The plans for the Grading System

[NickFaulks]

Nick,
I am reluctant to comment on private correspondence splashed onto the web, but I will help you out here. The clue is that YGS is on the chessnuts site. If you look at the leader board for the ChessRatings2 competition on the Kaggle website the evidence is there. Chessnuts is at 71 and Actual FIDE ratings benchmark is at 81.
Brian

Thanks, Brian. I know nothing about Chessnuts, but shall educate myself.

[Brian Towers]

What is the point of a monthly ECF grade?

It would make sandbagging, a topic in vogue on several other threads, much more difficult. If you have to get your rating "back in line" in the same month that you win a couple of weekenders with 5/5 that's going to be a lot trickier than if you've got 6 months to do it.

Leagues which insist on board orders being related to grading will just say the August list applies throughout the season, otherwise players will suddenly become ineligible for their usual team and there will be more defaults.

Bwahahahahaha!
I think you'll find clubs and league secretaries are a lot more intelligent than that. If a player is fast improving, either because he is a junior or perhaps hasn't played for a while and starts off ungraded, then rules which impose a maximum amount by which a lower board may outgrade a higher board will lead to the naturally just result of such a player moving up through the boards as the season progresses and their improvement is reflected in their monthly grades.

It would also make sense for the league to say in your example that in an U145 division grades at the beginning of the season apply to the U145 limit with the caveat that if a player improves to the extent of having a grade over 160, say, then he begins ineligible for the league.

We have a kind-hearted individual who arranges for the internal club events to be graded - if we tell him he has to do it on a monthly basis, he might be less kind-hearted.

I'm curious. What does "arranges for the internal club events to be graded" mean? Does he actually sit down himself and, like an old fashioned constipated mathematician, work them out with a pencil on paper? Does he collect the results and send them off? Does he feed them in to the computer? Does he just press the "go" button? More to the point, have you paid any attention to how the process is planned to work? Club captains and club secretaries will send off the results using a supplied computer program (League Management System, or whatever it is called) which will communicate directly with another computer program which will run automatically to calculate and generate the new grades. Or are you suggesting that the "kind-hearted individual" will be less kind-hearted because he will be redundant?

How long does a typical internal club event take? Is it typically over and done with in a month or does it drag on through half the season or maybe even the whole season? The proposal is that the new system should be loosely based on the FIDE system. I see you're a newly qualified FA. Do you have the faintest idea how the FIDE system works? Suppose the internal club championships were FIDE rated (something which is common in other countries, by the way) how would FIDE handle it? The standard FIDE approach is that competitions lasting longer than 90 days should, for reporting purposes, be split into sub 90 day segments. The season long club championship would have to report results quarterly rather than monthly. Of course the ECF doesn't have to slavishly copy every dot and comma. They could say that long running competitions have to report results once every 6 months or even once a year.

[Roger de Coverly]

Chessnuts is at 71 and Actual FIDE ratings benchmark is at 81.

Wasn't that Kaggle competition somewhat vulnerable to being gamed?

I believe the entrants were presented with a list of results and asked to predict results for games where they weren't told the results. If however you deduced or made the assumption that all games were from seeded swiss tournaments, you could reverse engineer to get the pairings and thus work backwards to find the earlier results that fitted these pairings.

[Carl Hibbard]

A tender with ownership of the actual source code is going to be rather expensive.

[Andrew Zigmond]

No one uses the live Yorkshire grades for that! The justification Jon gives for the live Yorkshire grades is maintaining peoples interest which seems slightly odd to me, but probably believable.

This isn't quite right as several teams have used live grades for board order purposes within Yorkshire (the Yorkshire League has a twenty point buffer which allows a reasonable amount of flexibility).

For a long time I was never a fan of live grades and was one of those who managed to get them removed from the main online grading list. As a captain I preferred a single season long grading list to work to; not only because it prevented squabbling about who should be in what team but because it allowed a stable board order throughout the season. However I suppose we ought to roll with the times; we live in an era of up to date information and if the rest of the world has up to date grades the UK should follow suit.

As others have said the problem with live grades is the quality of the reporting. I suppose you need to accept that some grades are going to be more live than others and if a player is unhappy that his grade isn't accurate due to a league secretary who isn't computer literate or a club secretary who thinks its easier to collate championship games at the end of the season then they need to raise that issue locally, not with the ECF.

[Roger de Coverly]

A tender with ownership of the actual source code is going to be rather expensive.

It's now up on the main ECF site.

http://www.englishchess.org.uk/league-m ... evelopment

From the ECF's viewpoint isn't the source code issue, one of future support ? Operating systems seem to have stepped back from the notion of backwards compatibility, one of the ECF's non-exec directors recently finding that an old BCF grading program would not run on a new computer.

If the problem is defined as one of getting league and other results into the grading system database on a monthly basis, is it taking the long route to build a league management system under the ECF's control when locally managed systems already exist and perhaps only need tweaking to provide a hassle free download into the ECF grading data? The assigning of new grading codes is possibly one of the bigger problems for the existing systems.

[Steve Rooney]

Our league amended the rules for this season to use the July grades for eligibility in teams throughout the season, but the latest list for board order. Some argued that we should stick with one list throughout, but it was felt logical to use the January list after it was published for board order only. I imagine the same would apply if there were more frequent lists.

If more frequent lists takes us on the road towards ELO-style grades then I think it would be a good step.

[MartinCarpenter]

As others have said the problem with live grades is the quality of the reporting.

No, there's a fundamental issue with live grades too. Most of the changes in any system of live grades are inevitably going to be noise, not information.

Which is fine just so long as you don't take it too seriously/make it formal! People often have a difference of 15 between their top and bottom live Yorkshire grades through a season, so even the generous +-20 isn't quite as much leeway as you'd think.

As for the kaggle competition, I can't imagine Jon gaming it (and checking online it was definitely Jon.). Chessnuts is basically BCF with a weighting towards recent results. The competition is seemingly to be about historic data then next three months results, which I'd have expected Chessnuts to do well.

I'm rather less sure how well it copes with things like the summer break people tend to take - doubt if 'micro' form swings often persist from Spring to Sept/October.

[Michael Farthing]

It would also make sense for the league to say in your example that in an U145 division grades at the beginning of the season apply to the U145 limit with the caveat that if a player improves to the extent of having a grade over 160, say, then he begins ineligible for the league

Can produce problems in smaller clubs like mine. You may find yourself unable to put out a team as a result: something which detracts from all teams in the league if the purpose of the activity is to have some good games of chess rather than winning a trophy on the basis of defaults.

It could also be argued that maybe the up-and-coming player has profited from lessons learned in the division and that this learning process is part of the event - if in the earlier part of a game you gain an insight into chess you are not regraded on the basis of your better skill and you can use that better skill in the second half of the game. Surely the same should apply to the first half of a competition and the second half?

[Roger de Coverly]

The competition is seemingly to be about historic data then next three months results, which I'd have expected Chessnuts to do well.

I'm rather less sure how well it copes with things like the summer break people tend to take - doubt if 'micro' form swings often persist from Spring to Sept/October.

It would be likely to cope better with overshoots, that's where typically someone on exceptionally good or bad form plays an intensive number of games over a short period, say the summer months. You are now seeing this with juniors, that by playing frequently during July and August they can take advantage of (K=40) to translate their improved play directly into rating points and the lower level OTB titles.

[Roger de Coverly]

Surely the same should apply to the first half of a competition and the second half?

If you had monthly grades, most, perhaps all leagues would define eligibility by reference to a fixed point. You might use 1st September to reflect the play of those active over the summer.

It would be less likely to achieve a consensus on board orders, if regulated, and on new players acquiring grades during the season. For its county matches, the SCCU requires ungraded players to be assigned a grade by the controller to make them eligible for grade restricted competitions. This is withdrawn if the January list takes them over the limit for the section.

[Michael Flatt]

A tender with ownership of the actual source code is going to be rather expensive.

It's now up on the main ECF site.

http://www.englishchess.org.uk/league-m ... evelopment

It shouldn't have been put out to tender yet. Currently, the Home Director has no authority to do this as he has not obtained the agreement of Council.

As Carl pointed out this could turn out to be an expensive exercise. The Budget is normally presented to Council in April.

[Mick Norris]

It shouldn't have been put out to tender yet. Currently, the Home Director has no authority to do this as he has not obtained the agreement of Council.

As Carl pointed out this could turn out to be an expensive exercise. The Budget is normally presented to Council in April.

Ok, but the sensible question in Council in April would be how much will it cost the ECF? So surely the tender process is an attempt to answer this problem

I'll be especially interested to see if John Upham tenders, given the Manchester & London leagues amongst others both use his system

[Roger de Coverly]

It shouldn't have been put out to tender yet. Currently, the Home Director has no authority to do this as he has not obtained the agreement of Council.

It could be argued that it was implied by his re-election. But it's a strategic policy direction for the ECF, not one that should be dependent on the election or otherwise of an individual.

The Budget is normally presented to Council in April.

If you don't know what it will cost, getting agreement to a budget to do it would be more problematic.

The regrading exercise of a few years ago would have had a cost, of volunteer's time, if nothing else. By contrast that went ahead without formal Council approval.

[NickFaulks]

I'll be especially interested to see if John Upham tenders, given the Manchester & London leagues amongst others both use his system

I think that's an excellent system, but how will he feel about losing ownership of the source code ( which I presume he has )? It will be a shame if that requirement means that someone has to reinvent this wheel.

If I were to design a League Management System - which of course I'm not - I'm sure it would end up looking suspiciously like John's. Would that lead to trouble?