New grades (split from Editorial thread)

[Roger de Coverly]

The graphs demonstrate visually for the non-mathematician (such as yourself perhaps?)

For what it's worth I do have a maths degree from a senior university.

One of the mathematical methods of testing the resilience of a theory is to find a counter example. So yes, the ratings histories of adult players for the last 14 years is relevant because it shows counter examples to the theory of deflation.

I don't think looking at one year's experience with an average of less than 10 games a player is enough to infer 15 years of deflation.

If you get 67% between players rated 25 points apart, I would want to know that you had ruled out the possibility that the "true" expected score is non-linear. It was non-linear in Elo's original analysis from the 60's.

[Howard Grist]

Has any work been done to say make the hypothetical changes back in 2003 and then run them against the actual results to see if they give a "better" outcome than the actuals?

The calculation of the new grades is being done on results from 2002-2006. Looking at the results for 2007, the new grades give a much better prediction than the current grades do.

I see all this but why does this prove deflation? Surely it might just as well be used as evidence of inflation?

I'm not sure that either inflation or deflation are correct words here. As the lower graded players are out-performing their expected scores, the range of grades is expanding. They need to be brought closer together to correct this. The absolute values that you end up with at the end depend entirely where you decide that the zero should go.

Once the grades are fixed the conversion formula needs changing back. It probably needs to become FIDE = ECF*8+650 (or 700) to account for the FIDE inflation - BUT IT MUST MUST BE FIDE = ECF *8 + x

The FIDE conversion formula will need to be changed, but it will probably be to FIDE = 7 x ECF + x. The FIDE system also has this expansion, albeit in a less severe form than currently exists in the ECF system. The effects of the expansion in the FIDE system can be clearly seen if you compare the ratings of the top players with those of 20 years ago.

Howard
ECF grading system programmer

[Matthew Turner]

I must apologise for my previous comments, I had either misread or mis-interpretted the impact of the proposed changes, of course if a player graded at 100 could go up to 125 or 130 then this will be hugely significant.
I am somewhat reluctant to post more comments which expose my ignorance, but I will give you my take on why inflation and then deflation have been created. An important factor in inflation was juniors competed at World and European Championships, or other foreign events, submitting (or having submitted) games against unrated (in terms of FIDE) players and counting these players as 175. Deflation was created my the new conversion system for ELO ratings. This is why Grandmasters are lttle effected by deflation, since most of their opponents are over 2200, the old conversion was still used. I cannot explain why this has had a greater impact on lower graded players, I could talk about 'trickledown' theories, but I am not totally happy with the transmission mechanisms.
If you accept this explanation then we can see that attempts by the ECF to rectify a very small problem with ELO conversions may have created a potentially more serious problem. I hope that the attempts to resolve these new grading anomalies will not have a similar impact.

[Sean Hewitt]

I must apologise for my previous comments, I had either misread or mis-interpretted the impact of the proposed changes, of course if a player graded at 100 could go up to 125 or 130 then this will be hugely significant.
I am somewhat reluctant to post more comments which expose my ignorance, but I will give you my take on why inflation and then deflation have been created. An important factor in inflation was juniors competed at World and European Championships, or other foreign events, submitting (or having submitted) games against unrated (in terms of FIDE) players and counting these players as 175. Deflation was created my the new conversion system for ELO ratings. This is why Grandmasters are lttle effected by deflation, since most of their opponents are over 2200, the old conversion was still used. I cannot explain why this has had a greater impact on lower graded players, I could talk about 'trickledown' theories, but I am not totally happy with the transmission mechanisms.
If you accept this explanation then we can see that attempts by the ECF to rectify a very small problem with ELO conversions may have created a potentially more serious problem. I hope that the attempts to resolve these new grading anomalies will not have a similar impact.

Matthew - this is one of the myths of the grading system. The FIDE conversion formula is not used to calculate grades. It is simply a way for tournament organisers to compare ratings.

[Sean Hewitt]

For what it's worth I do have a maths degree from a senior university.

Then you really should know better!

[Sean Hewitt]

Once the grades are fixed the conversion formula needs changing back. It probably needs to become FIDE = ECF*8+650 (or 700) to account for the FIDE inflation - BUT IT MUST MUST BE FIDE = ECF *8 + x

The FIDE conversion formula will need to be changed, but it will probably be to FIDE = 7 x ECF + x. The FIDE system also has this expansion, albeit in a less severe form than currently exists in the ECF system. The effects of the expansion in the FIDE system can be clearly seen if you compare the ratings of the top players with those of 20 years ago.

Howard
ECF grading system programmer

Howard, I know I dont need to tell you but the the factor between ECF and FIDE is 8.

[Matthew Turner]

Sean said Matthew - this is one of the myths of the grading system. The FIDE conversion formula is not used to calculate grades. It is simply a way for tournament organisers to compare ratings.
Can this really be true? If I play someone with a Welsh rating of 2000 are you saying they still count as 175 for ECF grading purposes?

[Roger de Coverly]

Matthew - this is one of the myths of the grading system. The FIDE conversion formula is not used to calculate grades. It is simply a way for tournament organisers to compare ratings.

Whilst this is correct for foreign players playing in British domestic events, I think Matthew T was referring to English players (particularly juniors) playing in foreign tournaments. One of the benefits of the higher levels of ECF membership is the facility to have one's overseas events included in the domestic rating. I've been trying to sabotage the domestic rating system by doing this for donkey's years.

I think his question was how overseas games against players with foreign international or domestic Elos should be included in the English system. It's optional as well - however you do have to gamble on a good result by alerting Howard G before the tournament.

[Howard Grist]

Matthew - this is one of the myths of the grading system. The FIDE conversion formula is not used to calculate grades. It is simply a way for tournament organisers to compare ratings.

Sean, it's not a myth. The FIDE conversion formula is used when ECF Members play in foreign tournaments and want to have their results included in their grade. It could be a source of error, but the number of games involved is very small compared to the number of games that are graded so is unlikely to throw the whole thing out of balance.

Howard, I know I dont need to tell you but the the factor between ECF and FIDE is 8.

This is only true if both systems are working correctly. At present neither of them is. Correcting the ECF system on its own will not restore the factor of 8.

If I play someone with a Welsh rating of 2000 are you saying they still count as 175 for ECF grading purposes?

If you play someone with a Welsh rating of 2000 in a Welsh tournament and you request that your results are included in your ECF grade, then he would be deemed to be 175 for grading purposes providing he does not have an ECF grade or FIDE rating. The FIDE conversion formula only applies to FIDE ratings - not foreign national ratings, where the traditional formula of ECF = (Elo - 600) / 8 is used.

[Matthew Turner]

Let me see if I understand. I have taken this from the ECF website

The following conversion applies to national Elo ratings and for FIDE Elo ratings greater than 2327:
(Elo - 600) / 8 = ECF
ECF x 8 + 600 = Elo

For FIDE ratings lower than this, and for ECF Grading purposes only, a different formula is used:
(FIDE - 1250) / 5 = ECF
ECF x 5 + 1250 = FIDE

If I play in the Cardiff open and play 5 players with Welsh rating of 2000 and get 2.5/5 I get a rating performance of 175, but if they all have FIDE ratings instead I come out with a grading performance of 150. Is this really true, is the grading system so sophisticated it checks whether all ratings submitted are National or FIDE?

[Howard Grist]

That's pretty much the size of it. You would send the results to the ECF International Rating Officer (another one of my hats) who would translate your performance into 5 games and 875 points if your opponents are lacking in other ratings or 750 points if your opponents have FIDE ratings of 2000. The games and points totals are then given to the computer.

[Matthew Turner]

Let me try another one on you. Megan Owens has an ECF grade of 114, when she plays in the 4NCL she has a rating of 1820, you can check this on the 4NCL website. This is presumably calculated by

114*5 + 1250

but you are saying that if I play her she counts as (1820 - 600)/8 = 152.5

so an ECF grade of 114 converts to an ECF grade of 152.5 - I am not sure I really understand this?

[Howard Grist]

If you play Megan Owens, then her ECF grade of 114 would be used for ECF grading purposes. The 1820 figure is a number calculated by the 4NCL for the purpose of board order, and does not come in to the ECF grading calculation at all. If you play a Welsh player with a 2000 rating (Welsh or FIDE) in the 4NCL - an ECF graded event - then this is also ignored by the ECF grading system, which will calculate a grade based on the player's performance in ECF graded competitions. The only use of the FIDE conversion formula in ECF grading is if an ECF member plays in an event outside England which is not ECF graded and asks me beforehand to include it in their grade.

[Matthew Turner]

I am finding out so much here. Lets say I have a grade of 200 and I play a Welsh player with a grade x who plays two games in English graded competitions in the grading year. In the other game he loses to a 220. If I beat him I have a grading performance of 210, if I draw a performance of 185 and if I lose a performace of 160? Interesting

[Roger de Coverly]

If I play in the Cardiff open and play 5 players with Welsh rating of 2000 and get 2.5/5 I get a rating performance of 175, but if they all have FIDE ratings instead I come out with a grading performance of 150. Is this really true, is the grading system so sophisticated it checks whether all ratings submitted are National or FIDE?

As it's probably hand computed by Howard, the answer is, I imagine, yes.

More to the point, do we know any players with welsh ratings of 2000 and international ones of 2125?

Actually I tried to find some, however a plausible list of players with a welsh average of 2062 had a Fide average of 1995 This is somewhat confusing.

I checked some names at the top of the Welsh list

Leighton Williams is 2420 welsh and 2389 Fide - only 208 English inactive

The Cobb brothers are both 220 (2360 on the traditional formula) but James (WLS) is 2402 welsh and 2425 international and Charles (ENG) is 2421 welsh and 2407 international.

Inflation in the welsh system?