The plans for the Grading System

General discussions about ratings.
NickFaulks
Posts: 8453
Joined: Sat Jan 02, 2010 1:28 pm

Re: The plans for the Grading System

Post by NickFaulks » Thu Feb 11, 2016 12:39 pm

Kevin Thurlow wrote: No - not if it's only one game.
Yes - even if it's only one game. That is the nature of statistics.

If the best estimate of your playing strength is 175 and then you win against a 180 opponent, it is quite reasonable to say that, in the light of new information, the best estimate of your playing strength has risen to 176 or 177. Nobody is suggesting that your actual playing strength has risen overnight, only that the best estimate, based on all known information, has risen.

Another way of looking at it could be that your new grade is a forecast of what your published rating would be at the end of the season if you didn't play any more games. That's meaningful, isn't it?
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: The plans for the Grading System

Post by Roger de Coverly » Thu Feb 11, 2016 12:41 pm

Kevin Thurlow wrote: Chess results are inconsistent.
Even if they don't use an Elo based system, which makes incremental changes, it would be likely that a minimum period and minimum number of games would be applied.

Part of the role of a rating system is to measure recent performance, so a bit of fluctuation reflects this. Once you report more frequently than annually, having the same games counted twice or more seems necessary, given that many events only take place once a year.

As far as reporting club championship results, reporting them to the ECF for grading should ideally be no more complex than informing participating club members. That would be reported more often than every six months surely?

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: The plans for the Grading System

Post by Roger de Coverly » Thu Feb 11, 2016 12:48 pm

NickFaulks wrote: Another way of looking at it could be that your new grade is a forecast of what your published rating would be at the end of the season if you didn't play any more games. That's meaningful, isn't it?
It would currently leave out leagues, but with Congresses now well trained to report results promptly, that's something that could potentially be implemented with minimal changes, apart perhaps from the uncertainties introduced by the presence of new players and more importantly the presence of juniors treated as new players. So assuming you played 30 games in the previous grading period, the new grade is your new result plus 29 notional games played at the previous period grade.

MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Re: The plans for the Grading System

Post by MartinCarpenter » Thu Feb 11, 2016 1:33 pm

Getting league results in in real time is easy enough. Yorkshire has been doing it happily for a few years now.

As for live grades, the big question is how heavily you weight recent results vs 'old' ones. A rolling 30 game window counts everything at full weight before dropping them entirely.

Chessnuts is time based and exponentially decays the weight of a game by 1/12 every month, so a 3 month old game counts for 77% of the amount in your current grade that a new one will. 6 months is 60 per cent, 12 30 per cent etc.

That 1/12 was chosen to be tidy rather than anything else :)

This approach firstly leads to people playing lots of games having hugely stable grades. That seems fine really. There is maybe an issue with people playing 10 games/year having perpetually jumpy grades. Some equivalent to the potential 3 year/30 game countback from the ECF system might be nice, but also fairly hard to do in a clean way.

Nick Grey
Posts: 1838
Joined: Thu Dec 01, 2011 12:16 am

Re: The plans for the Grading System

Post by Nick Grey » Fri Feb 12, 2016 12:11 am

Must admit going back to my junior years & having a month or two of lower grades I would have given up. That is probably a reflection of the competiveness of the school that I was at in every subject, sport or other activity - chess/bridge

The selling point to leagues was free system - so if not it may fail. Similarly have to deliver free system for clubs. Hopefully has some extras to make it interesting - like players being able to input games. Not sure that is the requirement & yes got a whole lot of hurdles.

I've come to live with my senior moments, lack of concentration, sugar levels in the wrong place, or too tired after work & travel. And not an excuse for any of my losses against any opponent (hopefully I have another quarter of a century playing left In me at what level who knows).

Some leagues & clubs & congresses will no doubt review team selections & orders & limits if it happens. Not sure what added value or costs.
Or any additional delays in getting a new player playing. Or a old player moving or coming back from a break.

Michael Flatt
Posts: 1235
Joined: Tue Jul 02, 2013 7:36 am
Location: Hertfordshire

Re: The plans for the Grading System

Post by Michael Flatt » Fri Feb 12, 2016 9:34 am

1. I understand that an informal survey had been carried to identify the various procedures and software in use to collect and submit results to the ECF grading system. Wouldn't it be helpful if the results of that survey were published in advance of the April Finance meeting to give background to members of Council?

2. Are there any Leagues, Clubs or Congresses that struggle to submit their results within the existing time limits?

3. The specification itself seems very detailed. Is it based on what might be supposed to be the preferred software identified from the survey of systems currently in use?

4. There don't seem to be any commercial terms identified in the invitation to tender.

5. Who are the members of the review board that will identify the winning bid? Do they have the necessary technical and commercial expertise?

6. What will be the criteria to determine the winning bid? It would be normal to include these in the invitation to tender so that the whole tendering process is 'above board'.

7. Does the ECF have sufficient resources to fund development of this software?

8. Will there be a further contract to maintain and update the software?

9. What documentation (User Manual, Programmer's Manual, etc.) will the supplier need to provide?

10. How will the software be validated and approved?

11. Is this software actually needed or is there an alternative solution?

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: The plans for the Grading System

Post by Roger de Coverly » Fri Feb 12, 2016 9:47 am

Michael Flatt wrote: 2. Are there any Leagues, Clubs or Congresses that struggle to submit their results within the existing time limits?
As an extension to that, how much time do local graders have to set aside in order to prepare and validate results for submission ? Whatever that time is, would on paper be multiplied by up to six for monthly submission, depending on how much graded activity took place over the summer months.

I get the impression that for Congresses, even arbiters who would feel lost without pairing cards to shuffle and argue about, have the resulting pairings and results input to a Congress management program which can supply grading files almost automatically.

Michael Flatt
Posts: 1235
Joined: Tue Jul 02, 2013 7:36 am
Location: Hertfordshire

Re: The plans for the Grading System

Post by Michael Flatt » Fri Feb 12, 2016 9:58 am

One of the most important features of a computer pairing program is the ability to automatically generate FIDE rating files. These can be accepted for ECF grading by making use of Howard Grist's results checking program.

Incidentally, pairing programs will pick up players rating/grading details directly from the FIDE and ECF grading lists.

Thus, Congresses will gain no benefit in using the proposed ECF League Management Software.

In fact, the need for an all singing all dancing LMS might be completely illusory.

Kevin Thurlow
Posts: 5821
Joined: Wed Apr 30, 2008 12:28 pm

Re: The plans for the Grading System

Post by Kevin Thurlow » Fri Feb 12, 2016 10:37 am

"As far as reporting club championship results, reporting them to the ECF for grading should ideally be no more complex than informing participating club members. That would be reported more often than every six months surely?"

Er, they look at the crosstable!

"Yes - even if it's only one game. That is the nature of statistics.

If the best estimate of your playing strength is 175 and then you win against a 180 opponent, it is quite reasonable to say that, in the light of new information, the best estimate of your playing strength has risen to 176 or 177. Nobody is suggesting that your actual playing strength has risen overnight, only that the best estimate, based on all known information, has risen.

Another way of looking at it could be that your new grade is a forecast of what your published rating would be at the end of the season if you didn't play any more games. That's meaningful, isn't it?"

I am not really disagreeing - I just think that the published amendment is more trouble than it's worth. I keep records every season and calculate my grade after each game, but that's really to look at trends over a number of games. That is not always a joy.

One issue is that if you have "stars barred" events and you are a couple of points below the limit, you might enter the tournament and book a hotel some way ahead (as accommodation can be hard to find), then a couple of days before the event, you go over the limit.

"The reason for going to monthly gradings is, as I have previously said, it would lead to about 10-20% more competitive chess being played.
I realise that is an unsubstantiated assertion. But we have evidence with the US and FIDE systems. Tell a 10 year old that they will get a grade in 6 months time is a real turn-off.
The live ratings for the top players attracts considerable interest, although it has no official FIDE status."

I don't see why people would play more. But Stewart is one of an ever dwindling band of people who know more about chess than me, so we have to take this seriously! If it does lead to more chess, than great. The numbers of games played will provide concrete evidence.

Anyway, time will tell - theoretically, monthly, weekly, daily, hourly grades would be great. But there are practical issues and I hope someone in authority takes those seriously.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: The plans for the Grading System

Post by Roger de Coverly » Fri Feb 12, 2016 11:08 am

Kevin Thurlow wrote: Er, they look at the crosstable!
How do you generate a crosstable, without knowing the results ?

Reporting should be as simple as making a list

Event, Round, Date, White Player, Black Player, Result

Stewart Reuben
Posts: 4542
Joined: Tue Apr 03, 2007 11:04 pm
Location: writer

Re: The plans for the Grading System

Post by Stewart Reuben » Fri Feb 12, 2016 11:10 am

Kevin > One issue is that if you have "stars barred" events and you are a couple of points below the limit, you might enter the tournament and book a hotel some way ahead (as accommodation can be hard to find), then a couple of days before the event, you go over the limit.

This is not a problem provided the entry form is properly designed. It is the rating/grade before the penalty fee date that is used. This may require some explanation when a player has gone above the limit after their entry has been accepted. The British Championships have this non-problem all the time as the grades and ratings change 1 July or 1 August. As there are always higher level events in the British, either the player can be promoted, or allowed to remain.

Trying again. Most people like to have a measure of their strength. Last night my team in Torremolinos played in a bridge event. We avoided coming bottom, just. In the grand scheme of things this is irrelevant. The prizes in bridge are only slightly above zero. We did well the first evening and came third. I was pleased to learn my results. Particularly in a chess Swiss, your raw score says very little. 5/9 tells us nothing. A TPR (Tournament Performance Rating) of 2100 will interest me, even please me. A TPG would be easy to do. Getting that information 6 months later would be of no interest to me.

With advances in computers, you could have each move graded! My variance must be from 3300 to 0.

NickFaulks
Posts: 8453
Joined: Sat Jan 02, 2010 1:28 pm

Re: The plans for the Grading System

Post by NickFaulks » Fri Feb 12, 2016 11:18 am

Kevin Thurlow wrote: One issue is that if you have "stars barred" events and you are a couple of points below the limit, you might enter the tournament and book a hotel some way ahead (as accommodation can be hard to find), then a couple of days before the event, you go over the limit.
This objection is thrown up a great deal, but cannot be difficult to solve. Just use a grading list long enough before the event to allow people to make plans. Yes, there would be people in the U190 playing off a current grade of 193, but that isn't difficult to explain.

The FIDE World Amateur Championship, for <2300 players, takes place in April and ignores lists after Feb 1st.

49 year olds are allowed to play in over 50 events on the grounds that they will be 50 at a defined date in the future. That really is odd!

edit : I see Stewart got there first.
Last edited by NickFaulks on Fri Feb 12, 2016 11:42 am, edited 1 time in total.
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: The plans for the Grading System

Post by Roger de Coverly » Fri Feb 12, 2016 11:23 am

Stewart Reuben wrote: This is not a problem provided the entry form is properly designed. It is the rating/grade before the penalty fee date that is used.
It's not a major problem when there's an event above the one you intended to enter. Suppose there's both an Open and an under 2040 or under 2050. Provided you are prepared to play in the Open, you can safely book travel and accommodation three months in advance even if you have a rating that might with plausibly favourable results go above the cut off. It's when, as at Hastings for the weekend events, there isn't an Open. You have to wait until you know the rating that's going to be used to determine eligibility.

I'd suspect that organisers might have to stipulate that monthly grading or rating list n would be used for pairings and rating or grading prizes, but the lowest of n, n-1 and n-2 would determine eligibility.

User avatar
Paolo Casaschi
Posts: 1186
Joined: Thu Jan 08, 2009 6:46 am

Re: The plans for the Grading System

Post by Paolo Casaschi » Fri Feb 12, 2016 11:37 am

Roger de Coverly wrote:It's not a major problem when there's an event above the one you intended to enter. Suppose there's both an Open and an under 2040 or under 2050. Provided you are prepared to play in the Open, you can safely book travel and accommodation three months in advance even if you have a rating that might with plausibly favourable results go above the cut off.
I think you are grossly exaggerating the issue.
Today we have grading lists published end January and end July. Do you really see events in February and August so badly impacted or have those organizers found a way to cope with the uncertainty of a new grading list published shortly before their event?

User avatar
Paolo Casaschi
Posts: 1186
Joined: Thu Jan 08, 2009 6:46 am

Re: The plans for the Grading System

Post by Paolo Casaschi » Fri Feb 12, 2016 11:53 am

Roger de Coverly wrote:It's now up on the main ECF site.

http://www.englishchess.org.uk/league-m ... evelopment
I have to admit I did not read the full tender document, but I thought of a slightly different approach to the problem.

The current proposal has an ECF owned event management system linked to the grading database; this has the drawback of upseting users of different event management systems and also frustrating developers of those other tools.

An alternative approach could be for the ECF to upgrade the grading website with public APIs to allow for automatic submission of results from compliant/certified applications. Then the ECF would allow the developers of all existing event management systems to upgrade their tools to use those API. Chances are the developers would include this functionality at no cost for the ECF, similarly to how they developed those systems in the first place. Costs for the ECF would be limited to defining and implementing those open access API, much lower than providing a full system with source code ownership. Chances are some developers of the existing systems will include the functionality, some will not; in any event end users will drive these developments and reward the tools that add the functionality of deemed useful. Users of tools that do not comply will have to submit results manually.

I'd be interested to hear from the people developing/maintaining these event management systems: would they take advantage of public APIs to submit result for grading automatically if those APIs were available?

Post Reply