Junior Grades

General discussions about ratings.
Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Junior Grades

Post by Roger de Coverly » Sat Jan 28, 2012 8:49 pm

Would anyone care to speculate how the grades for juniors in the new grading reports are derived?

They aren't the quoted July 2011 grades, nor do they seem to be the January 2012 grades either.

For instance at the Wycombe Congress, I played Maria Wang (15), James Holland (16) and Marcus Harvey (15). On my grading page their grades are shown as 173,204 and 205. On their profile pages their grades are shown as 175 (July)/177 (Jan), 204/208 and 212/209.

I played Phillip Archer-Lock (15) in a Berks cup match. He shows as 127 for me but 122/128 on the profile.

Junior increments come in somewhere, but I thought these varied by age.
Last edited by Roger de Coverly on Sat Jan 28, 2012 10:05 pm, edited 1 time in total.

User avatar
Christopher Kreuzer
Posts: 8806
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Junior Grades

Post by Christopher Kreuzer » Sat Jan 28, 2012 9:54 pm

No idea what is going on there, though I too noticed a couple of discrepancies between grades written on my scoresheets and the grades as they appeared online. Didn't think to check further than that, but it sounds like these were juniors and the same sort of thing that Roger has pointed out would be the case here as well.

Alex Holowczak
Posts: 9085
Joined: Sat May 30, 2009 5:18 pm
Location: Oldbury, Worcestershire
Contact:

Re: Junior Grades

Post by Alex Holowczak » Sat Jan 28, 2012 11:26 pm

Roger de Coverly wrote:Would anyone care to speculate how the grades for juniors in the new grading reports are derived?

They aren't the quoted July 2011 grades, nor do they seem to be the January 2012 grades either.

For instance at the Wycombe Congress, I played Maria Wang (15), James Holland (16) and Marcus Harvey (15). On my grading page their grades are shown as 173,204 and 205. On their profile pages their grades are shown as 175 (July)/177 (Jan), 204/208 and 212/209.

I played Phillip Archer-Lock (15) in a Berks cup match. He shows as 127 for me but 122/128 on the profile.

Junior increments come in somewhere, but I thought these varied by age.
Remember that juniors are treated as ungraded now when they are graded. Remember that grading is a two-step process:
(1) All ungraded players go through an iterative process to get an estimated grade
(2) Everyone's grade is calculated based on their previous grade, or the estimate provided by (1)

So, what you're seeing on your page for juniors and ungradeds, is the number associated with them after (1).

My opinion was that we should show the last published grade for everyone, because showing anything else would cause confusion. If necessary, leave it blank. I was outvoted. :)

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Junior Grades

Post by Roger de Coverly » Sun Jan 29, 2012 12:18 am

Alex Holowczak wrote: (1) All ungraded players go through an iterative process to get an estimated grade
(2) Everyone's grade is calculated based on their previous grade, or the estimate provided by (1)
The estimation process is long established and the forum has explored how it works, whether it converges and whether it's better expressed in matrix notation. What I don't follow is why the published Jan 2012 grade doesn't appear to be (estimate) + (junior increment). The point being that (estimate) is the number used in the calculation of adult grades, or is it?. Is there some undocumented cook based on activity? Thus (estimate) as used in the calculation of adult player grades is based on the most recent six months, whilst (published) includes games for earlier periods. Of course it might be the other way round.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Junior Grades

Post by Roger de Coverly » Sun Jan 29, 2012 12:24 am

Alex Holowczak wrote: My opinion was that we should show the last published grade for everyone, because showing anything else would cause confusion. If necessary, leave it blank. I was outvoted. :)
As it's not a particularly difficult process to download the data and reverse engineer the grading calculation in a spreadsheet, you need to publish the numbers which add up to the published grade. The very fact that players can do this and reproduce the grading calculations adds to confidence in the system. If strange and undocumented cooks are being applied to the grades of junior players, this reduces confidence.

User avatar
Christopher Kreuzer
Posts: 8806
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Junior Grades

Post by Christopher Kreuzer » Sun Jan 29, 2012 2:06 am

Roger de Coverly wrote:
Alex Holowczak wrote: My opinion was that we should show the last published grade for everyone, because showing anything else would cause confusion. If necessary, leave it blank. I was outvoted. :)
As it's not a particularly difficult process to download the data and reverse engineer the grading calculation in a spreadsheet, you need to publish the numbers which add up to the published grade. The very fact that players can do this and reproduce the grading calculations adds to confidence in the system. If strange and undocumented cooks are being applied to the grades of junior players, this reduces confidence.
FWIW, I did this calculation, and the numbers as published did add up to the calculated grade. This also applied to those players who I'd only been able to estimate grades for previously, but who were given grades in the online display that I plugged back into the spreadsheet and it all came out with the right number. It was just a bit disconcerting to see the numbers varying for juniors as well as ungraded players.

Roger, do you have a link to the thread that best summarises the most recent discussion on junior grades and the process for calculating them?

Neill Cooper
Posts: 1298
Joined: Sun Feb 24, 2008 4:43 pm
Location: Cumbria
Contact:

Re: Junior Grades

Post by Neill Cooper » Sun Jan 29, 2012 9:20 am

I think for Juniors the software uses the grade for that grading period only. It can be very different to any published grade. In the following example 42 points lower!

On 8/10/10 Joshua Caulcott-Cooper played Manibharathi Periasamymanjula. The calculation sheet gives Mani's grade as 121 (http://www.ecfgrading.org.uk/?ref=24127 ... 1900871021) when his published grade for July 2010 was 163 and for July 2011 was 141, and is now 150. http://www.ecfgrading.org.uk/?ref=252622A.

Mani's average grade for his 12 standard play games in 2010/2011, as listed at
http://www.ecfgrading.org.uk/?listid=19 ... ef=252622A, was 121 (no age correction!).

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Junior Grades

Post by Roger de Coverly » Sun Jan 29, 2012 11:15 am

Neill Cooper wrote:I think for Juniors the software uses the grade for that grading period only. It can be very different to any published grade. In the following example 42 points lower!

On 8/10/10 Joshua Caulcott-Cooper played Manibharathi Periasamymanjula. The calculation sheet gives Mani's grade as 121 (http://www.ecfgrading.org.uk/?ref=24127 ... 1900871021) when his published grade for July 2010 was 163 and for July 2011 was 141, and is now 150. http://www.ecfgrading.org.uk/?ref=252622A.

Mani's average grade for his 12 standard play games in 2010/2011, as listed at
http://www.ecfgrading.org.uk/?listid=19 ... ef=252622A, was 121 (no age correction!).
Thank you for that example. It's not really completely correct to say that published Junior grades are equal to the (new player) calculation plus a junior increment. If Mani had been a new adult, the published grade would presumably have been 121.

So in the online displays, there's an unexplained jump from 121 as a performance to 150 as a published grade?

I thought grades were supposed to enable players to be listed in order of strength.

Richard Bates
Posts: 3338
Joined: Fri Nov 14, 2008 8:27 pm

Re: Junior Grades

Post by Richard Bates » Sun Jan 29, 2012 11:19 am

Roger de Coverly wrote:
Neill Cooper wrote:I think for Juniors the software uses the grade for that grading period only. It can be very different to any published grade. In the following example 42 points lower!

On 8/10/10 Joshua Caulcott-Cooper played Manibharathi Periasamymanjula. The calculation sheet gives Mani's grade as 121 (http://www.ecfgrading.org.uk/?ref=24127 ... 1900871021) when his published grade for July 2010 was 163 and for July 2011 was 141, and is now 150. http://www.ecfgrading.org.uk/?ref=252622A.

Mani's average grade for his 12 standard play games in 2010/2011, as listed at
http://www.ecfgrading.org.uk/?listid=19 ... ef=252622A, was 121 (no age correction!).
Thank you for that example. It's not really completely correct to say that published Junior grades are equal to the (new player) calculation plus a junior increment. If Mani had been a new adult, the published grade would presumably have been 121.

So in the online displays, there's an unexplained jump from 121 as a performance to 150 as a published grade?

I thought grades were supposed to enable players to be listed in order of strength.
Presumably it's because his published grade includes a bundle of games from the July '10 list to get him up to 30 games?

Paul Cooksey

Re: Junior Grades

Post by Paul Cooksey » Sun Jan 29, 2012 11:35 am

er, I'm not quite getting this...

EDIT (oh, I am now, unhelpful question deleted)

That said, I still don't really like that you cannot work out your own grade, because the grades of any juniors you play are uncertain.
Last edited by Paul Cooksey on Sun Jan 29, 2012 11:56 am, edited 1 time in total.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Junior Grades

Post by Roger de Coverly » Sun Jan 29, 2012 11:42 am

Christopher Kreuzer wrote: Roger, do you have a link to the thread that best summarises the most recent discussion on junior grades and the process for calculating them?
If you go back into the archives of this section of the forum, there are several.

Here's one for example http://www.ecforum.org.uk/viewtopic.php ... 2&start=15

There aren't many new threads a month, so it's relatively easy to go back.

The grading team struggled with the treatment of juniors on their revaluation, not least because whilst their revaluation kept adult players and junior players in approximately the same ranking order as the previous list, it didn't provide a particularly rational way of connecting the two. The consequence was that earlier drafts implied a whole generation of Carlsens. It didn't help that the recursions used would give very high (and presumably very low) values to players who scored close to 100% or 0%. They released a set of grades at the start of August 2009 to criticism of some of the anomalies they had created. With time running out towards the end of August, they came up with the notion of disconnecting the published grade and the grade used in calculation for juniors. There wasn't to my mind any proper debate on the issue. By September 2009, you needed a list, any list, to get the new season under way. Anomalies like the one highlighted by Neil were just swept under the carpet.

To quote just one ongoing anomaly, one of the integrity tests is to sum the total grades published and show that the average hasn't moved much. If the grades as published are not used in future calculations, that test is devalued to some extent.

We will have to watch what is published for July 2012 as Congress and League results are uploaded over the next few months. Whilst results against adult players with grades will be unchanged when the July calculations are made, data for games involving ungraded players and juniors probably will.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Junior Grades

Post by Roger de Coverly » Sun Jan 29, 2012 11:57 am

Paul Cooksey wrote: I do find it weird that his grade counts as lower than his published grade in the last period, and his new grade.
The intention is that if a Junior with a grade of 175 starts playing to a standard of 200, then his opponents get credit for 200 opposition instead of 175. That's partly reasonable. The problem is that it works the other way round as well. Neil's example shows how wide the gap between published grade and calculations grade can become.

In my view it would be more rational to do junior calculations the same way as adults, possibly even without a junior increment, but have a rule that if the raw performance for a period is well above published performance, then do a rerun as a new player. The practical consequence is better value for the 40 point rule. So if as a 175 player, I find myself against a load of 125 players, they count as 135 for me. A 150 player with exactly the some opposition and results scores based on 125. It works the other way as well. If you play FMs and IMs, from 175 your maximum credit is 215. From 150, your maximum credit is 190. So the 40 point rule historically put a minor barrier in the way of improving players joining "the elite".

Richard Haddrell

Re: Junior Grades

Post by Richard Haddrell » Sun Jan 29, 2012 1:32 pm

Roger de Coverly wrote:In my view it would be more rational to do junior calculations the same way as adults... but have a rule that if the raw performance for a period is well above published performance, then do a rerun as a new player.
I'm inclined to agree, except that I'd say "well above or below". But I put it to the experts a while back and they weren't sure how feasible it was. It would mean iterations within iterations. I think.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Junior Grades

Post by Roger de Coverly » Sun Jan 29, 2012 1:56 pm

Richard Haddrell wrote:I'm inclined to agree, except that I'd say "well above or below". But I put it to the experts a while back and they weren't sure how feasible it was. It would mean iterations within iterations. I think.
Whether you would keep junior increments or not, I'm not sure, as they were intended as an anti-deflation hack to deal with supposedly improving players. Given that over 30 games, the statistical theory can say that a grade is only accurate to within 8 points, perhaps you have a rule which says that you revalue only if

under 18 or less than 1 /2 year's history
and
a change of 25 points (in either direction?)
and
30 games or more in the most recent 6 or 12 month period.

That makes the process

(1) Calculate grades with only new players estimated
(2) Inspect results for qualifying +25 players
(3) Rerun with these players flagged as new players
You might have to repeat if this flags any more players with a plus 25 result - this could be avoided by basing the plus 25 test only on games against established players.

Neill Cooper
Posts: 1298
Joined: Sun Feb 24, 2008 4:43 pm
Location: Cumbria
Contact:

Re: Junior Grades

Post by Neill Cooper » Sun Jan 29, 2012 4:03 pm

I think it would be worthwhile distinguishing between active and less-active juniors.
Active juniors - at least 10 games in the past grading period, should be treated as new players. So active juniors could have a rapid increase (or decrease) in grade.
Less-active juniors (less than 10 games) would be treated as adults, (if they had previously played graded games).

This would also avoid the problem that if you play a junior in their only graded game in that time period (even if they previously had a grade) then you get your grade for that game whatever the result.

Post Reply