Award

Discuss anything you like about chess related matters in this forum.
NickFaulks
Posts: 6060
Joined: Sat Jan 02, 2010 1:28 pm

Re: Award

Post by NickFaulks » Fri Feb 03, 2017 2:25 pm

Jonathan Bryant wrote: Playing a lot of non-elo rated opposition had a big impact on my elo rating
How is that possible?
If you want a picture of the future, imagine a boot stamping on a human face — forever.

Jonathan Bryant
Posts: 3248
Joined: Sun May 11, 2008 3:54 pm

Re: Award

Post by Jonathan Bryant » Fri Feb 03, 2017 2:30 pm

NickFaulks wrote:
Jonathan Bryant wrote: Playing a lot of non-elo rated opposition had a big impact on my elo rating
How is that possible?

Nick, try removing the rating points you gained from your last 14 wins and 5 draws. Only count the games you lose for rating purposes. See what it does to your rating.

NickFaulks
Posts: 6060
Joined: Sat Jan 02, 2010 1:28 pm

Re: Award

Post by NickFaulks » Fri Feb 03, 2017 3:09 pm

Oh, I see. You're saying that if, in some parallel universe, these opponents had been FIDE rated, with ratings equivalent to their national grades, and the games had gone the same way, then your FIDE rating would be higher than it is. That is true, but is a long way from saying that those games caused your rating to drop.
If you want a picture of the future, imagine a boot stamping on a human face — forever.

Jonathan Bryant
Posts: 3248
Joined: Sun May 11, 2008 3:54 pm

Re: Award

Post by Jonathan Bryant » Fri Feb 03, 2017 4:08 pm

NickFaulks wrote:Oh, I see. You're saying that if, in some parallel universe, these opponents had been FIDE rated, with ratings equivalent to their national grades, and the games had gone the same way, then your FIDE rating would be higher than it is. That is true, but is a long way from saying that those games caused your rating to drop.
A rating programme that systematically doesn’t count games you win and does count games you lose will cause your rating to drop. At least it did mine. That system is quite typical in London /South East over last years I suspect. If only because those without ratings are likely to be less experienced / good than those with. The knock-on effect of this and junior issue mentioned above is that everyone becomes under-rated.


Fact is: at at time when my elo was in 1900s i was losing rating points at a time when I was scoring 200+ TPRs on the ECF system.

We could argue, I suppose, that these games might not have had the same result had every other detail been the same except for my opponent having an elo-rating. I would find that quite a difficult position to defend, though.

Anyhoo, I’m absolutely convinced I’ll see 200 ECF before I get back above 2000 elo - unless I stop playing locally and only play abroad. Of course to do either I’ll have to stop being crap, but that is the aforementioned 'other matter'.

User avatar
Christopher Kreuzer
Posts: 7532
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Award

Post by Christopher Kreuzer » Fri Feb 03, 2017 4:16 pm

Why do you think you played so well (200+ TPRs) and then went into a (relative and comparative) decline? Does it take a lot of time and effort to get good and to stay good (or rather, to improve from whatever level you play at innately)?

Nick Burrows
Posts: 1155
Joined: Sat Aug 14, 2010 12:15 pm

Re: Award

Post by Nick Burrows » Fri Feb 03, 2017 4:20 pm

I had in mind ecf rating rather than Fide when making my point. Fide ratings I agree are often grotesquely misrepresentative.

If adults playing in the South East and London suffer from rating deflation due to facing many more underrated juniors, then they should fare well vs similar rated adults from other regions. I have not noticed this phenomena in my playing experience.

Anyone else think this is the case?

Roger Lancaster
Posts: 1068
Joined: Tue Mar 17, 2015 2:44 pm

Re: Award

Post by Roger Lancaster » Fri Feb 03, 2017 5:54 pm

Having already noticed that four of our club juniors were listed on the ECF website as among the 50 "fastest risers" (meaning standard grades improved by 20+ points) in the second half of 2016, I thought I'd check average performance

Nineteen of our juniors had standard grades both at 1.1.16 and 1.7.16 and these collectively increased by 200 points for an average of a 10.5 point rise. Twenty-one had standard grades both at 1.7.16 and 1.1.17 and these collectively increased by 191 points for an average of a 9 point rise. Ages ranged from 7 to 14. If one can extrapolate our club figures to other juniors in the south-east, junior grades are increasing on average at roughly 20 points per year. Our J4NCL experiences do not particularly suggest that this is untrue of other parts of the country.

This throws no light on whether juniors are under-graded at the point the gradings appear (although I suspect they are) but, even if the grades are initially accurate, they will tend to be 10 points too low by the end of the grading half-year.

Roger de Coverly
Posts: 19192
Joined: Tue Apr 15, 2008 2:51 pm

Re: Award

Post by Roger de Coverly » Fri Feb 03, 2017 7:13 pm

Roger Lancaster wrote: This throws no light on whether juniors are under-graded at the point the gradings appear (although I suspect they are) but, even if the grades are initially accurate, they will tend to be 10 points too low by the end of the grading half-year.
Don't forget that what's published isn't used in future calculations. Instead they recalculate the grade credited to the opponent by reference to fresh results. It's the same method as used for new players. The further piece of cooking is that there's an age adjustment, so what's published is higher than what was achieved.

Brian Valentine
Posts: 478
Joined: Fri Apr 03, 2009 1:30 pm

Re: Award

Post by Brian Valentine » Fri Feb 03, 2017 8:12 pm

Roger de Coverly wrote:
Roger Lancaster wrote: This throws no light on whether juniors are under-graded at the point the gradings appear (although I suspect they are) but, even if the grades are initially accurate, they will tend to be 10 points too low by the end of the grading half-year.
Don't forget that what's published isn't used in future calculations. Instead they recalculate the grade credited to the opponent by reference to fresh results. It's the same method as used for new players. The further piece of cooking is that there's an age adjustment, so what's published is higher than what was achieved.
But the age adjustment is in both lists. Between July 2016 and January 2017 the average grade of juniors in categories A-E increased by 1.6 points. Those in A-E in both lists increased by 4.2 (a fairly normal return). New entrants and those leaving the list had lower average grades and since new entrant exceeded leavers the new average ameliorated the increase of those in both lists.

The crude increase in average grade was a bit concerning, but it looks as though the Isle of Man and London Classic FIDE Open attracted an small but unusual number of strong overseas juniors. This disrupted the usual stability.

This hides the regular trends in that those move into more active categories increase by more than average and those moving to less active categories often see their grades going down. For instance there were six juniors in category E that moved to A and their increase was 33 on average.

Roger Lancaster
Posts: 1068
Joined: Tue Mar 17, 2015 2:44 pm

Re: Award

Post by Roger Lancaster » Fri Feb 03, 2017 10:01 pm

Brian has pre-empted me on one point, which is that the grades of the most active juniors (who, almost by definition, are those an adult is most likely to encounter) increase by more than the average. A fast-increasing grade correlates strongly with a fast-improving chess performance so, if the increases at our club are any guide, adults in the south-east are most likely to encounter those juniors improving at more than the 20-points-a-year rate which I cited earlier.

It's interesting that Brian has calculated a 4.2 point junior increase over the past six months. That's calculated across the whole country so our club figures look unrepresentative of juniors nationally although they may be a fair indicator for the south-east. Then again, they may not. But, whichever figures you pick, the principle seems the same and the question is simply of degree.

NickFaulks
Posts: 6060
Joined: Sat Jan 02, 2010 1:28 pm

Re: Award

Post by NickFaulks » Fri Feb 03, 2017 11:38 pm

Jonathan Bryant wrote: Anyhoo, I’m absolutely convinced I’ll see 200 ECF before I get back above 2000 elo
This assertion made me wonder about the FIDE = 7.5*ECF+700 conversion formula, which sets 2200 FIDE = 200 ECF.

I looked at the 24 ENG players with active ratings from 2185 to 2215 who have ECF grades. They have an average grade of 207, quite a serious disparity. Has any proper work been done on this?
If you want a picture of the future, imagine a boot stamping on a human face — forever.

Roger de Coverly
Posts: 19192
Joined: Tue Apr 15, 2008 2:51 pm

Re: Award

Post by Roger de Coverly » Sat Feb 04, 2017 12:04 am

NickFaulks wrote: Has any proper work been done on this?
It was a (rounded) least squares fit not so many years ago. The 7.5 multiplier (rather than 8 ) was justified by looking at the conversion implied by the FIDE tables for players around 1 in 4 apart. So that's saying that if you rate a game in the FIDE system and simultaneously grade it in the ECF system, what results do you get and shouldn't they be the same?

NickFaulks
Posts: 6060
Joined: Sat Jan 02, 2010 1:28 pm

Re: Award

Post by NickFaulks » Sat Feb 04, 2017 12:15 am

So either

a) my sample is too small and everything's ok, or
b) 2200 was not equivalent to 200 when the formula was introduced, or
c) something has happened since then.

If anyone who works with the ECF rating system is interested, I'll leave this thought with them.
If you want a picture of the future, imagine a boot stamping on a human face — forever.

User avatar
IM Jack Rudd
Posts: 4122
Joined: Tue Apr 17, 2007 1:13 am
Location: Bideford

Re: Award

Post by IM Jack Rudd » Sat Feb 04, 2017 12:27 am

24 is too small a sample size for any serious statistical work. (Which isn't to say there is no disparity, just that you can't hope to prove it with that amount of data.)

Alistair Campbell
Posts: 331
Joined: Sat Mar 06, 2010 12:53 pm

Re: Award

Post by Alistair Campbell » Sat Feb 04, 2017 12:30 am

Jonathan Bryant wrote:Playing a lot of non-elo rated opposition had a big impact on my elo rating
This is interesting.

Presumably, rather than basing a grade on (say) 90 results, were one to base the grade on a random sample of (say) 30 games, the results should not differ too much.

Of course, in this case they do (but the sample is biased). This suggests either there is a systematic difference in grading, or that there is a significant degree of over/underperformance against certain subgroups. (Or both).

I think it is agreed that the former holds (both in overgrading adults in the past, and undergrading juniors in recent history).

I suspect the latter is also true. People may over/under perform against certain types of player or in particular competitions. Playing for the team on a cold winter's night after a long day at work may tend to different results than playing for yourself in a weekend tournament.

I'm not sure what the difference would be (in terms of Elo points)

Post Reply