GRADING

ECF News SCCU home page

updated 11.10.06
ECF GRADING
GRADING UPDATE
11.10.06, and already in its second edition
This has been in Open Forum and on the Atticus website, and - perhaps belatedly - popped up the day before yesterday on the ECF site (Grading page). Now that the question is ECF-recognised, your Webmaster would like to throw in some remarks of his own. Readers will know he is not a statistician.
     The problem is that players don't score what they ought to. An analysis of last season's results has shown that players outgrading their opponents by (say) 25 points have not scored the theoretically expected 75%. They have scored more like 68%.
     Why? No one knows. Something has crept in to distort the system. The ECF site suggests possibilities, some of them lost in the mists of time. Apparently no one even knows if the anomalies are real, till the statisticians have had another go at them, but this Website would not mind betting they are. Suggested sources of distortion include the treatment of new players and that of juniors.
     Does it matter? Not sure. As the ECF site points out, people are playing on the average as many players above them as below, so you would expect the system to remain static regardless of distortions. (How about new players? If there is distortion, why don't they tend to counter it? Maybe they do.) At all events the system doesn't remain static, or at least doesn't seem to have over the years. Perhaps it matters.
     Is the ECF grading methodology flawed? Our answer, in the small hours of this morning, was "Of course it isn't. Not essentially." We're not so sure now, having grazed again in the higher pastures of the Atticus site. (It's worth going to, if you haven't.) One possible cause of deflation, touched on there but made clearer to us by someone else, is that even with a static pool of players the average grade can fluctuate. We've always known this but its significance had escaped us. It might not matter if the fluctuation was locally random, but there is some reason to expect a downward drift. We think. There seem to be other issues, but go and graze.
     Would switching to Elo calculations solve the problem? Not obviously, though we're less sure than we were in the small hours. Your Webmaster likes ECF because he can do it in his head, but maybe that's not the only consideration. (Switching to Elo so our grades would look the same as everyone else's is a different question, and we wouldn't vote for it.)

How will this impact on English competitions? You may know that Sean Hewitt, the unopposed candidate for election as Home Director at the ECF Council meeting 21.10.06 - and thus prospectively in charge of grading - was minded to go ahead at once with an adjustment to the 2007 grades. The impact on leagues and congresses would have taken effect from next season. You may also know that he informed the ECF 9.10.06 that he was withdrawing his candidature. His decision was not prompted by the paper on the ECF website. He didn't know about it.
     He has so far resisted attempts to persuade him to change his mind. What the effect will be we don't know, but the impetus for change may be lessened. Some say more checks are needed, and they may be right. Is there time for more checks if we are to have adjustments in 2007 and still give organisers adequate notice? It seems unlikely.
      But we are coming round to the view that adjustments may be necessary.

ECF ONLINE GRADING LIST: UPDATE
The list appeared in provisional form on 23rd July 2006, and was updated 23rd August to take account of corrections and events reported late. (No games from 2006-7 have been included.) The revised August list is official for ECF Standardplay competitions throughout the season. The printed book, on sale from the start of September, mirrors it exactly.
     So what has changed since the original July list? Just for fun we give updated stats below. What they don't show is the number of people with changed grades since July. It is 560-odd, not counting the 40-something who were not in the July list at all. If you think that's a lot, don't forget that late events have been added. They include a fair-sized congress and a fair-sized league, as well as a number of smaller things. It should be noted, further, that a single correction (to a player's identity, say, in a large league) can affect the grades of a dozen people.
     Reflect also that, two years ago, you got a printed list in July with no attempt at corrections. We've come on since then. The online list has not only enhanced accessibility, it has enhanced accuracy as well.

One thing distorts the 2005-2006 comparison. Last year, "halfgames graded" meant halfgames graded. This year the issue is complicated by the Northern Specials. These are northern events which have paid no Game Fee. Calculations have been done for everyone in these events, but the results of non-Members have not been included in their grades. The 2006 “halfgames graded” figure excludes these. The total number of halfgames calculated is some 12,000 higher.

Halfgames graded
Players active
Players published

2005
272457
16426
12237

July 2006
269022*
15607
11842

Aug 2006
270605*
15650
11873

STANDARD
Halfgames graded
Players active
Players published
A grades
B grades
C grades
D grades
E grades

208369
  13810
  10608
    1859 (129)
    1558 (121)
    2473 (115)
    2606 (105)
    2112 (  97)

200134*
  13045
  10352
    1774 (129)
    1608 (120)
    2419 (115)
    2627 (107)
    1924 (  97)

201539*
  13089
  10383
    1799 (129)
    1600 (120)
    2422 (115)
    2637 (106)
    1925 (  97)

RAPID
Halfgames graded
Players active
Players published
A grades
B grades
C grades
D grades
E grades

  64088
    5483
    3677
      418 (  86)
      232 (  89)
      525 (  95)
    1451 (  73)
    1051 (  67)

  68888*
    5641
    3587
      473 (  84)
      232 (  99)
      460 (103)
    1221 (  80)
    1201 (  50)

  69066*
    5627
    3573
      477 (  84)
      233 (  99)
      458 (103)
    1211 (  80)
    1194 (  50)

* For “halfgames graded”, see note above about Northern Specials.
Figures in brackets are mean grades. "Players active" means players with at least one graded game in the most recent season.

Earlier material is in the Archive.

Back to ECF News Back to SCCU home page