet". Another difficulty is that the movement of a game between one period and the next can often have disproportionate effects on the grade given - a good principle in my view is that grades should not have significant variation as a result of such minor administrative arrangements.

I do not think anyone believes that the current system is not capable of improvement. Our grading team clearly give a lot of thought to the problems presented to them and their objectives are manifestly to produce a good system. I believe they do listen to genuine concerns, but please calm your anger and make a reasoned case.

[Maurice Lawson]

You are badly missing the point. First of all (in the arbitrary example given) it's the difference between 185 and 195, so it can be quite significant differences. But additionally there are a whole load of distortions introduced into the system (much more than you might first think of). Also it's about the anti-incentive aspects. If I spend a lot of time on my chess and make a 6 or 7 point improvement to reach a grading level I have never achieved before, then that is hugely important to me and to many others too.
If you don't think that a few points is important then you are missing the whole point of the grading system. For many the ECF grade has been the very most important measure their chess performance and status (especially given how slow the FIDE rating reacts). And anyway if you don't think five points is important then why would you want to smooth it? The performance and the incentive are far more relevant to competitors than is statistical smoothing.

[David Williams]

How important is the difference between 190 and 195?

Possibly two bad moves in thirty games.

[Maurice Lawson]

Michael, I have been making a reasoned case, but every response I get is mostly dependent only on the what people think are the "noise" aspects. Just because I am one of the few who realises the associated negative implications it doesn't mean that my reasoning is not valid. I am indeed in discussion with Brian, who is certainly open to the points I raise, and I will be continuing with that discussion. Yes I know there were other aspects involved, but the change is a sledge hammer to crack a nut. And if you have noticed that I am angry then you have to ask yourself why someone should become so angry about what you think is just a statistical change?

[Carl Hibbard]

How important is the difference between 190 and 195?

At 192E and inactive huge

[Mike Truran]

then you have to ask yourself why someone should become so angry about what you think is just a statistical change?

I accept I'm no mathematician, but isn't it just a statistical change?

[MartinCarpenter]

Basically statistical yes. People do get silly about things

As far as I can tell Elo is rather more conservative in terms of quick grading changes than even traditional ECF. Quite purposefully so imagine and for good reasons with very strong players.

If you've genuinely improved your chess and are playing a halfway meaningful number of games a season you'll see it soon enough on ECF! You just have to sustain it for a little while. Not long if you're playing 30+ a season.

On a statistical basis, the previous rule set basically contended that:
a) If you play n>30 games in a year, the first n-30 of those are entirely useless in terms of predicting your strength in the next season,
b) If you play n<30 games in a year, then the best way to make that up to 30 is to try to take your last 30-n games from the previous year rather than the average of all the games you played during that year.

Those assertions really are very hard to logically defend if you want the grade to be as good an estimate of next years strength as possible.
(Non trivial playing strength in(de)creases are fairly slow/rare, big/terrible performances in one event are common place.).

Yorkshire run a fully live grading list, which can make even the previous ECF list look stable!, but the grader likes it that way He views grades as a way of keeping people interested. Not how I'd do it but it isn't an insane attitude at an ameteur level.

[Maurice Lawson]

It is not "basically statistical". It is a structural change in the parameters which has shifted the whole purpose. It is now interested in smoothing, rather than being a major influence on players' incentives. I don't particularly rate the statistics of something which could be based on 30 games for one person and 120 games or even more for someone else. The mechanism has been put together by persons who have talked about one year being the most important parameter, instead of a number of games being most important.
There is room to argue (for example) that about 35 games, or 40 games, etc might be a more sensible figure (although I continue to favour 30) but there is no logic at all in suggesting that players choose their own number, which is what the present system does!
Of course the FIDE/ELO systems have parameters which imply an effective number in the history (without stating it explicitly) and this is at least logical. FIDE of course publishes monthly, recognising that players are extremely interested in how their rating progresses. FIDE clearly therefore recognise that players are interested in that incentive and FIDE do not therefore pretend that ratings stay the same even month to month - let alone for a whole year. And more recent games are implicitly given substantially higher weighting than older games, all of which is logical from a competitive point of view.
The ECF change is nothing more than the consequence of people focussing far too heavily on smoothing as opposed to sporting interest. For that reason it is a backward and bad change. The system has been rendered anodyne.