English Chess Forum

The new junior grades are producing erratic results, particularly for those who don't play 30 games:
One example, an improving junior has just got a very dispiriting grade change of -30:

Category Standard Category Rapid
Jan 2010 ... ... ... ... B 111
July 2009 A 137... ... B 141
Jan 2009 ... ... ... ... B 85
July 2008 A ..79... ... C 73

Is there going to be a review of the new method of producing junior grades?

I see you have raised this via the feedback form Neill, so I would appreciate you also posting the follow up on that when you receive it...


I cannot comment on that one sorry!

_________________
Cheers
Carl Hibbard

It is not the sort of thing that the grading officer deals with. I should raise it directly with the Manager of Grading or the Home Director.

Ok...

_________________
Cheers
Carl Hibbard

A bit more analysis shows:
In 2008 list 40 adults and 4 juniors had their Rapidplay grade go down by more than 10 points
In 2009 list 57 adults and 89 juniors had their Rapidplay grade go down by more than 10 points

So no great change for the adults, but a dramatic increase in the number of juniors with significantly falling grades

Entirely to be expected as the system for juniors has weighted the balance between lag and randomness much more in the direction of randomness.

To amplify this, all grades are potentially wrong so if you get results which don't agree with the previous published grades, do you
(a) presume the results to be random fluctuations and make a small change to the previous grade
or
(b) presume the previous grade to be incorrect and make a large change to the previous grade.

or
(c) Forget about it and just play the game for fun?

Perhaps that is too idealistic.

_________________
February 16-17: 2013 BUCA Team Championship
April 20: 2013 British Blitz Championship
July 27-August 10: 100th British Chess Championships

Alex,
I'm playing for fun tomorrow, tonight I'm looking at this thorny issue.

Roger,
I think there is a deeper point in Neill's issue. Overall those juniors with grades for both periods have increased their sum of grades. However those going down are roughly in the ratio of 5 for every 4 increasing, on adults it's more like 2-1 . There is the feature in a milder form in the long play changes. The situation is very difficult to pin down because of the high number of players who only appear in one of the two lists and therefore may be changing the strength of the pool. However one measure of deflation might be that the average grade (sum of grades/# of players) is falling which it is with the junior rapid play grades for those involved.

There are other measures of deflation, but on the face of it based rapid play grading list, there is evidence that people playing rapid play are getting weaker.

I thought that general reasoning had established that the "new" method of grading juniors is unstable. Is this perhaps a correction factor from the 2009 grades being too compressed (or over inflated).

My guess is that the grade change now is highly dependent on the strength of the opposition played for juniors (particulary those playing less than 30 games in a season) and the weaker the tournament they play in relative to their grade the more likely their grade is to fall as they in effect lose the 40 point protection from their previous grade. In rapidplay games they are more likely to produce a slip than in standard games and this has a snowball effect dragging down their grade but also the average grade for the year on which the 40 points rule is calculated.

If a player with a grade of 160 plays one tournament all against players of 190 say and loses all 6 games and 3 tournaments against players graded 90 say and wins all 18 games I think their grade as a new player under the new system would be less than 140 or 140?

Under the old system it would be ((170 x18) + (140x6))=163

Although the above is an extreme example it is my interpretation of what is happening.

Roger,
I agree the process is unstable, especially when juniors may even be in the majority (depending on the influence of entrants and leavers. However I think Neill is pointing out another issue in that more players (that are measured) are declined than those who have grading increases. This is consistent with those playing most games being those who have rising grades.

This situation may or may not represent the reality. But it doesn't do much to enthuse the next generation of chessplayers.

My posting above is wrong forgot about the itterations - tired, but the really sad thing is that I don't really understand the new player process enough to work out what the grade would be in my example. Does anyone know? I know it is probably a stupid example but what is the answer anyone ?

I believe the answer is 163 (isn't it) as the process for calculating performance has not changed as far as I am aware. If the graded players were juniors then you would be unable to calculate the performance (and therefore the grade) at all.

If I've thought this through correctly a juniors grade under the iterative basis is made up of a performance bit plus a weighted average of previous grades of some graded players, not necessarily just the juniors own rated opponents. The weights will sum to 1 and the average at present appear to be something sensible like 100 or 170. However the performance related bit is not restricted to -50 -> +50 as it brings in the performance of opponents and performance of opponents of opponents etc. The performance bit may stack up by a series of +ve or -ve parts and produce very high or very low grades depending on who the juniors opponents are. At some stage a junior grade of +300 and substantial negative grades are not unlikely.

The main point I am making is under the previous system junior grades were stable and therefore could useful to indicate their progress. Junior grades are now unstable as not only is their old grade ignored, but for those who play less games only a maximum of 20 games are used not 30. The result is that some junior were either overgraded last time or undergraded this time, or both. Felix Ynojosa has dropped 12 points, Maria Wang 20 points.

[Edited to correct name]