Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Junior Grades?
Quote:Why is it that juniors' grades are so "under-rated". Surely it creates an inconsistency with the grading data, thus compromising its statistical accuracy? Perhaps something which needs looked at? Anyway; that's for another topic.

Where does this come from? Have you just been toasted by a Jx ?
lol No! I haven't played Chess since Lanark actually. Big Grin

It was just a general thought: you always hear people talking about juniors' grades and how under-rated they are and perhaps it's time we did actually consider it? It's probable that juniors improve more quickly than established adults so perhaps updating their "published" grades more frequently would be a good idea?

That is what the junior addion addresses.
Does it though? That's the point I was making.

Incidentally: is the junior addition added to a junior's grade when they enter a Congress - in order to determine which section they can play in. I don't think it is?
Well, it partly does, but juniors are always given the grade they 'earn' in the previous season. The 'underrated' thing is mostly a function of having a grading system which is (for excellent reasons) updated once a year. Fast-improving players, a disproportionate amount of whom are juniors, are always going to seem (be) underrated, but that's what the additions (and the 200up rule for those of us who fall victim to the highest risers) address.

The additions were presumably worked out according to some vague statistical analysis at one time or another, and it might be worth re-examining that to see if things have changed significantly (if that isn't the sort of thing that goes on anyway), but I'd caution against any assumption that they aren't fair. There's a temptation to focus on what I suspect are the relatively few cases of people improving considerably above the junior addition, while ignoring the less visible cases of plenty of juniors whose ratings are improving 'only' by the standard number of points, or less, a year.

Sure, every now and again we all get sandbagged by someone shooting up at a rate of knots, but that's a) broadly covered by the 200up rule and b) one of those things, I think. Overhauling the entire grading system seems unnecessary.

The junior addition isn't considered in sections as far as I know, no. Don't think it should be either, learning how to win events is a worthwhile skill, and they don't need any more encouragement to enter events above their station =)
The junior addition works reasonably well with most juniors - it assumes the junior's grade will increase by about twice the addition for his/her age over the course of a season, so that if games are spread evenly over the season the average disparity will be roughly the amount of the junior addition, and on average the junior's opponents will get appropriate compensation. The values of the junior additions (+120 points at ages up to J12, declining to +20 at J20) were set decades ago, and I think they are still realistic enough to continue to be used.

The problem concerns those talented juniors whose grades are rising far faster than the rate assumed in the junior addition. Already this season, in the space of only 3½ months, three players (Kai Pannwitz, Euan Gray and Andrew McCusker) have triggered the "200-up" rule, meaning that their current grade is no longer linked to and constrained by their start-of-season published grade. They have, effectively, achieved a full year's expected improvement in under four months. And there are quite a few other juniors who will soon go 200-up.

Where the 200-up rule kicks in, opponents will (at the end of the season) be considered to have played against these juniors at their end-of-season grades rather than at their start-of-season grades, which will be more beneficial (or less damaging) for the opponents than the normal operation of junior additions would have been. But even under the 200-up rule, some junior grades will be seriously understated, because early-season results will still be given the same weight as late-season results - if a player starts the season playing at a strength of 1000 but finishes at a strength of 1500, and the numbers of games played early and late are about equal, the grading system will give the player a grade of 1250 at the end of the season. The player will therefore go into the following season with a published grade of 1250, but a true strength of 1500+ and rising! Even the maximum junior addition comes nowhere near bridging that gap.

In the long term, players' grades do generally reach a level that is a fair indicator of their playing strength, but it is probably not good for the grades of some (even if not all) juniors to lag far behind their true strength.

Publishing juniors' grades twice a year rather than once would reduce the discrepancy between published and actual strength, and might ensure that published grades caught up with actual grades more quickly. Another approach might be to use only a junior's most recent 30 (or even 20) games in calculating his/her grade. How easily such calculations could be incorporated into the CS grading system is another matter.
My comment in another thread about Juniors was meant in jest. To us old yins, the bit about there being two of them is a reference to an extremely old joke, so old that perhaps I am the only person who recalls it.

As far as the Minor section is concerned, the more Juniors who play up a section, the better. In fact I think that I shall propose a new rule, anyone who scores more than 50% in a section, can never revert to a lower section. That should leave the problem of these 'pesky kids' to other, more highly graded players.

Having played, and lost to, many of the youngsters in question, all I can say is good luck to them all. Oh yes, and remember the RAF motto 'Per Ardua ad astra.'
Also worth noting that the junior additions cannot be accurate immediately at the beginning of a season. How can a junior who has just earned a new published grade suddenly be 120 elo higher than that grade - because surely if that were true then their new published grade would be higher? Immediately after grades are published should statistically be the time when they are the most reliable. Whereas by the end of a season their grades could be wildly different from their actual playing strength.

As stated, fixing this "problem" (if it even is a problem), might be an entirely different matter. But before that could ever possibly even be considered, we have to first establish if there is a problem or not. Perhaps things have changed from when it was first devised? Smile

This should be an interesting discussion. =)
The comments above cover most of the issues regarding the current situation.

The junior additions we currently have are the result of analysis of how juniors performed on average in the course of a year. This was done several years ago and has not been updated recently.

The system makes some assumptions which are obviously not applicable to all players but are simply the best guess of what happens on average. It assumes adults do not improve and are stable from year to year and it assumes that juniors will increase by these average number of points.
So by all means these ideas could be tweaked to feed in junior points with a bias towards the end of the season. However you would need a date marker on every game (we dont have that currently on league and club championships). Adults do improve up to the age of about 35 and then post 40 its all downhill - on average. So a young adult addition and an older player subtraction may make the system more accurate (albeit not that attractive an adjustment to sell to the established regulars).

The drift figure which you see on calculations is meant to add back points which may have leaked out of the system. The 200 up idea is to adjust the most wrong grades quickly to a more appropriate level.

The twice a year or more processing has been discussed several times. You need answers to all the practical issues related to it before you decide the change is worth it.

Regarding whether tournaments use junior additions for entry requirements. That's entirely up to the tournament what it wants to use. The live grades rather than published are used by many junior events now.

It would be useful if there was output listing all games with actual and expected score. We could then see exactly how adults are performing v juniors. eg I just did a quick (manual) run through of all the adult games v juniors on the bottom two sections at Lothians - over 52 games the adults scored 27 but were expected to score 30.4. However that's a biased set of data since Lothians results were one of the tournaments where the perception of underrated juniors was identified.

What is needed is data for several full seasons with every game v juniors not just the good juniors who play weekenders.
To Andy McCulloch:

I recognised the "trap" joke - I remember hearing it in 1967, though it wouldn't surprise me if it was actually far older, and was simply updated from time to time depending on who is fighting whom.

Going back to junior grades, the fact is that for a fast-improving player the published grade is already about six months out of date at the start of the season, and in some cases the difference between published grade and true strength is greater than the junior addition even at that time. As anyone who has watched them or played against them can testify, there's no way Kai was a 1230 at the start of this season, or Andrew a 1217 - their current grades (calculated under the 200-up rule) of 1620 and 1516 respectively are probably not far away from their true start-of-season strength (their junior additions make them only 1350 and 1327).

OK, we're talking about a small number of players, and there would be no point in putting a lot of effort into changing the grading system to make a small number of grades more accurate for a time. But if one or two quick and easy tweaks to the system would increase the accuracy of fast-risers' grades, that might be of lasting benefit.

Forum Jump:

Users browsing this thread: 3 Guest(s)