Future Grading Systems

General discussions about ratings.
Jon Griffith
Posts: 10
Joined: Sat Aug 11, 2007 11:01 am
Contact:

Future Grading Systems

Post by Jon Griffith » Mon Nov 29, 2010 5:14 pm

Future Grading Systems

I have been developing the chessnuts grading website for the Yorkshire CA for the last five years.
http://www.chessnuts.org.uk

The objectives of chessnuts are broader than the previous grading system:

Grading - more frequent grade updates
Access - all data on open web pages
News - timely publication of fixtures, results and tables
Management - automate clerical tasks for event organisers

Priority has been given to handling leagues, which generate most results (19,000 per year). Of the eleven Yorkshire leagues, the largest seven are now entering their own fixtures and match results directly into chessnuts online forms. In most cases the league secretary enters result details within a few days of play. The organiser benefits in that chessnuts takes care of the league table and of publication. The players benefit by finding fixtures and current results on-line; and the grader benefits because the grading input data is exposed to immediate public scrutiny. Next year team captains will be filling online match cards for the approval of the league secretary.

Congresses provide 10,000 results annually. Chessnuts now computes swiss cross-tables, and in future the prize-lists could also appear. But yours truly still spends far too much time copy-typing pairing cards. A bottle-neck here is that english arbiters have not yet found an acceptable swiss pairing computer program. For what is possible look at http://www.chess-results.com/ Why does Yorkshire not use electronic swiss management when there are 82 nations who do?

Club events contribute 2,000 results, but chessnuts does not yet really help them. Eventually it should be possible for an organiser to register his event, and publish his fixture program online. Players should be able (if they wish) to report their results to him online, and there should be an online page that looks just like the result-sheet on the club notice-board. One day, the sheet on the notice-board will be printed from the web. Nobody will ever again have to make a grading submission when the results are already in the server.

Which brings us to Live Grades, which were launched on chessnuts this month. Using a new grading algorithm, a robot recalculates all Yorkshire players' grades twice daily, and the grading list shows "Live" grades alongside the "Official" grades. In fact, the Official grade is the Live grade from 31st May. So we are liberated from the annual grading report deadline, which might eventually be replaced by "report your result within seven days". But the new algorithm has not yet passed the test of time, and players and organisers have yet to come to terms with the idea of grades which change while you watch.

So until my peers and betters decide otherwise, live grading is interesting but experimental. Meanwhile I enjoy tilting at shibboleths.
Jon Griffith
Yorkshire Grader

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

Re: Future Grading Systems

Post by Roger de Coverly » Mon Nov 29, 2010 9:03 pm

Jon Griffith wrote: Which brings us to Live Grades, which were launched on chessnuts this month. Using a new grading algorithm, a robot recalculates all Yorkshire players' grades twice daily, and the grading list shows "Live" grades alongside the "Official" grades. In fact, the Official grade is the Live grade from 31st May. So we are liberated from the annual grading report deadline, which might eventually be replaced by "report your result within seven days". But the new algorithm has not yet passed the test of time, and players and organisers have yet to come to terms with the idea of grades which change while you watch.
Some time you should tell us what the spec is - both how far back you go to calculate today's grade and whether the live grade has any effect on future calculations.

Those of us who have played on-line are familiar with continually moving grades. If you are trying to organise a league or congress, you do however need grades which don't change in the run up to an event. FIDE's approach of a fixed list published bi-monthly or monthly is the more practical approach for these purposes. FIDE do of course show the results to be included in the next monthly update, so they are pseudo-live.

For play not on a server, you will always have the problem of results being reported in the wrong order. So if a player plays in a league game tonight and finished a weekend congress yesterday, what is the impact of the league game being reported first?

Jon Griffith
Posts: 10
Joined: Sat Aug 11, 2007 11:01 am
Contact:

Re: Future Grading Systems

Post by Jon Griffith » Mon Nov 29, 2010 10:05 pm

The full spec of the grading method, and some comments are here
http://www.chessnuts.org.uk/ny5/Rulesof ... ading.html

I would be happy to discuss the method, but perhaps we should start a separate thread.
Jon Griffith
Yorkshire Grader

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

Re: Future Grading Systems

Post by Roger de Coverly » Tue Nov 30, 2010 1:35 am

Jon Griffith wrote:The full spec of the grading method, and some comments are here
http://www.chessnuts.org.uk/ny5/Rulesof ... ading.html
Noting that you use 5 years of results, does the daily grade discard results from five years ago on a daily basis? Don't you then suffer from the problem with any moving average grading system that the grade can change without a single game being played?

Brian Valentine
Posts: 574
Joined: Fri Apr 03, 2009 1:30 pm

Re: Future Grading Systems

Post by Brian Valentine » Tue Nov 30, 2010 5:10 pm

There are several features of the Yorkshire grading system that I like. Nonetheless I’m not sure that trying to fix the ECF system by making it more complicated is the best way forward.

It does seem to address the issue of getting games on the system more quickly, which is the foundation of more regular grading lists.

It has a sensible approach to new entrants and the depreciation idea adds a nifty extra sophistication (although I reserve my views on whether the enhancement is worth the complication, and on the suitability of the weight).

Roger talked me last weekend about the difference between a grade – the figure used for administration of titles restricted tournaments etc – and performance – the player’s latest historical record of progress. (I hope I’ve got that right, RdC.) The ECF is about the former whereas I suspect the YCA goes too far in the other direction.

I have some concerns about the scaleability of the admin, if the number of sources of data were to rise, particularly in a transition.

I do have concerns about what appears to be a systemic deflation and stretch in the ECF system and I’m going to have to think a lot harder to form an opinion on whether the YCA adjustments would tend to fix or enlarge these issues.

It is useful that a working alternative model is around to shed light on the modifications required to get out more regular national grading lists. However in total they are a major change and it might be better to jump to a modified Elo, which might use some of the YCA ideas on new entrants and on improving juniors.

Alan Walton
Posts: 1393
Joined: Mon Nov 02, 2009 8:33 pm
Location: Oldham

Re: Future Grading Systems

Post by Alan Walton » Fri Dec 03, 2010 9:11 am

Jon, you have to be congratulated in your efforts in the Chessnuts grading system.

There will always be pros and cons with any system, but it does seem that it is a more transparent and comprehensive system compared with the very simplistic ECF average calculation (which has been derided over the last couple of years with the recent recalculation)

Maybe it is worth expanding by contacting other leagues like the Manchester League to see if they would start inputting into your system, there is already a high volume of players who play in Yorkshire events as well as this league

DavidCordover
Posts: 1
Joined: Mon Mar 28, 2011 2:10 am

Re: Future Grading Systems

Post by DavidCordover » Wed Mar 30, 2011 11:52 pm

The next step would be to have the pairing program integrate with the gradings so:
a) Arbiter puts results in
b) Pairing software calculates gradings
c) Automatically updates website

Then nobody needs to do any work! :D

Sean Hewitt

Re: Future Grading Systems

Post by Sean Hewitt » Thu Mar 31, 2011 9:43 am

DavidCordover wrote:The next step would be to have the pairing program integrate with the gradings so:
a) Arbiter puts results in
b) Pairing software calculates gradings
c) Automatically updates website

Then nobody needs to do any work! :D
FIDE does this already!

Jon Griffith
Posts: 10
Joined: Sat Aug 11, 2007 11:01 am
Contact:

Re: Future Grading Systems

Post by Jon Griffith » Fri Apr 01, 2011 11:47 am

DavidCordover wrote:
The next step would be to have the pairing program integrate with the gradings so:
a) Arbiter puts results in
b) Pairing software calculates gradings
c) Automatically updates website

Then nobody needs to do any work! :D
Chess-Results http://chess-results.com does all this already.
It is used in more than 100 countries, but never in Britain!
Jon Griffith
Yorkshire Grader

Sean Hewitt

Re: Future Grading Systems

Post by Sean Hewitt » Fri Apr 01, 2011 11:53 am

Jon Griffith wrote:
DavidCordover wrote:
The next step would be to have the pairing program integrate with the gradings so:
a) Arbiter puts results in
b) Pairing software calculates gradings
c) Automatically updates website

Then nobody needs to do any work! :D
Chess-Results http://chess-results.com does all this already.
It is used in more than 100 countries, but never in Britain!
The pairings it produces are wrong. They are being used here in Aix-les-Bains.

David Sedgwick
Posts: 5249
Joined: Mon Apr 09, 2007 5:56 pm
Location: Croydon
Contact:

Re: Future Grading Systems

Post by David Sedgwick » Fri Apr 01, 2011 11:58 am

on Griffith wrote:Chess-Results http://chess-results.com does all this already.
It is used in more than 100 countries, but never in Britain!
It was used at this year's Tradewise Gibraltar Festrival, an event under ECF auspices.

Sean Hewitt wrote:The pairings it produces are wrong. They are being used here in Aix-les-Bains.
Are the wrong pairings being corrected, or are they being allowed to stand?

Sean Hewitt

Re: Future Grading Systems

Post by Sean Hewitt » Fri Apr 01, 2011 12:04 pm

David Sedgwick wrote:
Sean Hewitt wrote:The pairings it produces are wrong. They are being used here in Aix-les-Bains.
Are the wrong pairings being corrected, or are they being allowed to stand?
They are allowed to stand as the tournament regulations say that the pairings to be used are those produced by SwissManager.

A good example is the second round pairings. Take a look at the half point score group which should be simple to pair correctly.

David Sedgwick
Posts: 5249
Joined: Mon Apr 09, 2007 5:56 pm
Location: Croydon
Contact:

Re: Future Grading Systems

Post by David Sedgwick » Fri Apr 01, 2011 12:22 pm

Sean Hewitt wrote:
David Sedgwick wrote:
Sean Hewitt wrote:The pairings it produces are wrong. They are being used here in Aix-les-Bains.
Are the wrong pairings being corrected, or are they being allowed to stand?
They are allowed to stand as the tournament regulations say that the pairings to be used are those produced by SwissManager.

A good example is the second round pairings. Take a look at the half point score group which should be simple to pair correctly.
Yes, three of those pairings are indeed bizarre.

To the best of my knowledge we didn't notice anything similar in Gibraltar. Possibly that was because we used accelerated pairings for the Masters.

I presume that the problem has been, or will be, reported to FIDE.

Edit: "two" changed to "three" following Roger's post below.
Last edited by David Sedgwick on Fri Apr 01, 2011 1:04 pm, edited 1 time in total.

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

Re: Future Grading Systems

Post by Roger de Coverly » Fri Apr 01, 2011 12:28 pm

Sean Hewitt wrote: A good example is the second round pairings. Take a look at the half point score group which should be simple to pair correctly.
The last three boards look really odd

106 94 GM Buhmann Rainer 2590 ½ ½ Lappas Konstantinos 1949 341
107 144 IM Muzychuk Anna 2528 ½ ½ GM Yakovich Yuri 2546 127
108 324 Webb Matthew D 2062 ½ ½ Denisov Ivan 2300 249

Alan Walton
Posts: 1393
Joined: Mon Nov 02, 2009 8:33 pm
Location: Oldham

Re: Future Grading Systems

Post by Alan Walton » Fri Apr 01, 2011 1:27 pm

I think you were looking too narrow, if you look at the last five boards then all of them had played each other in round 1, presumably the computer program did the best possible pairing and unfortunately you would get at least one strange pairing if you also consider colours

Post Reply