Junior Grades?

[Andy Howie]

As regards your ECF comments, firstly I have never advoctated following their model and secondly even if I had chess hasn't died south of the border, so why would it here?

Watch what happens next year when it is implemented in it's current form.

[Douglas Bryson]

I agree that a rolling grade would not give an accurate reflection of the strengths in the scenario above, it isn't perfect. But neither is the current system, as if you substitute an adult for one of the above juniors then the junior will still be graded 1000 while the adult will have an increased grade.

With regards to deflation, I believe it could be dealt with by drifting adult grades. As adult grades are more stable, it makes sense to keep the points they have in the system at the same level while letting the junior points fluctuate. A weekly drift would be better and more inkeeping with a rolling grade than an end of season adjustment.

In the current system the adult will not be expected to score 50% against the junior. If the junior addition is +120 then the adult will be expected to score 0.337 on every game ie 10.11/30 The junior will genuinely have improved so his grade will go up since he will be expected to score 50% on the arithmetic but in reality will score more than 50%. The adult is stable and his grade will not go down since he has been compensated by the junior bonus points reducing his expected scores.

At the end of the grading period you will have two grades which add up to more than 2000 ie the system will reflect the fact the pool of players has improved in true ability - you have (in theory) grading stability. (The actual level of the additions may not be right but we're just talking of the theory here).

Will you get that with a rolling grade system that doesn't have additions? Wont you just get rapidly deflating grades as the juniors take points unfairly from adults who have not changed their skill level?

"I believe it could be dealt with by drifting adult grades" - not sure what that means, Gary.

[Gary McPheator]

Dougie, what I mean by drift is a redisdribution of the points lost by adults to juniors back to the adult players as a whole on an equal basis. I disagree with junior additions because of their arbitary nature. While it works at the general level, it does not at the individual level. Players will improve at different rates and at different times through out the season. With a rolling grade, adults who lose to juniors will take the individual hit of the junior's grade but the junior's grade will be rising through the season.

[Andy Howie]

Gary,

A few years ago, a statistician did a study of our grading system and found it to be statistically sound. We have mechanisms in place to counteract grading inflation and deflation which are the scourge of the FIDE system. As Dougie has pointed out, we have a mechanism in place that allows for the improvement of Juniors. I fail to see what the benefit is of having a rolling grade system.

For the grading system to work, it requires a sample size, that is the nature of the grading. The calculation uses n as the sample size where n is 30 unless more then 30 games have been played, on those occasions then n is the number of games played. FIDE has a similar idea, the K factor. Using rolling grades in the calculation would cause grades to fluctuate widely and destroy the stability and mechanisms we have in place.

As I mentioned earlier, out of interest I simulated what would happen if we were to grade juniors twice a year instead of once. I picked the arbitrary dates of 1 July and 1 Jan. I deliberately picked the top 20 junior risers from last season as that should have given the greatest change in final grade. I did this as I have been an advocate of 2 gradings a year for juniors for some time. To my surprise, it had little to no effect. The Final grade was always in the same ballpark as this years published grade.

Several of the Juniors went +200 points from their starting grade in both periods and given the games played, I used a value of 30 for N.

Thinking this through logically, it should have meant that we saw far greater grading rises. We didn't. What that tells me is that the current system we are using for juniors works. What puzzled me is why the following season, their grades took off again. By watching my daughter learn how to read and write at school, it dawned on me. The juniors are continually learning. I have seen Eilidh go from not being able to write (a beginner) to being able to write a sentence in a year. That is her starting point for this year and she will build on that. The same happens in chess.

I also tested by removing the junior additions for games against juniors, unsurprisingly it meant the final grades dropped. I'd argue this would exacerbate the problem.

I have now changed my opinion and I see no merit in having 2, 3 or 4 grading periods a year. Statistically it only appears to marginally benefit people who play 60+ games who do not increase their grades by 200 points.

[Gary McPheator]

Andy, to make a grading system to be as sound as possible surely we need to compare like with like. The current system compares people who play a different amount of games as well as giving some people arbitrary bonus points.

A rolling grade would also use a sample size, but it would be the same for all players.
It could be set at the 30 last games played although I would prefer a lower figure of 15 or 20 to be in line with the amount of games played per season by more players as well as allowing outlying results to be discarded quicker.

Juniors will gain more points when playing adults than they do at the moment which should along with up to date grading counteract or surpass any loss of points that they may occur when playing fellow juniors.

I am not suprised you found little difference in grades in your simulation. Whether that makes me illogical I couldn't possibly comment on!

[Andy Howie]

Hi Gary,

I was surprised. I was under the belief that the more games a junior played in a season, the harder it was towards the end of the season to affect that grade. I was pleasantly surprised when I ran the simulation to find that this was not the case.

I didn't quite understand what you meant by rolling grades, now I do from your last post. This is the system they use in the ECF at the moment!

Mathematically the use of N = 30 keeps the grade change between+- 24.5 grading points for games 1-30

If we change to 20 it becomes +- 36 and 15 becomes +- 48! (using change = 800*(actual-expected)/n )