Chris Majer returns!

[Alex Holowczak]

Leagues will presumably catch on with submitting grading results twice a year once the system is established.

Well, every League and County has been informed by e-mail. Graders will be informed by Richard H at some point, I'm sure. So if the leagues remain oblivious, I'm fairly sure the graders will pass the message on.

[Roger de Coverly]

Will this make much difference to the grades? Even if there are few results for the 6 month period it will still be averaged over 30 games.

It isn't clear that it will be averaged though. Even in the current system, the practice is now to go backwards in time to count the most recent 30. It doesn't matter for A grade people (30 games in one season), but for B and below, games played at the end of the season ( March, April, May) appear more frequently in the grade than those at the start of a season. With two cutoffs a year, this effect is amplified.

General reasoning suggests that on average the effects on consistent active players should be limited. Treatment of new players and by implication juniors will be different, the effects of this remain to be seen (or not if they don't parallel run).

[Mick Norris]

Leagues will presumably catch on with submitting grading results twice a year once the system is established.

Well, every League and County has been informed by e-mail. Graders will be informed by Richard H at some point, I'm sure. So if the leagues remain oblivious, I'm fairly sure the graders will pass the message on.

The Manchester League grader has been informed, but don't know in what capacity or by whom

[David Sedgwick]

Traditional theory on how to implement changes to computerised calculation systems would indicate the need for a parallel run.

So after processing the May 2011 files using the yearly approach, you then subdivide the data into before and after November 2010 and produce both a hypothetical Jan 2011 list and a parallel July 2011 list. You then compare the results of the actual with the parallel. If you get significant major differences, some rethink may be in order. You might even want to continue the parallel processing with the January 2012 list.

This was one of the main reasons for deferring the change for a year. However, it would appear that no such exercise is planned.

If the exercise were to be undertaken, it wouldn't give the same answers.

Someone who plays 29 games in each half year period will still have a A grade and each of his/her grades will be based on the 58 games played in the last twelve months.

However, someone who plays 31 games in each half year period will have an X grade and each of his/her grades will be based only on the 31 games played in the last six months.

That's how things currently work with the Rapidplay List, but, as you said up thread, no-one cares enough about his/her Rapidplay grade for that to matter very much.

It will be very different with Standard Play grades if this bizarre anomaly is not addressed.

It's entirely possible that I've misunderstood and that what I've written above is competely wrong. However, I've raised the point with more than one of the ECF officers responsible and it's not been denied so far.

[Roger de Coverly]

Well, every League and County has been informed by e-mail.

You might hope that if leagues and counties were paying attention, they would already be aware that the ECF was going to approach them for more frequent grading data at some stage. This might cause timing and date-stamping problems for adjudication or adjournment leagues or those whose result management systems didn't include automatic or near automatic production of computer grading files.

The wider question is to ask how the ECF are going to use the additional data and what are the likely practical consequences? I'm reminded that the CEO wrote http://englishchess.org.uk/farthing/?paged=2

The grades will be calculated in the same way, so the length of the rolling period being used for grading will be the same.

That's not my understanding, at least not for those who play 30 games or more in a six month period, but the point lacks official clarification.

We managed to get a description of the rapid-play system, namely that the software goes backwards in time for a maximum of six half years until it finds sufficient events to accumulate at least 30 games. The term "event" being used because a 5/6 round tournament is either completely included or totally left out.

[Roger de Coverly]

Someone who plays 29 games in each half year period will still have a A grade and each of his/her grades will be based on the 58 games played in the last twelve months.

I had the different impression, namely that the software would go backwards to find the game or event last played in the previous grading period. That's the way it's said to work for B,C,D and E grades at the moment. It is a bit important to clarify these things, particularly with the parallel plan to put detail results on-line in the style of FIDE.

The 29 game example is a good one, would it mean that January 2012 was based on 29 games from November 2010 to May 2010 and 29 games from June 2010 to November 2010? If so, then it's more "out of date" than one based on 29 games from June 2010 to November 2010 with either 1/30th of the August 2011 grade added in, or the very last game or event of the 2010-11 season.

[Ian Thompson]

... or the very last game or event of the 2010-11 season.

If it's last event, how you choose that event is not obvious. Many players will play in two or more leagues which officially finish on 31 May, so which league do you choose? What happens if a league finishes on 31 May, but the last game the player played in it was weeks or months (for an occasional player) earlier, so he might have more recent congress games than the last league game he actually played?

[Alex Holowczak]

This might cause timing and date-stamping problems for adjudication or adjournment leagues or those whose result management systems didn't include automatic or near automatic production of computer grading files.

I'm privately campaigning to shift the cut-offs from the end of November and May to the end of December and June. Hopefully this won't cause so much of a problem, because I think even the most staunch adjourners would avoid adjourning over the Christmas/New Year period. At the very least, there'll be far fewer of them than there would be a month earlier.

[Steve Rooney]

Someone who plays 29 games in each half year period will still have a A grade and each of his/her grades will be based on the 58 games played in the last twelve months.

I had the different impression, namely that the software would go backwards to find the game or event last played in the previous grading period.

I too thought the system counted back to get at least 30 games not the whole of the previous grading period. But whilst it's good to know precisely how the calculation is done, I'm still not clear how much difference it makes as the grade is worked out from a reasonable number of games, not just a handful.

Players grades can go up and down and it's only a guide to past performance not a precise estimate of how well they are going to play when you face them over the board. Mountains and molehills come to mind with some of the comments being made here.

[Roger de Coverly]

If it's last event, how you choose that event is not obvious.

If I understood Richard Haddrell correctly, he said that every game has an "event date". It's very obvious for a one-day tournament, perhaps less so for a league or Congress. A league might date its games by the actual day they were played or by the start or end of the month or start or end of the season. There's a need to standardise these conventions, but the sorting method was to order games by date and take the most recent ones first. So if two leagues both dated games to 31st May, all games with that date would be included.

I was referring to the revelations in this thread http://www.ecforum.org.uk/viewtopic.php?f=6&t=2572

That thread infers that the going back in time rule only applies to rapid-play games. But it's also been asserted that the six-monthly grades should use the system established for rapid-play.

I'm sure many of us have seen IT horrors at first hand. Implementing new systems without a clear idea of how you want them to work is usually a recipe for disaster.

A system where you

(a) Established a performance rating over each six month period
(b) Published a grade based on higher weightings to the most recent performances, up to and including just using the most recent six months

would be reasonably logical.

It wouldn't be the same as currently, for example a player with 40 games, but 20 in each half year would be counted as two-thirds the most recent six months and one-third the six months before that.

[Roger de Coverly]

Players grades can go up and down and it's only a guide to past performance not a precise estimate of how well they are going to play when you face them over the board.

It's a statistic that's used to

determine board orders
seed pairings in tournaments
determine eligibility for events
determine eligibility for prizes
award titles ( only ECF ones)

Players regard it as important that it should be correctly calculated. If you don't know what "correctly" is supposed to mean, confidence in the system is lost.

[Roger de Coverly]

I'm privately campaigning to shift the cut-offs from the end of November and May to the end of December and June. Hopefully this won't cause so much of a problem, because I think even the most staunch adjourners would avoid adjourning over the Christmas/New Year period. At the very least, there'll be far fewer of them than there would be a month earlier.

For leagues, moving the cut-off to the end of June wouldn't make much difference. It would at least get late-running cup finals into the correct season. It puts pressure on the grading team though, given the expectation of new grades to become available at or before the British.

A December 31st or 24th cut-off would make sense for leagues because it coincides with a period of low activity. I'd suggest 24th rather than 31st do as not to split Hastings in two. I have no recent experience of serial adjourners, aren't they likely to want to postpone resumptions of games from late November and early December into January?


(edit) Some of the same ground was covered in http://www.ecforum.org.uk/viewtopic.php?f=4&t=2310 where I started by asking whether there was a precise proposal on how six-monthly grades would operate. We seem little the wiser six (!) months later. (/edit)

[Mick Norris]

See here for info:

http://www.sccu.ndo.co.uk/grad.htm

[Richard Bates]

See here for info:

http://www.sccu.ndo.co.uk/grad.htm

I don't understand why anyone thinks it is a good idea to have two officially endorsed grading lists in operation at the same time. If the January list is to be seen as a valid, indeed superior, list in its own right (which is the implication of grading over the six month period as opposed to operating a rolling list, or alternatively characterising January as an "update" with the August list being calculated over a full year as before), then grading limited competitions/leagues using the August list are going to run into disputes.

Also seems a bit strange that leagues will be allowed to submit all their games at the end of the season and have them graded in the August list. Clearly it undermines the principle of six monthly lists, but for leagues intending to use one list (the August one) it is an entirely rational thing to do! For the accuracy/relevance of the grading list that they use, they would obviously want all games played in their league included in that grading list.

From one perspective, I find it vaguely bizarre that it says that the ECF are intending to use the August list for the national stages of the Counties Championship. Which doesn't start until April!

[Alex Holowczak]

I don't understand why anyone thinks it is a good idea to have two officially endorsed grading lists in operation at the same time.

We do with FIDE-ratings in the 4NCL. The season lasts longer than the rating list. It doesn't cause that much difficulty.

Also seems a bit strange that leagues will be allowed to submit all their games at the end of the season and have them graded in the August list. Clearly it undermines the principle of six monthly lists, but for leagues intending to use one list (the August one) it is an entirely rational thing to do! For the accuracy/relevance of the grading list that they use, they would obviously want all games played in their league included in that grading list.

If a league doesn't submit its results in January, how are the graders supposed to know? It might have lapsed, for example. So there'd be no results to actually chase. So they submit in one big lump at the end. Is Richard then supposed to refuse the games because they didn't submit the pre-Christmas games pre-Christmas? That would be churlish. Besides, the precedent is that if a grading file is submitted late, then it ends up being graded as an event late to report anyway. So these would be events reporting late. How do you distinguish between the two types of late reporting?

From one perspective, I find it vaguely bizarre that it says that the ECF are intending to use the August list for the national stages of the Counties Championship. Which doesn't start until April!

Remember though that the Union stages start in October(ish) the previous year, and would use the June list. It would be ridiculous for someone to be able to play in (say) the Under 140 team at the Union stage with a grade of 139, then get kicked out of the team at the national stage if his grade rose to 141 in the January list.