Rating/grading bounds correlations

Discuss anything you like about chess related matters in this forum.
Paul McKeown
Posts: 3388
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)
Contact:

Rating/grading bounds correlations

Post by Paul McKeown » Sun Sep 01, 2019 6:55 pm

What are people's view of the best correlations between bounds for tournament sections and/or grading prizes, please?

One could use the formula of FIDE = 7.5 * ECF + 700.

However, that seems to put most ECF grades - to my eye anyway - at too high a FIDE level.

What do others suggest?

Roger de Coverly
Posts: 18353
Joined: Tue Apr 15, 2008 2:51 pm

Re: Rating/grading bounds correlations

Post by Roger de Coverly » Sun Sep 01, 2019 10:39 pm

Paul McKeown wrote:
Sun Sep 01, 2019 6:55 pm
However, that seems to put most ECF grades - to my eye anyway - at too high a FIDE level.

What do others suggest?
The 4NCL seem to agree with you by imposing an additional conditions on eligibility for their under 2000 tournaments, that players have to be under 175 as well.

The rankings implied by the ECF grades are likely to be more reliable than those implied by FIDE ratings, particularly for improving but less active juniors and those who play them.

John McKenna
Posts: 3883
Joined: Tue May 17, 2011 2:02 pm

Re: Rating/grading bounds correlations

Post by John McKenna » Mon Sep 02, 2019 12:05 am

"One could use the formula of FIDE = 7.5 * ECF + 700"

Alternatively -

FIDE = (7.5 * ECF) + 600

works much better for me.

I'm not a good yardstick, though.
To find a for(u)m that accommodates the mess, that is the task of the artist now. (Samuel Beckett)

MSoszynski
Posts: 164
Joined: Thu Apr 05, 2007 4:43 pm

Re: Rating/grading bounds correlations

Post by MSoszynski » Mon Sep 02, 2019 4:48 pm

An empirical survey could be done, comparing current players with both a FIDE rating and an ECF grading.

JP Wilkinson carried out such a survey, but it was a very long time ago, last century, in one of the British chess magazines. I no longer have the details except for his formula itself: (ECF * 5.701) + 1136 = Elo rating.

John McKenna
Posts: 3883
Joined: Tue May 17, 2011 2:02 pm

Re: Rating/grading bounds correlations

Post by John McKenna » Mon Sep 02, 2019 5:44 pm

That's food for thought...

Though I have my doubts whether any ECF 'statto' would even bother to repeat the experiment at present...

JP Wilkinson's formula, above, shows its age when the results are churned out...

21st century developments have overtaken it and a race to the biggger bottoms of the grading & rating scales/lists have left it hanging in the upper air...
To find a for(u)m that accommodates the mess, that is the task of the artist now. (Samuel Beckett)

Brian Valentine
Posts: 455
Joined: Fri Apr 03, 2009 1:30 pm

Re: Rating/grading bounds correlations

Post by Brian Valentine » Mon Sep 02, 2019 7:03 pm

We are in danger of going over old ground. We've certainly chronicled the work done in this thread:

viewtopic.php?f=4&t=3994

Basically I started the discussion in 2012 as an interested observer then when I "volunteered" as grading manager, instituted a change after listening to forumites. The formula was reviewed in 2017.

When determining sections I know organisers aren't that happy with it, as basically the "averaging" means 50% of players are overgraded in one list. However some organisers need to convert the other way round. Moving to a four digit grade will simplify the situation (I hope).

Brian Valentine
Manager of ECF Grading

Paul McKeown
Posts: 3388
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)
Contact:

Re: Rating/grading bounds correlations

Post by Paul McKeown » Mon Sep 02, 2019 7:59 pm

Brian, thank you for all the work that you have put into improving the grading system over many years, now, and good luck with the conversion to four digits (hugely overdue*) and monthly updates (one hopes this one is not a "Bridge too Far").

What would you suggest are suitable corresponding ECF boundaries mapping to the following FIDE ratings:
  • 2200
  • 2000
  • 1800
  • 1600
  • 1400
Any tips from your expertise in this would be hugely appreciated...

* ironically drawing us into Caissic alignment with our neighbours, as a chaotic and inchoate narrow political faction struggles insanely, spittle-flecked, to sunder the country's political and economic alignment with those selfsame neighbours

Paul McKeown
Posts: 3388
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)
Contact:

Re: Rating/grading bounds correlations

Post by Paul McKeown » Tue Sep 03, 2019 12:16 am

Ok, I'm going to make a proposal:
  • 2200 approx 210 (official formula gives 200)
  • 2000 approx 185 (173)
  • 1800 approx 160 (147)
  • 1600 approx 135 (120)
  • 1400 approx 110 (93)
Does this reflect reality?

Am I off-beam?

Or am I just attempting the impossible, riding two horses, the stably rated adult population and the under-rated junior population?

I suspect the 2200 equivalent is probably more like 205, but prejudiced in favour of straight lines!

Roger de Coverly
Posts: 18353
Joined: Tue Apr 15, 2008 2:51 pm

Re: Rating/grading bounds correlations

Post by Roger de Coverly » Tue Sep 03, 2019 12:51 am

Paul McKeown wrote:
Tue Sep 03, 2019 12:16 am

Or am I just attempting the impossible, riding two horses, the stably rated adult population and the under-rated junior population?

I suspect the 2200 equivalent is probably more like 205, but prejudiced in favour of straight lines!
If you use the FIDE ratings as a cut off, you need to be pragmatic. Check how many sharks with lower ECF grades you might attract and whether you want them.

as a general premise, if you wanted to rank English players and players active in England by relative strength, the ECF grades are a more reliable guide.

Paul McKeown
Posts: 3388
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)
Contact:

Re: Rating/grading bounds correlations

Post by Paul McKeown » Tue Sep 03, 2019 1:12 am

Roger, rather a gnomic utterance. Pinning you down on a number is the devil's own job. Would you care to make an actual suggestion? (I'm trying to set fair cut-offs for grading prizes and section entries, not some abstruse "ranking" exercise.)

Roger de Coverly
Posts: 18353
Joined: Tue Apr 15, 2008 2:51 pm

Re: Rating/grading bounds correlations

Post by Roger de Coverly » Tue Sep 03, 2019 1:39 am

Paul McKeown wrote:
Tue Sep 03, 2019 1:12 am
Would you care to make an actual suggestion? (I'm trying to set fair cut-offs for grading prizes and section entries, not some abstruse "ranking" exercise.)
You could fall in line with the 4NCL Congresses or alternatively diverge from them.

4NCL Congresses wrote: FIDE Rated U2000: Open to (1) FIDE rated players rated below 2000 FIDE; (2) players without a FIDE rating graded below 175 ECF *

FIDE rating graded below 175 ECF *
* If FIDE rated players also have an ECF grade, their ECF grade must be below 175.

John McKenna
Posts: 3883
Joined: Tue May 17, 2011 2:02 pm

Re: Rating/grading bounds correlations

Post by John McKenna » Tue Sep 03, 2019 2:04 am

Paul McKeown wrote:
Tue Sep 03, 2019 12:16 am
Ok, I'm going to make a proposal:
  • 2200 approx 210 (official formula gives 200)
  • 2000 approx 185 (173)
  • 1800 approx 160 (147)
  • 1600 approx 135 (120)
  • 1400 approx 110 (93)
Does this reflect reality?

Am I off-beam?

Or am I just attempting the impossible, riding two horses, the stably rated adult population and the under-rated junior population?

I suspect the 2200 equivalent is probably more like 205, but prejudiced in favour of straight lines!
Reminds me of -
Even if parallel lines do meet and I see it myself, I shall see it and say that they’ve met, but still I won’t accept it.
(The Brothers Karamazov)
Last edited by John McKenna on Tue Sep 03, 2019 2:11 am, edited 1 time in total.
To find a for(u)m that accommodates the mess, that is the task of the artist now. (Samuel Beckett)

Paul McKeown
Posts: 3388
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)
Contact:

Re: Rating/grading bounds correlations

Post by Paul McKeown » Tue Sep 03, 2019 2:08 am

Thanks, Roger, but isn't that just per the current formula (FIDE = 7.5 * ECF + 700), rounded to the nearest nice number?

(It's also agrees exactly with the old formula, FIDE = 8 * ECF + 600, which held from the 1970s will into the 2000s...)

Do you think 175 and 2000 are equivalent? (My impression is that 2000 FIDE is a generally a little bit - not hugely - better than 175; lower down, though, the divergences can be much greater, particularly with kids.)

Brian Valentine
Posts: 455
Joined: Fri Apr 03, 2009 1:30 pm

Re: Rating/grading bounds correlations

Post by Brian Valentine » Tue Sep 03, 2019 7:39 am

Paul McKeown wrote:
Mon Sep 02, 2019 7:59 pm
Brian, thank you for all the work that you have put into improving the grading system over many years, now, and good luck with the conversion to four digits (hugely overdue*) and monthly updates (one hopes this one is not a "Bridge too Far").

What would you suggest are suitable corresponding ECF boundaries mapping to the following FIDE ratings:
  • 2200
  • 2000
  • 1800
  • 1600
  • 1400
Any tips from your expertise in this would be hugely appreciated...

* ironically drawing us into Caissic alignment with our neighbours, as a chaotic and inchoate narrow political faction struggles insanely, spittle-flecked, to sunder the country's political and economic alignment with those selfsame neighbours
I am of the view that it is up to the organiser and what they are trying to achieve. As I obliquely stated above it depends on whether the primacy is FIDE or ECF. The 4NCL approach given above is tried and tested. With a modest bit of mathematics a more complicated approach can be derived. If a player has only a FIDE or ECF number then use that. If they have both use the average of FIDE and converted ECF.

Roger de Coverly
Posts: 18353
Joined: Tue Apr 15, 2008 2:51 pm

Re: Rating/grading bounds correlations

Post by Roger de Coverly » Tue Sep 03, 2019 9:57 am

Brian Valentine wrote:
Tue Sep 03, 2019 7:39 am
If a player has only a FIDE or ECF number then use that. If they have both use the average of FIDE and converted ECF.
Perhaps that could be followed through into the number used for pairings. It would avoid the apparent randomisation when the top half by FIDE rating meets the bottom half in the first round, but you get a game between players with identical ECF grades.

Extremities of where players are out of line between their FIDE ratings and ECF grades are improved juniors who haven't played much FIDE rated chess. These would have ECF grades much higher than their FIDE ratings. Organisers need to decide whether they are happy with 180+ players in a tournament perhaps intended for under 150s. The other extremity is older players returning to FIDE rated chess and reactivating a higher rating from years or decades earlier. Does the organiser insist they play in a higher section than implied by their ECF grade?

Post Reply