Page 1 of 2
Summary of the new grading system?
Posted: Fri Jul 17, 2020 9:38 pm
by Wadih Khoury
Where could I find a summary of the new grading system?
Of interest, I am particularly keen to understand the relationship between win% and grade differential.
In the current ECF system, it is rather easy: a difference of ten implies a 60/40 win ratio, a 20 point implies a 70/30, up to 50 where the implied ratio is roughly 100/0 (at least for juniors).
For Fide, a 70 delta translates to about a 60/40 win ratio, 150 a 70/30 up to 90/10 just under 400.
Re: Summary of the new grading system?
Posted: Fri Jul 17, 2020 9:53 pm
by Roger de Coverly
Wadih Khoury wrote: ↑Fri Jul 17, 2020 9:38 pm
Of interest, I am particularly keen to understand the relationship between win% and grade differential.
In the current ECF system, it is rather easy: a difference of ten implies a 60/40 win ratio, a 20 point implies a 70/30, up to 50 where the implied ratio is roughly 100/0 (at least for juniors).
For Fide, a 70 delta translates to about a 60/40 win ratio, 150 a 70/30 up to 90/10 just under 400.
As far as anyone is aware, they will be using the same Elo tables as FIDE. They aren't using identical methods though as there's a proposed tweak or two to attempt to deal with juniors and other rapidly improving players having out of date ratings.
Re: Summary of the new grading system?
Posted: Sat Jul 18, 2020 1:56 am
by John McKenna
Wadih Khoury wrote: ↑Fri Jul 17, 2020 9:38 pm
Where could I find a summary of the new grading system?
Of interest, I am particularly keen to understand the relationship between win% and grade differential.
In the current ECF system, it is rather easy: a difference of ten implies a 60/40 win ratio, a 20 point implies a 70/30, up to 50 where the implied ratio is roughly 100/0 (at least for juniors).
For Fide, a 70 delta translates to about a 60/40 win ratio, 150 a 70/30 up to 90/10 just under 400.
You may find something in the link to Monthly Grading in the link below -
https://englishchessonline.org.uk/ecf-o ... lculation/
Re: Summary of the new grading system?
Posted: Sat Jul 18, 2020 11:35 am
by Mike Gunn
If you divide the (4 digit) grading difference by 8 or 7.5 to get the equivalent (3 digit) ECF grading difference and then apply the old rule you won't be far out.
Re: Summary of the new grading system?
Posted: Sat Jul 25, 2020 12:43 pm
by Paul Dupré
Wadih Khoury wrote: ↑Fri Jul 17, 2020 9:38 pm
Where could I find a summary of the new grading system?
Of interest, I am particularly keen to understand the relationship between win% and grade differential.
In the current ECF system, it is rather easy: a difference of ten implies a 60/40 win ratio, a 20 point implies a 70/30, up to 50 where the implied ratio is roughly 100/0 (at least for juniors).
For Fide, a 70 delta translates to about a 60/40 win ratio, 150 a 70/30 up to 90/10 just under 400.
I wish you'd told me that before I played at the London Chess Classic Weekender - I scored 33.33% against players I should have scored 99.99%.
Re: Summary of the new grading system?
Posted: Sat Jul 25, 2020 8:45 pm
by Reg Clucas
Details of FIDE Elo rating calculations are
here. Table 8.1b shows expected score according to rating difference. This table is used to calculate rating changes as described in section 8.5.
Re: Summary of the new grading system?
Posted: Mon Oct 19, 2020 11:05 am
by Roger de Coverly
The list of entrants for the Hull 4NCL weekend online event is at
http://www.4ncl.co.uk/fide/online/entrants_hull20.htm
At the time of writing, every single entrant with ratings in both systems has a higher ECF rating than FIDE one. It's a relatively small sample, so the mathematics of coincidences comes into play, but even so it rather looks as if the ECF system has pitched itself above the FIDE one, at the lower ranges of ratings anyway.
Re: Summary of the new grading system?
Posted: Mon Oct 19, 2020 11:49 am
by Tim Spanton
Roger de Coverly wrote: ↑Mon Oct 19, 2020 11:05 am
The list of entrants for the Hull 4NCL weekend online event is at
http://www.4ncl.co.uk/fide/online/entrants_hull20.htm
At the time of writing, every single entrant with ratings in both systems has a higher ECF rating than FIDE one. It's a relatively small sample, so the mathematics of coincidences comes into play, but even so it rather looks as if the ECF system has pitched itself above the FIDE one, at the lower ranges of ratings anyway.
Interesting. In my experience playing in Europe, most players have a higher Fide than national rating
Re: Summary of the new grading system?
Posted: Mon Oct 19, 2020 1:02 pm
by NickFaulks
Roger de Coverly wrote: ↑Mon Oct 19, 2020 11:05 am
At the time of writing, every single entrant with ratings in both systems has a higher ECF rating than FIDE one.
The entrants to an online event form a biased sample. I don't know what effect that would have.
Re: Summary of the new grading system?
Posted: Mon Oct 19, 2020 3:13 pm
by Wadih Khoury
Roger de Coverly wrote: ↑Mon Oct 19, 2020 11:05 am
The list of entrants for the Hull 4NCL weekend online event is at
http://www.4ncl.co.uk/fide/online/entrants_hull20.htm
At the time of writing, every single entrant with ratings in both systems has a higher ECF rating than FIDE one. It's a relatively small sample, so the mathematics of coincidences comes into play, but even so it rather looks as if the ECF system has pitched itself above the FIDE one, at the lower ranges of ratings anyway.
This is not a surprise. Excluding the noise at the lower end, when I correlated the ECF and fide databases, the linear formula was more like 7.5x+600 rather than +700.
Obviously, there will be case where fide is higher or lower by more than a 100 points, but I'd expect the ECF grade at conversion to be 100 higher on average.
Remains to be seen how it will evolve with time.
Re: Summary of the new grading system?
Posted: Mon Oct 19, 2020 5:14 pm
by Roger de Coverly
Wadih Khoury wrote: ↑Mon Oct 19, 2020 3:13 pm
Excluding the noise at the lower end, when I correlated the ECF and fide databases, the linear formula was more like 7.5x+600 rather than +700.
The plus 700 formula is not that old and was based on a goodness of fit between grades and ratings. The discussion will be archived on the forum. The breakthrough was the realisation that a multiplier of 8 wasn't necessary as if you look at the Elo table used by FIDE. In the range 0 to 200 Elo difference, the "equal gain for equal results" premise points somewhere between 7 and 8.
The implication is that a gap had opened up between FIDE rating and ECF grade. The ECF had adopted the drastic step of annually throwing away all previous results data for the most active juniors whilst FIDE only revalued at the pace of a K=40 parameter.
Re: Summary of the new grading system?
Posted: Mon Oct 19, 2020 5:35 pm
by Wadih Khoury
Roger de Coverly wrote: ↑Mon Oct 19, 2020 5:14 pm
Wadih Khoury wrote: ↑Mon Oct 19, 2020 3:13 pm
Excluding the noise at the lower end, when I correlated the ECF and fide databases, the linear formula was more like 7.5x+600 rather than +700.
The plus 700 formula is not that old and was based on a goodness of fit between grades and ratings. The discussion will be archived on the forum. The breakthrough was the realisation that a multiplier of 8 wasn't necessary as if you look at the Elo table used by FIDE. In the range 0 to 200 Elo difference, the "equal gain for equal results" premise points somewhere between 7 and 8.
The implication is that a gap had opened up between FIDE rating and ECF grade. The ECF had adopted the drastic step of annually throwing away all previous results data for the most active juniors whilst FIDE only revalued at the pace of a K=40 parameter.
When I looked, I don't recall juniors being the main culprits. It's been a while and I don't think I still have the analysis, but I probably removed all/most scores under 1400 and/or all ages under 16/18 (you can get the age when correlating the Fide data with the ECF).
Re: Summary of the new grading system?
Posted: Mon Oct 19, 2020 5:47 pm
by Roger de Coverly
Wadih Khoury wrote: ↑Mon Oct 19, 2020 5:35 pm
When I looked, I don't recall juniors being the main culprits.
It's a second order effect. You have adults playing juniors whose ECF grades are broadly in line with the results they achieve, but whose FIDE ratings are well on the low side. The consequence is that the ECF grade can hold steady, but the adults' FIDE ratings are deflated.
Re: Summary of the new grading system?
Posted: Thu Oct 14, 2021 3:40 pm
by Roger de Coverly
With a few games now having been played, I took a look to see how the variable K factors for juniors work out in practice.
It's done one month at a time, If a junior's rating would go up based on a complete month of results, their K factor would be 40. If on the other hand it would be going down, their K factor is 20.
The effect would be overall to increase junior ECF ratings above what they would be with an invariant K. I cannot help but thinking that it's going to help widen the existing gap that's developing for players whose results go into both FIDE and ECF systems.
Re: Summary of the new grading system?
Posted: Thu Oct 14, 2021 4:26 pm
by NickFaulks
Roger de Coverly wrote: ↑Thu Oct 14, 2021 3:40 pm
The effect would be overall to increase junior ECF ratings above what they would be with an invariant K. I cannot help but thinking that it's going to help widen the existing gap that's developing for players whose results go into both FIDE and ECF systems.
Yes, but in that case, instead of relying on the randomness of results ( which is by its nature random ), why not just give every junior who plays in any month 10 bonus points?