New Grades

Request amendments or pass comments on this free service.
Sean Hewitt

Re: New Grades

Post by Sean Hewitt » Wed Sep 03, 2008 10:04 am

Roger de Coverly wrote: The junior adjustments seem to be factoring in improvements which haven't happened yet.
Hasn't that always been the purpose of the junior adjustment?! Calculate Little Johnnys performance, and then stick a junior increment on to reflect the fact that he is likely get better over the next 12 months. Its used to be a flat 10 points for all juniors, then it became gradiated according to age. Now, with accurate data available they have been able to calculate what the increment should actually be, rather than just stick a finger in the air. I haven't seen the calculations, but I have no reason to doubt that they are correct.
Roger de Coverly wrote:
If I've understood the process correctly, parallel grades have been calculated for 2008 (published) , 2007 and 2006 (unpublished). The sceptics need to see the results for 2006 and 2007 to demonstrate that rampant inflation has not been permanently injected into the system.
I dont know whether the ECF will publish this data or not. What I do know is that it wont help you one jot in determining if rampant inflation has been introduced. Simply having these parallel grades on their own wouldnt allow anyone to prove inflation, deflation, or anything else. You would need far more information than that, including all of the individuall results for a start.

Given that you dont even believe that deflation exists, despite the weight of statistical evidence, it's difficult to believe that you would be be able to examine the evidence objectively.

Paul Stimpson
Posts: 109
Joined: Mon Feb 25, 2008 10:52 pm
Location: Essex

Re: New Grades

Post by Paul Stimpson » Wed Sep 03, 2008 10:23 am

Secondly the method of iteration used seems to be entirely scientific and thoroughly justified.
Hmmm, Not sure that's true. For instance what was the starting grade used for the iteration process? Also at what grade did the Juniors enter the process in at?

If you do a detailed look at the list some odd characteristics appear. If you rank all players in terms of losses and gains in the grading process the gains go as high as 78 points but nobody loses more than 9 points (this looks cooked to me)

However it's the junior grades that to me show something has gone horribly wrong in the process, probably an oversight, something silly might have been overlooked.

It appears (to me) that the junior grades have been corrected in some way to eliminate negative grades mainly. Then the grading process run from 2006. The problem I think is that junior players play at lot of games against their own age groups and this caused the really good junior players to leap their grades up massively, unfortunately I think this is not in context with the whole grading population and hence groups of Juniors at each age bracket are now (in my opinion) graded too highly.

The knock on effect of this would be that if you were lucky enough to have played junior players then you will be a high gainer in grading points. It would be interesting if the ECF could run a report plotting Adult grade increase against Number of Juniors played?

If this is indeed the problem then I think the best solution would be to have all Junior only events graded together in a Junior Grading List and exclude them from the main Adult List, Juniors could still appear in the main list but only by entering events with adult players present.

Paul Stimpson
Posts: 109
Joined: Mon Feb 25, 2008 10:52 pm
Location: Essex

Re: New Grades

Post by Paul Stimpson » Wed Sep 03, 2008 10:44 am

Actually it might be possible to disprove or prove the junior over graded argument by running this years junior grading results against ONLY adult players for those who have played 10 or more games against adults. If their starting grade for 2007-8 fits then not many junior grades should decrease? If the majority of Junior grades decrease then they would have been overrated for the 2007-8 season. How many of the top 5 in each age group decrease in grade taking only games against adult players?

If there isn't grading losses for the majority of juniors during this process then it will prove these high Junior grades I am seeing are more accurate than I had assumed.

Sean Hewitt

Re: New Grades

Post by Sean Hewitt » Wed Sep 03, 2008 1:32 pm

Paul Stimpson wrote:If you rank all players in terms of losses and gains in the grading process the gains go as high as 78 points but nobody loses more than 9 points (this looks cooked to me)
If you accept the premise that the grading system was suffering from deflation, that the grading distribution had become stretched and that the new grades are an attempt to correct this problem, then it should be no surprise that everyones grade goes up as a result of the correction.

In fact, I am surprised that anyone has gone down by 9 points!!

User avatar
John Upham
Posts: 7179
Joined: Wed Apr 04, 2007 10:29 am
Location: Cove, Hampshire, England.

Re: New Grades

Post by John Upham » Wed Sep 03, 2008 1:45 pm

How does the algorithm discriminate between non-improving juniors and improving ones?

I've noticed a number whose rating had remained low for the last 2 - 3 seasons and static (because they are weak and not getting any better) who have climbed above adults who consistently beat them (on the chess board).

Is the algorithm saying that these non-improving juniors have improved because their rating has increased?

Is there somewhere in the code a line somewhat similar to

new_rating = old_rating + ((18 - age_in_years) * unpublished_fudge_factor * RND(1.0))**(postcode) ?

Obviously, the variable names have been changed to protect their privacy.
British Chess News : britishchessnews.com
Twitter: @BritishChess
Facebook: facebook.com/groups/britishchess :D

carstenpedersen
Posts: 48
Joined: Fri May 23, 2008 10:20 am

Re: New Grades

Post by carstenpedersen » Wed Sep 03, 2008 3:26 pm

Is there somewhere in the code a line somewhat similar to

new_rating = old_rating + ((18 - age_in_years) * unpublished_fudge_factor * RND(1.0))**(postcode) ?
It does say in the new grades explanation that the average adjustment is "grade * 0.79 +45" for adults and "grade * 0.79 + 64" for Juniors.

Presumably this would mean that if a junior and an adult with identical grades in 2007/08 played against exactly the same opponents last year (say in a scheveningen tournament) and achieved exactly the same results then the junior's "new" grade this year would be 19 points higher than that of the adult.

User avatar
Dean Madden
Posts: 39
Joined: Fri Apr 20, 2007 1:06 pm
Location: Birmingham

Re: New Grades

Post by Dean Madden » Thu Sep 04, 2008 11:16 am

Going forward, are we still going to be using the +50/-50 formula to calculate next seasons grade? Presumably this is what has introduced the deflation in the first place? Or is the process used to calculate the 'new' grades going to be repeated every year?

I'm happy the grading system is being overhauled, but am concerned with the confusion it's going to cause, and don't understand that if it's being changed then why not go straight to an ELO system?

Tim Spanton
Posts: 1205
Joined: Thu May 01, 2008 11:35 am

Re: New Grades

Post by Tim Spanton » Sun Sep 07, 2008 5:51 pm

Is it really true that the average grade has gone from the long-established approximate mark of 110 (old style) to 130 (new style)?
Now that's what I call a cure for deflation ...

User avatar
JustinHorton
Posts: 10364
Joined: Mon Aug 04, 2008 10:06 am
Location: Somewhere you're not

Re: New Grades

Post by JustinHorton » Wed Sep 10, 2008 5:42 pm

I think mine's only going up by nine. I demand a recount!
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com