Junior Grades

[Roger de Coverly]

I resisted mentioning Elo It would be possible, and it would make rating junior events easy.

Elo scales have the advantage of going some 600 or more Elo points below the equivalent ECF ones, thereby giving more scope to rank players who would be zero on the ECF approach. It still has a zero and the potential for negative ratings. If you wanted continuous or monthly ratings, an Elo approach can work better. It doesn't solve problems of result reporting or players improving to a standard several hundred Elo points above their published rating.

[Adam Raoof]

Using ECF grading it would be very simple as a lot of junior events use Tournament Director.

Do you make new roads for old cars, or for cars not yet launched? If any ECF policy is predicated on Tournament Director then it will take a very long time indeed to change, if indeed any change is desired!

Fair point. So, let us have a separate grading list for junior-only events. One for Phil Ehr, I think!

[Richard Haddrell]

Since all juniors start the grading process with no grade, then if you have 'pockets' of juniors who just play other juniors then they will all end up with grades of 0. It will not matter if they lose all their games or win all their games.

Not true. Juniors "start the grading process with no grade" in the sense that their original grade is ignored, but in its place they get a calculated "start-of-season-grade" based on available games over the last three years. It might be 0 or it might be 226. It makes no difference whether their opponents are juniors.

There is a special case where a pocket of ungraded players - they need not be juniors - have played entirely among themselves. But that means real ungraded players. No grades at all, not even calculated "start-of-season-grades". Since these players have no contact with the outside graded world, grading their games is impossible and not attempted. Even so, they don’t get grades of 0. They just don’t get grades.

[Paul Cooksey]

Do we know yet if the new system for juniors is achieving its aims?

[Peter D Williams]

Using ECF grading it would be very simple as a lot of junior events use Tournament Director.

Do you make new roads for old cars, or for cars not yet launched? If any ECF policy is predicated on Tournament Director then it will take a very long time indeed to change, if indeed any change is desired!

Fair point. So, let us have a separate grading list for junior-only events. One for Phil Ehr, I think!

Is there enough junior only chess events to want to do this?

[Matt Mackenzie]

There is a special case where a pocket of ungraded players - they need not be juniors - have played entirely among themselves. But that means real ungraded players. No grades at all, not even calculated "start-of-season-grades". Since these players have no contact with the outside graded world, grading their games is impossible and not attempted. Even so, they don’t get grades of 0. They just don’t get grades.

Just out of interest, does this ever actually happen??

[Ian Thompson]

There is a special case where a pocket of ungraded players - they need not be juniors - have played entirely among themselves. But that means real ungraded players. No grades at all, not even calculated "start-of-season-grades". Since these players have no contact with the outside graded world, grading their games is impossible and not attempted. Even so, they don’t get grades of 0. They just don’t get grades.

Just out of interest, does this ever actually happen??

Yes. It's definitely happened once in the 12 years I've been doing the Surrey Border League grading (and I've got a feeling its happened twice). I've had groups of 2 or 3 new players who have only played amongst themselves, and haven't played any other graded games outside the league.

[Roger de Coverly]

Do we know yet if the new system for juniors is achieving its aims?

Without knowing what the aims were, it's very difficult to say.

I can give a potted history.

In the early days of grading (1950s to late 60s), there was no special treatment for juniors. It was perceived that this caused the whole scale to spiral downwards. This was observed in particular in the lists published of players with grades above x, that the list had fewer names every year. It was suggested that the dilution effect from improving players coming into the top echelons with lower grades than established players had something to do with this. As a hack, for calculation purposes, you would score (their grade) +/- 50 + 5 (later 10) when playing under 18s. For board order, eligibility and pairing purposes, it was often the practice to take the grade + 10 for juniors.

Fast forward to the mid to late 1980s and the national computerisation and centralisation of the grading system. It was observed that the "centre" of the grading system was edging upwards from the 110s into the 120s and towards the 130s. A perceived reason for this was that the +10 applied in junior calculations was too generous at the older age ranges and thus was acting as an inflationary factor (using inflation to mean a shift of the whole scale upwards). Thus it was decided to refine the age additions by making them age related. At the same time, it was considered too complex to apply the age scale per player per event, so the age addition was worked out just the once and applied at the publication stage. So for juniors, what was in the published list was no longer their performance over the measured period, but their performance plus the age related component.

Undoubtedly change had again taken place by the 2000s, since grades under 100, previously rare, had become commonplace, also the negative grade issue was appearing. It's my opinion that these changes were not properly investigated, rather they jumped to the conclusion that "spread" was to blame and rushed off with the revaluation project.

[E Michael White]

Remember that juniors are treated as ungraded now when they are graded. Remember that grading is a two-step process:
(1) All ungraded players go through an iterative process to get an estimated grade
(2) Everyone's grade is calculated based on their previous grade, or the estimate provided by (1)

So, what you're seeing on your page for juniors and ungradeds, is the number associated with them after (1).

On the basis of what is written on the grading pages, the ungraded do not go through the reiterative process unless they played at least one ungraded opponent.

Do you happen to know whether or not the estimate, which comes out of the reiterative process, is set to 0 where the estimate was negative, before going into the last stage ? This would give a new source of anomalies.

[Alex Holowczak]

Do you happen to know whether or not the estimate, which comes out of the reiterative process, is set to 0 where the estimate was negative, before going into the last stage ? This would give a new source of anomalies.

My understanding is that the negatives remain after (1), and grades are only reset to 0 after (2). I'm not certain of that, though. One for Richard/Howard.

[E Michael White]

Surely you cannot get a minus grade for losing a game. We assume the minimum grade on the system to be "Nil" (as it clearly is otherwise a large number of these juniors would have a minus grade).

My understanding is that the negatives remain after (1) ( ed. the reiterative new starter process), and grades are only reset to 0 after (2). I'm not certain of that, though. One for Richard/Howard.

I expect this is the source of some anomalies. If AH is correct, you may see a grade of 0 but a negative number may be used in the calculation of other grades.

The ECF grading team, like their predecessors in the 1960/70s, seem to believe the major source of deflation anomalies to be improving juniors. In their haste to control this, they have produced a system for juniors, which spreads the grades and puts the top juniors where they think they should be placed but in so doing they don’t seem to be considering the deflationary effect of the juniors in the lower reaches. Half flooring the system in the way they have is similar to using increased junior k factors, in that both are flawed and won’t achieve the objects.

[Roger de Coverly]

The ECF grading team, like their predecessors in the 1960/70s, seem to believe the major source of deflation anomalies to be improving juniors. In their haste to control this, they have produced a system for juniors, which spreads the grades and puts the top juniors where they think they should be placed but in so doing they don’t seem to be considering the deflationary effect of the juniors in the lower reaches. Half flooring the system in the way they have is similar to using increased junior k factors, in that both are flawed and won’t achieve the objects.

This prompted me to look at my personal games for the most recent rating period. Of the 49 games counted, 10 were against juniors with * by their grade. The grades totalled 1786. Looking at the published grades for August, they totalled 1782. Two players were outside my 40 point cut-off, excluding those gave totals of 1532 against 1539.

Whilst I don't really see the point of recalculating grades for junior players already effectively in the adult pool it doesn't on the face of it appear to introduce a bias on my very limited test. Do you suppose they run the test across the whole set of data?

[E Michael White]

Roger

I wouldn’t expect much difference at your level of play. Most of your opponents will have a higher proportion of games against graded players in the calculation of their own * grade. Its lower down where there can be issues, when juniors largely play against other juniors and become underrated by some large amounts, as a result of the process. Unfortunately there aren't many obvious tests that can be done as standard statistical theory wont work when the scale is not uniform, as in the ECF grades, so quantities such as the mean and variance won’t hold water. You could try developing a curvilinear branch of statistics to fit the grading system but you could do a lot of work and find that the ECF board or council prefer the old methods.

[Paul Cooksey]

Half flooring the system in the way they have is similar to using increased junior k factors, in that both are flawed and won’t achieve the objects.

What is the objection to increased k for juniors?

I ask in all innocence. I was viewing this as a simple way to reduce the lag on the grades of improving juniors.

[Alex Holowczak]

What is the objection to increased k for juniors?

I ask in all innocence. I was viewing this as a simple way to reduce the lag on the grades of improving juniors.

If k is higher for juniors than adults, then consider the following:

(1) Junior beats adult. Total rating points in the system goes up, an indicator of inflation.

(2) Adult beats junior. Total rating points in the system goes down, an indicator of deflation.

(3) Junior draws with adult.
(a) If junior is lower rated, then junior's rating goes up more than the adult's rating goes down. Total points in the system goes up, an indicator of inflation.
(b) If junior is higher rated, then junior's rating goes down more than the adult's rating goes up. Total rating points in the system goes down, an indicator of deflation.

There are other effects with (3). A draw between players with different k could result in the player who started ahead of the other now being below the other player. This can never happen when k is the same. Nonetheless, repeated draws between the two will converge to a rating, but I can't work out in my mind whether this is the same rating as if the k was the same.

It could be argued that, if a junior is improving, (1) is more likely than (2), and (3a) is more likely than (3b). I've no statistics that show that is the case, but that's what I would expect. If the expectation is indeed reality, then if a junior has a higher k than an adult, then while yes, his rating will increase more quickly, you will also have the side-effect of creating an inflation within the system. This isn't a disaster - the differences are the key, not the magnitude of the ratings - but I would imagine that it's non-desirable.