Back to SCCU home page     Forum (general)     Forum: Mobile Phones

Updated 17.12.06
OPEN FORUM: GRADING


From John McKenna
16.12.06
Richard,
Having been involved in the earlier ECF grading debate I was interested to read Trevor Jones's and Sean Hewitt's recent contributions...
     Once again Sean Hewitt's statistical investigations seem to confirm what Elo predicted in a system like ours - namely chronic deflation. The +/- 40-point cut-off to our linear system might be making a significant contribution to the deflation being increasingly larger the lower the grading. The player rated more than 40 points below his opponent is limited in what he can gain from a result (draw or win) whereas the higher player is limited in what he can lose if he doesn't win. Those graded 0-40 points have no band of players below them to gain points from and each class of players above them is protected from the bands below to a limited but probably significant extent. In a system where each band of higher rated players is "insured" against "accidental" loses the deflationary trends will tend to be concentrated in the lower end and become larger the lower the grade.
     (We also reinforce the divisions by organising tournaments on a kind of class system with Open, Major, Minor and other sections.)
     Trevor's comments on the history of the system are a bonus and if he can trace his article in the old BCF Newsflash it might shed more light on the subject. It would be nice to know the actual year of the introduction of the 40-point rule. I recall that someone once mentioned that an extension of the BCF grading system, to below 175 points, was done in the dim and distant past. Could it account for the exacerbation of the inherent deflationary trends which were then skewed to the lower end of the grading scale? Was it also done in the early 1970s?
John McKenna johnrobert_2000@yahoo.com
John's letter has been abbreviated.


From Sean Hewitt
14.12.06
Richard,
Trevor Jones got me thinking. His point about the 40 point rule is of course correct - mathematically it is totally unsound but as it is reciprocal, it should not affect the overall integrity of the grading system. But I started to wonder if there was a way to demonstrate clearly that the problem with the grading system is real and ongoing. So I examined the 1996 and 2006 grading lists, interested only in players who are in both lists so as to compare true like with like. I naturally excluded players who were juniors in 1996 on the premise that they would be expected to improve. This left me with 4815 players, so a pretty decent sample size. I looked at how much their grade had gone up (or down) in that 10 year period, and then banded the results as follows.
1996 grade
200-217
190-199
180-189
170-179
160-169
150-159
140-149
130-139
120-129
110-119
100-109
90-99
80-89
70-79
60-69
50-59
0-49
Change in grade
4.90
1.77
-0.89
-1.80
-3.76
-2.46
-2.40
-3.74
-4.66
-5.29
-4.28
-8.43
-9.01
-12.19
-14.48
-14.87
-22.21
This tells us that below 190 every grading band has suffered deflation since 1996. This proves in my mind that the grading deflation problem is ongoing and is not simply an old problem that is not getting worse as some have suggested. This should not really be a surprise as it is only recently that we have become aware that the FIDE conversion formula no longer works. We can also see that the lower graded the player, the worse the deflation is. This correlates with my previous findings, and adds further support for the correction that I have advocated and published previously.
     However, we can also deduce something else. Since my correction would increase players' grades above the 1996 levels, we can say that this problem must have been in existence since before that time. In fact, my correction increases grades by approximately 3 to 3.5 times the degradation shown above for the last ten years. We can therefore suppose that the problem started approximately 30-35 years ago. So the cause(s) possibly first surfaced in 1970 – 75.
Regards
Sean Hewitt sean.hewitt@virgin.net
Leicester


From Trevor Jones
10.12.06
Richard
Being only on dial-up I don't spend time browsing websites for fun, so receipt of the latest SCCU Bulletin is the first I've heard of recent discussions of our grading system, so thanks for including all those letters.
     Presumably the references to people not getting the percentage scores they ought to according to grading differences are to say that. for example, a 175 player doesn't score 75% against a 150 player (where the straight-line system would expect the 25-point difference to give a +25% score above 50%). However it must be remembered that because of the unavoidable cut-off (so that you don't lose out by beating someone miles below you, in fact dealt with by the 40-point rule), our grading system is sort of Z-shaped rather than merely a straight line. Whether you are 40 or 140 above your opponent, the system assumes you should score 90% (40% above 50%) against him, which obviously isn't true in reality constantly across all such grading differences. Conversely, on the straight line section, to compensate for this, it will also be true that the straight line is an approximation and the reality will probably be some sort of normal distribution resulting in an S-shaped curve that will approximate to the straight line near the 50% mark or equality of grades but diverge the more the further away you are from the middle. This would suggest that the 175 should score less than 75% against the 150. Conversely the 175 should score more than 25% against a 200.
     But I guess that these divergences ought to be symmetrical, and if they are not then that is something that perhaps ought to be looked at.
     Incidentally I too have a copy of Elo's book, bought in the days I was Grading Programmer for the SCCU (before a national BCF system other than for top players), and sometime in the mid to late 1970s I remember writing a statistical article for the then "BCF Newsflash" duplicated newsletter edited by the BCF Secretary as far as I can remember. I'm sure I must still have a copy hidden somewhere but cannot quickly trace it. I was also, incidentally, a member of the BCF Grading Committee at the time the 40-point rule was agreed, prior to which the system was simply to discount 50+ differences if (and only if) you got the "right" result! I remember it being stated that 40 was the mathematically correct figure rather than just a convenient figure, although I never saw the evidence.
H Trevor Jones htjones@raildev.fsnet.co.uk
Guildford


From John McKenna
17.11.06
Roger de Coverley's question, below, about whether to "forget" grades of 50 and under, at the start of every season, may result in "throwing the baby out with the bath water". Some juniors (and low-graded adults) must genuinely have stable grades between 1 and 50 ECF grading points and may not be on a fast track to grades of 100 points and above, nor be on the slippery slope to a grade less than zero. They could be deprived of the chance to gradually build up their grade by it being "forgotten", which is not the same as "throwing away" meaningless grades of zero. How many players currently have grades between 1 and 50 (or zero grades for that matter) and are their grades rising, falling or relatively stable?
      On the other hand, "forgetting" grades of 50 and under could reduce deflation in the system, as a whole, because it would split off the "under 50s" (points not years) into a satellite system of their own, and some players, e.g. rapidly improving juniors, would enter the main system at higher grades than they currently do. Would this approach deal with the purported "deflationary trend at the foot of the list" better than Sean Hewitt's sliding scale of (one-off) corrective points for all players under about 217?
     In The Rating of Chessplayers (1978) A Elo wrote, "Application of the rating system to the entire membership of a national federation requires a range wide enough to cover all proficiencies... from novice to Grandmaster, and enough ballast numbers (points - JM) so no rating ever goes negative." Even if the FIDE rating system is extended down to 1001 it would still have 1-1000 points as "ballast". By comparison, grades in the ECF system now so fully utilise the available scale that there are no positive "ballast" points left, except the decimal points between zero and one! In old printed ECF grading lists I am fairly sure I saw negative grades actually being published for a while. If all such grades are now published as zero, and are "thrown away" at the start of the next season, then the current grading system does not deal with them meaningfully and their owners are all lumped together in a kind of "zero" grading limbo.
     I (still) maintain that Elo foresaw the problems that would eventually plague systems like the one the ECF are (still) using and that he designed a better system. The range and flexibility of Elo's system means that it has been successfully applied to both entire national rating pools (e.g. USCF) and the more exclusive international FIDE rating pool. Again, in 1978, Elo wrote, "The FIDE pool includes only players rated over 2200... which in a national pool might represent only the upper one percent... it mainly consists of players performing close to their ultimate capacities. Very few of them are juniors. Dramatic rating changes are quite infrequent... The FIDE pool is more stable than any national pool is likely to be... its deflation and deflation control procedures are far less acute than in a national pool... Although the FIDE pool has expanded tenfold in the 1970-76 period careful monitoring has revealed no significant deflation... Thus the FIDE pool can serve, with a good degree of confidence, as a common basis of comparison for national pools and a true international standard."
     However, now that FIDE have extend their list down from its original base of 2200 Elo points to 1400 {and below?} their rating pool is becoming more like a national one and the nature of the beast will inevitably change. (Reminiscent, of the demise of the gold standard and its replacement by the mighty dollar - in a system of floating international currency exchange rates - in the 20th century!?) But will it change for the better or, more likely, the worse? Perhaps the appropriate title will soon be the FIDE multinational (instead of international) rating system. Gary Cook wrote, below, that GM Nakamura has a USCF national rating of 2709 points and a FIDE international rating of 2640 points. Although USCF and FIDE ratings are measured on equivalent Elo scales, the USCF system was designed to prevent deflation - mainly due to large numbers of improving players, mainly juniors. While the FIDE system, with far few juniors, was originally designed to prevent inflation. Hence Nakamura's USCF rating is higher than his FIDE one. To me that seems quite understandable and appropriate.
     Gary and Roger also expressed their personal views on the ECF grading system, but they reflect the views of above average players (assuming the average grade is about 120?) and the effects of deflation on them and their grades is much less than it is on players in the bottom half. How long since they played in an U-120 section, and if they ever did were there many precocious juniors playing whose grades were, say, about 80-100 points for a whole season but the following season were suddenly too highly graded for that section anymore?
     At the end of August this year in the revived Surrey Congress, at Guildford, the U-130 section was won by a female junior with a new grade of 117, but she was in the top 5 graded players in it, so fair enough. However, first to fourth equal places in the U-160 section were taken by juniors graded A134, A134, A139 and A128 - 21st, 23rd, 15th and 25th in the grading order respectively. So watch out Gary! And, in the Open section, in the 1st round the top player (ECF 234, FIDE 2503) was defeated by a junior (ECF 174, FIDE 2071), so Roger had better watch out, too!! The advance guard of the growing column of (under-graded?) juniors is reaching the top of the tree.
     If anything, having better juniors requires a better (grading/rating) system. Should we praise the ECF for its successful development of junior chess while we bemoan their effect on an ageing grading system?
John McKenna johnrobert_2000@yahoo.com


From Roger de Coverly
5.11.06
Richard
Some additional thoughts on rating and grading systems.
Underlying model
I rather think Elo's model of rating assumed players of predictable and stable strength. This assumption feels a bit doubtful when juniors or rapidly improving players are involved so one should be careful with mathematical and statistical assertions.
Evidence of inflation
On a personal basis I don't see evidence of deflation in the ECF grading system at my level. For example my first international rating was 2130 in December 1993 and it is now 2097. My British grade is 176 now and was the same then. Over the last 13 years my international rating has usually been above the British grade on the traditional 8* +600 conversion. Does this indicate that British grades are deflated? No I don't think so. Rather as Kevin Thurlow alluded to, the international rating process was selective both in the inital set of games included and the subsequent adjustors. Basically you needed to score points against a high quality field to get on the list in the first place and thereafter you had to do well enough in international tournaments to even play rated opponents. My conclusion is therefore that the bottom end of the international scale is inflated. I imagine the extension of the international scale down to 1400 will eliminate some of these effects over time but in the meantime there are plenty of career 150 players with ratings in the 2000s. I should also add that my ECF grade is based on considerably more games than my international rating.
Rapidplay
I think the ECF could (and perhaps should) switch the rapidplay list to Elo methodology if it intends frequent publication. Rapid play is almost exclusively tournament based for which the Elo approach is designed. It's rather less obvious how to fit leagues into an Elo structure. That said, I don't think the ECF method would work for standard play on a publication basis more often than annual. The reason being that the averaging needed to bring the total up to 30 games would be needed for many more players.
Minimum Ratings
As regards minimum grades, would it be acceptable to "forget" all grades of 50 or under? In other words if a new player scores 25 in their first season, this is published but ignored and they are treated as new players again in their second. So if they play to a 100 standard in their second season, their opponents get full value for their strength. If there's a deflationary trend at the foot of the list, this would help to correct it.
Roger de Coverly rdc@rdc200.fsnet.co.uk


From Gary Cook
5.11.06
Richard/John
I agree that there may be some problems with the ECF system, though I imagine it has affected everyone in a similar way and therefore anyone close to my grade (which amounts to 99% of the people I play) will be out by the same amount and therefore any recalculation will only be cosmetic.
     You are quite right about FIDE and USCF ratings being comparable since they use the same basic model, what I meant to say was that US players have two different ratings, one FIDE and one USCF (for example Nakamura is 2640 FIDE but 2709 USCF) so moving to an Elo system would not mean moving to one grade as some people hope.
Gary Cook ncclsecretary@yahoo.co.uk
Secretary, North Circular Chess League
PS The ECF should be here to serve English chess not the other way round. Sorry if this is old fashioned, but I believe an organisation should serve the needs of its users not the users serve the needs of the organisation.


From John McKenna
4.11.06
Gary/Richard,
True, the ECF grading system isn't "exactly broken" but ECF experts aren't happy with all of its results anymore. Sean Hewitt's statistical analysis shows that if established players' 2005-6 results are recalculated (using successive approximations) then all players below 217 points should have higher grades than those published in the 2006 list. So, comparing current ECF grades isn't as simple and straightforward as it is supposed to be. The ECF, recently, came to the conclusion that "there is a need for further analysis... " and that "... this will be the responsibility of the grading team and in particular the new Manager of Grading and Rating.")
     Replacing the ECF grading system with the FIDE rating system could be a long-term objective, but would probably require major changes to the organisation of chess in England and a dashing new general (Director of Home Chess) with a small army of extra volunteers to do it. Somehow I cannot see the ECF leading that charge, though it could happen gradually over a number of years. On the other hand, replacing it by a national Elo rating system - like the USCF one - would require a good chief scientist (new Grading Manager mentioned above) and a lot of technical work by an enhanced, dedicated band of boffins, so that too probably won't happen, soon, either.
     In the short to medium term the current grading system will probably grind inexorably on and down. However, a small number of changes (including the introduction of an absolute minimum grade of, say, zero ECF points and, possibly, Sean Hewitt's corrections) could help restore its integrity and raise its quality. A change to a 4-figure grade, using the same modified ECF methodology as the Welsh, might also be worthwhile. English and Welsh grades would be comparable, even if the two federations didn't share data, and if they did Welsh and English who played in each others events could have one (combined) grade instead of two.
     (NB - any and all such changes should be tested and checked, as thoroughly as possible, and only introduced to the system one at a time with suitable time gaps between them.)
     By the way, USCF and FIDE ratings are comparable with each other because they are based on the same Elo methodology. ECF and WCU grades are based on a different methodology and not comparable with FIDE and USCF ratings. All that can be done is a rough conversion - see Kevin Thurlow's contribution of 29/10/06 for a critique of that. Regards
John McKenna johnrobert_2000@yahoo.com
PS The way forward is not to ask, "What does the ECF do for me?" But, "What can I do for the ECF?"
     rjh: - We do have a minimum grade of zero (actually 1), in the sense that grades lower than this are thrown away. I don't know whether this is the sense John intends.


From Gary Cook
1.11.06
Richard
Maybe I am missing an important point somewhere. Are we talking about moving to replace the ECF system with the FIDE system or a FIDE type (ie 4 digit) system?
     If the former what is the point? It would exclude probably a third of players and would not be usable for the evening leagues. If the latter I again have to ask why? It doesn't really matter to me if I have a 3 digit or 4 digit grade, so long as I have something to compare against my opponent(s). Why fix something that isn't exactly broken (too much)? The Americans use a FIDE type rating but it isn't comparable with the official FIDE, so their players have two grades both 4 figures.
     Could I make the suggestion that the new(ish) ECF regime start by dealing with the old "what does the ECF do for me" problem, maybe then everyone would jump at the chance of being a member and one of the hurdles to bringing in FIDE ratings would be dealt with.
Regards
Gary Cook ncclsecretary@yahoo.co.uk
Secretary, North Circular Chess League
rjh: - To answer Gary's question. If I'm not mistaken Paul Keevil suggests the former course, and John McKenna suggests (one form of) the latter.


From John McKenna
31.10.06
Richard,
Paul Keevil made some very relevant points about the state of chess in the UK and the merits of (dare I mention) a continental approach. I would like point out that the recent election of a majority of "Way Forward" candidates to positions of authority in the ECF means, in principle, a commitment to promoting FIDE-rated events and looking into the feasibility of moving to a FIDE/Elo rating system within a reasonable timescale, I believe.
     However, as David Sedgwick has stated, below, ideas in principle are not always implemented in practice and the fact that the only candidate for the important position of Director of Home Chess, withdrew 12 days before the AGM (over what may have been grading-related issues) does not inspire confidence. It seems that, going into the AGM on 21st Oct. just past, neither the old guard, nor the new brooms at the ECF had anyone prepared to take on a role that is absolutely central to so many aspects of English chess in general and the grading system in particular. Who is going to persuade organisers to run FIDE-rated events if, as Paul suggests, that is the best way to make progress? By analogy with economics and politics, it is as if England were a peripheral member of the EU and UN, in a position of splendid isolation, with an introverted and backward-looking system in need of modernisation.
     The new brooms at the ECF implicitly acknowledged all of this in their election manifesto which also points out an internal, north-south divide which has been rumbling on in the background for ages. Will things now change or simply revert to the old status quo, soon? Watch this space!?
Regards
John McKenna johnrobert_2000@yahoo.com
PS Chris Majer wrote that he believes the Welsh Chess Union uses a modified ECF methodology to generate its four-figure ratings and Paul pointed out that federations should share information. If the ECF quickly moved to the Welsh system it could act as a stepping stone (admittedly mainly cosmetic) to a FIDE/Elo system. (Probably what the WCU had in mind when implementing their current system.)


From Kevin Thurlow
29.10.06
Hi Richard
I think there are time-limit restraints on FIDE rating, so evening leagues may not be eligible. Does anyone care about the Rapidplay grading list?
     No grading or rating system works pefectly, and I wish people would recognise that all the grading or rating does is establish an approximation of performance (not ability or strength) over the period measured. The main reason people have low gradings is because they lose games.
     You cannot convert accurately ratings/gradings from one system to another as they are calculated in entirely different ways. It's not like Fahrenheit to Celsius or inches to centimetres, where the scales are linear in each case.
     Having said that, I think the new Elo/ECF conversion is based on a flawed premise, and I recommend returning to the old one. I think the new conversion arose when the lower limit of Elo was 2000. Someone noticed that some 160s were rated 2100 (instead of 1880), and changed the conversion so that instead of 175 ECF = 2000 Elo, then 150 ECF = 2000 Elo. But I believe what they overlooked was that some 190s (eg) did not achieve Elo ratings as they scored below 2000. So the results were skewed - instead of getting an average, half the results were discarded, and it was not just any half - it was one particular half. Doubtless someone will dispute that. And of course, some of us play better at a game a day played in splendid conditions, compared with an evening league where you rush back from work and rush out again.
Best wishes
Kevin Thurlow Kjt2300@aol.com
Redhill


From Paul Keevil
29.10.06
Gary/Richard
Thanks for replying to my thread on this forum. My views may be a bit naïve, but for the life of me I really don’t understand why a person should have more than one rating. It just seems absurd. I do not presently have a FIDE rating. However, in recent years I have obtained Part FIDE Ratings of 2050 at Coventry and 2060 at Chorley, yet my last Welsh Chess Union rating (from 7 years ago) was 1780-ish and I believe I may have had a BCF rating of something like 1950. So the question I have to ask myself is, am I a 1780 player, a 1950 player or a 2055 player? The whole thing is just daft. If I play in Wales against an 1800 opponent and draw, my opponent is going to lose points. If I play an 1800 player in England and draw, my opponent will gain points, yet if I play in a FIDE competition against an 1800 opponent and draw, my opponent will gain no benefit at all as I am technically unrated.
     Is it right that should I choose to (I wouldn’t) I could enter a Welsh Major Event (i.e. up to 1800) as a player without a FIDE Rating and rely on my WCU rating some 7 years earlier, and potentially walk away with prize money? I don’t think so, but I am aware of some individuals who have done just that.
     That was the rationale behind my thinking. The sooner everyone has one distinct rating the better!
In respect to streamlining, Gary, I understand your point. But as a businessman, I always ask myself these questions:
     1) Are we running as efficiently as we can?
     2) Can we do it better of differently?
     3) Can we cut any red tape?
     4) Can we offload anything that WE do not have to do.
No, I don’t think we are. Results go from tournament organiser to local grader to central grader to publisher. That’s 4 people. I am suggesting that it should go straight from the tournament organiser to FIDE, via the IRO. Why should we worry about the calculations, when FIDE can do it and produce a list 4 times per year!
     In respect to the ECF, I would be interested to learn why people don’t join their member organisation. Do the ECF rate the games of individuals who are not members of the ECF? If so why? As some of you will appreciate, I am not a great lover of the ECF. But the cost of membership is negligible for what you get in return - £16 or £1.30 per month – Some people spend more than that on a Saturday night!
     This is an interesting topic. I am not sure that I have all the answers, but it’s interesting discussing the pro’s and con’s.
Regards
Paul Keevil Paul Keevil paulkeevil@capitalgrp.freeserve.co.uk
Wales
rjh: - It seems to be two or three topics. One or two of them are getting repetitive. No one has denied we'd be better off without multiple systems. I believe it myself, though I'm not sure the answer is as simple as Paul suggests. As for his spin-off benefits in efficiency, I think them illusory and he knows why. I do have some experience of the ECF system. Please no more contributions on these two points unless you've got something new to say.


From Gary Cook
29.10.06
Richard
Paul Keevil makes some interesting but in my opinion rather naive comments.
     He says that local graders would not be needed if we moved to a FIDE rating system "as the results would be submitted by the tournament organiser or arbiter" - since in most cases that is exactly how it works under the current ECF method I am not sure where the streamlining would come from.
     Anyone under 1400 "would not be 'saddled' with the burden of a low rating". Now by my old fashioned calculation 1400 roughly equals 100 ECF, so a move to FIDE ratings would do away with the U100 county matches, not to mention all those grading prizes for a whole raft of players, but at least they would not be saddled with a low grade.
     Finally Paul would be surprised if becoming a member of the ECF would be a major problem for most people - this seems to have been a major issue for a lot of people for the past few years with no end in sight. How many members does the ECF currently have as a percentage of those with a grade?
Gary Cook ncclsecretary@yahoo.co.uk
Secretary, North Circular Chess League
rjh: - There were just under 1500 Members at the last count. But some of them don't have grades, and a couple of hundred don't appear to play chess at all. Among graded players, about 1200 out of 11,800 are Members. It will be going up, no doubt, with Membership Schemes round the corner.
     Organisers and arbiters don't usually report results. Mostly they give the pairing cards, or whatever, to a specialist local grader who knows how to do it. But I agree Paul's streamlining is not apparent.


From Paul Keevil
27.10.06
Richard
In your editor’s note [see letter below], you raise some important questions. I had considered them, but perhaps hadn’t put my view across eloquently enough to cover them. So apologies for that.
     My vision is a system by where all leagues and tournaments are FIDE rated. I believe this is workable as there are enough FIDE rated players in the UK for this to work, and as time elapses more and more people will obtain a FIDE rating. I would anticipate that upon the cessation of a tournament, the arbiter/organiser would compile the results in the prescribed FIDE format, which popular software can accommodate, and report the results to the IRO for checking and onward submission to FIDE.
     So in answer to one of the valid questions, I do not believe that there would be the need for individual graders as the results would be submitted by the tournament organiser or arbiter, who are involved in the event anyway (irrespective as to whether it is FIDE Rated or not). Therefore, the whole process would become more streamlined.
     The 2nd point is that all individuals would have to become members of a federation. I would be surprised if that is a major problem for people.
     The 1400 cut off is a very important point, and in answer to it I would say that you are either rated or are not. Under my vision, an unrated experienced player should be able to obtain a FIDE Rating very easily. However, a lesser experienced player would not be “saddled” with the burden of a low rating (of say 1250) following an average start to their first ever tournament. Instead, they would benefit from remaining unrated until such time as they improved and be able to move forward once they have obtained a half decent rating.
     I firmly believe that this process is very workable and would take a lot of the admin out of our game. In Spain, for example, there are over 340 Internationally Rated Tournaments per year. There is no reason why we cannot adopt the same approach.
Paul Keevil paulkeevil@capitalgrp.freeserve.co.uk
Wales
[rjh: - Readers will know what I said, and I have nothing to add to it.]


From Paul Keevil
25.10.06
Richard
On the subject of grading, one question that puzzles me is “why the need for 2 grades or in some instances 3?” At the present time it is possible for someone to be registered with the ECF and by virtue of playing in the 4NCL receive an ECF rating. If you play your chess in Ireland, Scotland or Wales, you may also pick up a “Local Rating”. Experience tells me that federations very rarely share their games with other federations, meaning a disparity straight away and if two ratings are not enough, many players now also have a FIDE Rating making 3 in total.
     Surely only one rating is needed?
A simple answer would be to make every competition in the UK a FIDE Event. The cost is relatively minimal and the end result would be that everyone who performs above 1401 will have a FIDE Rating. Apart from other benefits, this will mean that there will be no need for individual graders - as FIDE will do all the rating calculations. There will be no need for discussions as to whether ratings are “moving out of sync” as, again, this is something FIDE will control.
     Surely we should make it easy for ourselves and let FIDE do the work? After all we pay them enough in annual registrations.
Regards
Paul Keevil paulkeevil@capitalgrp.freeserve.co.uk
Wales
rjh: - It's not clear that a switch to FIDE would eliminate the need for local graders. Graders don't do the ECF calculations, the central computer does. Graders report results. Someone's got to do that whatever the system.
     All players would have to become Members of their national federation, of course. One disadvantage of the switch, currently, would be the 1400 cutoff. Another would be the lack of a FIDE Rapidplay list.


From John McKenna
25.10.06
Let me start by saying that the names Chris Majer, Howard Grist and Richard Haddrell are all known to me, by reputation, for the good work they have done on behalf of English chess. They, however, probably don't know me from Adam. So, by way of a very short intro I'll just say that I am, indeed, John R McKenna (BSc. Computing) of Crystal Palace Chess Club.
     Below, Chris mentions the eminent statistician Sir Richard Clarke (father of the former Labour Minister Charles Clarke) who was also the founding father of the BCF grading system circa 1958. (And a good father he was, too, since he continued to administer the system through its infancy, I believe. By the way, it would be interesting to have a brief history of its life after that.) Since those halcyon days it would seem that the necessary statistical monitoring has not been carried out. Any such system must be statistically monitored - no matter what the methodology (Harkness, Ingo, Elo, Thompson, Sonas, etc.) - or it will be in danger of not reflecting, reasonably accurately, its members' ratings, and the sources of deflation, inflation and other anomalies will go undetected, as Elo stated in his book of 1978.
     I agree with Chris that a statistical "MOT" of the current system is the first thing to do - even though it may only reveal if a systemic problem exists and, hopefully, indicate what action could be taken to remedy it. (Sean Hewitt's proposal, below, seemed a bit drastic to me which is why I got involved in this debate. Although by then David Sedgwick had already announced, again below, Sean's withdrawal and the ECF's return to inertia as far as grading was concerned.) I also agree that if, as Howard proposes, an examination of historical data is done that the findings may indicate the origin of any such problems and how best to avoid them in future.
     Even if that cannot be done or is inconclusive, just following Elo's advice in Chapter 3 of The Rating of Chessplayers (1978) could restore a good deal of confidence in the future integrity of the existing ECF system. But, by historical analogy, if Sonas is to be our rating Einstein then isn't Elo our Newton and Sir Richard our (very own) Ptolemy? Which then begs the questions why are we using a "Ptolemaic" rating system and when might we ever expect to progress beyond it?
     Having said all that, I still think Howard can be proud of his authorship of the current grading program and that Chris and Richard can be proud of their contributions to the ECF system which I, for one, appreciate and freely admit is much better than no system at all [Gee, thanks, Guv - rjh]. Still, I'd like to make a final plea for serious consideration - LET'S, AT LEAST, HAVE AN ABSOLUTE (PRACTICAL & THEORETICAL) MINIMUM ECF GRADE OF ZERO POINTS, PLEASE! (A minimum concept for Elo points is justified in Elo's book - on p.166 - using two pools of data one from the USCF and one Belgian Chess Federation.)
     Finally, I'd just add that the move to the online ECF grading system was a big step forward in its presentation and accessibility. Though, like the televising of Parliament, it is no guarantee of the integrity and quality of what goes on inside, at least we can see it better.
PS To me "diehards" are those who wish to perpetuate something for its own sake (or personal reasons) rather than allow change for the common good.
John McKenna johnrobert_2000@yahoo.com
rjh: - I think what Chris wondered was who they are. Never mind.


From Chris Majer
23.10.06
John
In response to John McKenna (two letters down)
I have in fact already obtained a copy of Elo's book to assist me with conducting a statistical MOT of the existing grading system, which I hope to conduct shortly. I know that Sir Richard Clarke corresponded with Prof. Elo when setting up the grading system. So presumably he knew about and understood the limitations of the simplifications. Plainly, this is something to revisit. Equally, the anomalies may be linked not to the grading methodology but to some other aspect. It would be very foolish to change the methodology and then discover that the problem was still there.
     I'm not sure who you mean by the die-hards. There are about 100 Graders who submit results to the ECF central system. Any radical change is going to have to be done by consultation with them, not to mention the leagues and congresses.
regards
Chris Majer CEMAJER@aol.com
PS my understanding is that the WCU use a "BCF methodology" and then convert to a 4 figure grade.


From Howard Grist
23.10.06
Richard,
I am perplexed by the large number of people, including John McKenna, who are attracted by the complexity of Elo's probability curve over the ECF's much simpler straight line as a predictor of a game's outcome. Jeff Sonas has done some investigation into using ratings as a predictor of a game's outcome and reached the conclusion that a straight line makes for better predictions. You can read the article at http://www.chessbase.com/newsdetail.asp?newsid=562. (His website http://db.chessmetrics.com/ is also worth a visit.)
     Obviously, with today's improved computing power and chess databases it is much easier to do this sort of analysis now than it was in the time when Elo wrote his book. It should also be noted that rating systems are inherently deflationary - a player will improve in playing strength as he gains experience and most players stop playing with a higher rating than when they started - so measures have to be taken to ensure that this does not occur in practice which is why the ECF system adds 6-10 points to the grades of junior players.
     No-one has really investigated how good the ECF grading system is at predicting a game's outcome until very recently as the data to do so has not been readily available. However this situation has now changed and I will shortly be examining the grading database of some 500,000 results to investigate where the problem(s) with the ECF system lie.
Howard Grist howard.grist@talk21.com
Southend
Disclosure: I am the author of the current ECF grading program and also the ECF International Rating Officer (and so deal with the FIDE rating of English events). I have a degree in Maths.


From John McKenna
21.10.06
To: Open Forum c/o the SCCU webmaster
A plea for sanity in the ECF grading debate
Questions have been raised by a number of people, on the SCCU and other websites, about possible problems the system and its integrity. In his classic 1978 book, The Rating of Chessplayers, Arpad Elo covered the whole subject in an excellent practical and theoretical treatment which addressed many of the questions (including the "specter" of systematic deflation) being debated in English chess circles today. Whoever takes over the official positions concerned with grading in the ECF, after today's AGM, should make a study of, at least, Chapter 3: Rating System Administration before changing a single thing. A copy of the book might be found in the new National Chess Library, if not I strongly suggest one is obtained.
     Proper attention should be paid to the qualitative and quantitative differences between an Elo rating system and the ECF grading sytem. In the 1950s (prior to the introduction of a national Elo rating system) the US Chess Federation used the Harkness System (on which the old German Ingo System and the BCF [now ECF] system were also based) and this is what Mr Elo wrote, 28 years ago, about it in his seminal work, "At first glance it seems appealingly simple, but thoughtful examination reveals that a strong player can lose (grading) points even with a perfect score (against significantly lower graded players - JM) and a weak player can gain points by losing all games (against significantly higher graded players - hence the BCF's introduction of the 40 point rule, in the past, to remedy these two abominations - JM)
     The equation (still used in the ECF system today - see below) yielded invalid results and its continued application developed uncertainties which became disturbingly larger than could be expected from common statistical variation. Although the system was broadly used, covering many thousands of USCF players between 1950 - 1960, and adopted by a number of other national federations its results never lived up to its objectives, and it has since been replaced by the Elo system almost universally."
     I leave readers to draw their own conclusion, but mine is that - like the Irish, Scottish, Welsh and other federations in these isles - the ECF should upgrade (!?) to a statistically-streamlined Elo rating system with a probability [Rp = Rc + D(P)] drive, rather than waste more resources tinkering with the ECF grading version of a clapped-out Ford Prefect with deflationary tyres that is erratically powered by a crude linear [Rp = Rc + 50 (W-L)] motor with its +/- 40 point governor.
     Of course, the mere fact that to do so would be regarded by many (diehards) as the equivalent to driving on the same/wrong (delete as appropriate) side of the road as most of the rest of the world, will probably ensure that it could only happen here over their dead bodies.
     Let's face it, the owners of the current system cannot even tell us if it has been passing or failing its (statistical) MOT so why should we all be forced to continue to use it any longer?
John McKenna johnrobert_2000@yahoo.com
Crystal Palace CC, we believe. John does not state his credentials.
There is, of course, lots of other material about this on the Atticus site, some on the ECF site, and who knows what elsewhere. Somebody ought to be doing a co-ordination job.


From David Sedgwick
14.10.06
Dear Richard,
I'm afraid that the SCCU website is losing its reputation as the only place to go if you want to know what's happening in the ECF. From the excellent Atticus website I learnt on Tuesday 10th October that Sean Hewitt had withdrawn his candidature for the post of ECF Director of Home Chess on the previous day. The paper by David Thomas and Chris Majer was also published on that day on the ECF website. Strange as it may seem, the timing was in fact a coincidence.
     I've been around long enough to see lots of new people with fresh thinking take up positions in the ECF, only for them to depart the scene fairly rapidly when they find that they can't implement their ideas quickly, or even at all. However, I think that Sean Hewitt has just set a new record for brevity of tenure - minus twelve days.
     With regard to the grading issues Sean had convinced me - and I'm not the most revolutionary of people - that doing nothing was not an option. He hadn't convinced others and it looks as though nothing is precisely what the ECF will now do. Even Martin Regan appears to consider this the best short term course of action.
     It's all rather ironic. For months we've been complaining that there is never anything on the ECF website to tell us what's going on. Then we suddenly get a paper which makes it clear that the only candidate for an important Directorship would not have had the support of key colleagues.
     It's also very sad. In my view Sean had a lot to offer.
David Sedgwick david.sedgwick@amserve.com
23 Tierney Court, Canning Road, Croydon CR0 6QA
David's letter was sent 14th October but got held up three days in transit.


From no one
9.10.06
Not a letter to the Forum. But the ECF Home Director and Grading Manager have put out a statement on the ECF website (Grading page) which relates to the topic of the last two letters. Discussion continues on the Atticus site.


From Sean Hewitt
2.10.06
Richard
As “the only candidate for the job of Home Director” I must reply to Roger de Coverly’s recent letter.
     The ECF Grading Manager and others on the ECF grading team had been investigating the goings on within the system, prompted by the failure of the old FIDE conversion formula. During the course of that investigation they became concerned that higher graded players were not scoring as highly as they should against lower graded opposition, and were also worried about the high number of players who have grades below zero. As a result of my background in statistics, I was asked if I would investigate and my methodology and findings are published at the address Roger gives. Any one who cares to read the paper will see that rather than suggesting that “lower rated players should be given 10 or 20 more points” I have suggested a correction formula which effectively will raise the grades of all players currently below 217. Indeed Roger himself would receive 9 points on his “low” grade of 176!
     Roger is of course entitled to his opinion that such a correction would “seriously annoy most of the country's active chess players” but in my conversations with players I have found no-one who seriously objects to having their grade increased!! Indeed, many players have told me that they knew that something had gone wrong with gradings, and they all seem to put the time scale to the last 20 years.
     Organisers certainly need time to assimilate new information and to decide how it affects them. With that in mind the correction would need to be finalized and details published hopefully by the end of January, with the new list in July accommodating the changes.
     Perhaps it is merely a happy ‘co-incidence’ that my proposed correction to the grading system would mean that the old FIDE conversion formula of BCF * 8 + 600 = FIDE would work again. But then again, if my suggestion is actually correct, perhaps that shouldn’t be so much of a surprise!
Sean Hewitt sean.hewitt@virgin.net
Leicester
PS - If I am elected, this will be an important year for the grading team. If anyone is interested in getting involved with the grading team, particularly those with a Maths or Stats background, I would welcome the opportunity to discuss the matter further.


From Roger de Coverly
29.9.06
to the SCCU website
The only candidate for the job of Home Director seems to believe that the ECF grading system is "incorrect" and that all lower rated players should be given 10 or 20 more points. There's a paper which supposedly "proves" the assertion at http://www.lrca.org.uk/lrca/Grading/Grading%20Correction1.doc. There's more discussion at the Atticus site http://www.atticuschess.org.uk under Forum | General Chat | [poll] Grading. See also under ...General Chat | Gradings.
     Personally I think that the ECF tampers with the grading system at the extreme risk of seriously annoying most of the country's active chess players and organisers. Perhaps those closer to the grading system could comment as to whether the rerating proposals have any merit.
     I am aware that the international ELO ratings have drifted out of line with ECF grades . I had always supposed that because of the 2000 minimum rating (as was) and the exclusion of unrated players in the ELO calculation, the international rating only reflected a 2000-ish player's best performances and therefore an ECF grade which reflected all performances would "obviously" be lower. One supposes that with international ratings now extending to 1401 this effect would disappear over time.
Roger de Coverly rdc@rdc200.fsnet.co.uk
rjh: - I think the drift out of line with FIDE ratings is a separate question. The candidate Director's contention is that ECF grades themselves have got "stretched". I'm glad Roger has raised it, because I've been wanting to. However, while I wear an ECF Grading hat I have no expertise in statistics that you'd notice. I hope ECF people better qualified than I will feel able to respond.


Earlier material (lots of it) is in the Archive.


Back to top      Back to SCCU home page