Grading games where the scoreline does not total 1

General discussions about ratings.
E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Grading games where the scoreline does not total 1

Post by E Michael White » Tue Oct 13, 2015 12:27 am

These will normally arise under FIDE Law 12.9 or FIDE Competition Rule 10.1 or equivalent in Event or National Rules.
On another thread Brian Valentine wrote:I don't think anyone would want another layer of arbitration on top of those that have already been seen in this case!
Graders grade games submitted by organisers (with some responsibility to check for administrative error). Organisers should record results as agreed by the players or as ruled by the arbiter if they don't. Players should have no recourse to change the result for grading purposes.
Hello Brian,

In these cases I think you need to take a different view and not initially use scorelines set by arbiters for grading. Since the new basis for new starters and juniors (about 2009) the alteration of one result will affect the grade of all juniors and adult new starters and potentially thousands of other players. You may remember an alteration to junior results about 2 years ago when -
Sean Hewitt wrote:
Daniel Young wrote:A revision of the January 2013 list has gone up today (4th Feb) "following the discovery of numerous incorrect grades." And indeed, a glance at the amended grades section reveals no less than 2,099 (!) grades have changed. A very large proportion of these are junior players, which leads me to wonder if there was a slip-up with the junior calculation process somewhere?
Two things have happened. Firstly, Cumbria have submitted their league results for grading and these have now been included resulting in changed grades for most Cumbrians. But far and away the biggest impact was the discovery that one junior player had a rogue grade allocated to her in the grading system. This affected her grade, that of every player she played, and every player who played someone she had played, and so on and so on. Ultimately, 2000 grades were affected; most, I'm told, by 1 point.
The effects are more than Sean suggested; when 1 junior grade is altered it is quite likely that every junior grade and every grade of any player, who played any junior, will be altered. In addition grades of many adult new starters would also change and thus grades of adults and juniors who played those adults etc. In many cases the revisions would be lost in rounding, being too small to change the published grade. All this of course assuming the system still works the way I think it does.
Roger Lancaster wrote:It's not relevant to the main issue but the game at the heart of this dispute was graded as a win for one player and a draw for the other. On discovering this, I queried it with Richard Haddrell who told me it was an occasional occurrence - you could count the number of occasions this (ie. a total award of 1.5 points) had happened over the past 10 years, he suggested, on the digits of one hand.
Richard's anecdotal notes from earlier than 2009 would not contribute positively to calculations performed on the new junior basis. If the 1.5 scoreline was used in 2014 then many pairs of juniors could potentially have had grades shown in the reverse order.

Scorelines of ½-1 etc should be viewed as a basic result which totals 1 plus an arbiter bonus/penalty to ensure the event positions/prizes reflect what the arbiter considered fair. I guess what needed to happen in the case of the 2014 scoreline of ½-1 would be to use the nearest compatible basic result, perform the grading for all players and then adjust the 2 grades for the arbiter bonus or penalty without altering the remaining grades. This is what I was getting at in the Aber 2014 thread. Using a basic 0-0 has some merit as neither competitor wished to continue and the position could not be agreed and ½-½ has some merit as the competitors could have agreed a draw had they known the alternative would be 0-0.

User avatar
David Shepherd
Posts: 912
Joined: Fri Nov 23, 2007 3:46 pm

Re: Grading games where the scoreline does not total 1

Post by David Shepherd » Tue Oct 13, 2015 9:43 am

My view is that games should only be graded where the outcome can be established. With the game in question it seems that effectively the players agreed a draw by not wishing to continue, but what could not be established was whether prior to that point the game had already ended by an illegal move. The score given of 1-0.5 seems to imply that there was a reasonable/some chance an illegal move was made, but to me seems incorrect, as a decision was needed one way or another - based on whether there was enough evidence available to confirm the illegal move (particularly given it was a national championship).

However no clear result was established and a decision was been made to record it as 1-0.5 which is a mixture of the two results. In these circumstance I believe it would be wrong to grade the game at all. This is on the ground that no result could be established and the score used was just for the purposes of recording the game in the tournament in a way as to be as fair as possible to the two players (although clearly unfair in relation to all the other players in the tournament - but for most of lesser importance).
Last edited by David Shepherd on Tue Oct 13, 2015 10:18 am, edited 1 time in total.

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

Re: Grading games where the scoreline does not total 1

Post by Roger de Coverly » Tue Oct 13, 2015 10:15 am

David Shepherd wrote: but what could not be established was whether prior to that point the game had already ended by an illegal move.
That was where it started to go wrong. The fine print of the new FIDE Laws contains a specific process for claiming a win by illegal move. If that hadn't been followed, the arbiter would have been justified in rejecting the claim of a win and ruling that the game should continue. Much the same if someone had claimed a draw by repetition of position outside the prescribed manner. There isn't a dispute that something happened, but it doesn't terminate the game.

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

Re: Grading games where the scoreline does not total 1

Post by Brian Valentine » Tue Oct 13, 2015 5:48 pm

Hi Michael,
I agree with your views about such results as ½-1 that they do contaminate the grading system. The Graders do discourage awarding such results.

However as the January 2013 issue shows we are mere mortals and therefore have no higher authority than those who are present when these decisions are made.

The grading system is based on results. Often underneath such results events distort the performance of the players in a game for instance: not understanding the time limit, letting a mobile phone go off, or occasionally an arbiter’s decision. I see nothing to be gained from revisiting results correctly submitted.
Brian Valentine
Manager of ECF grading

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

Re: Grading games where the scoreline does not total 1

Post by NickFaulks » Tue Oct 13, 2015 7:07 pm

In my view this issue is very simple, to the point of not being an issue.

The Elo rating system is mathematical in nature and is based on a few simple tenets, one of which is that the scores in a game add up to one. If you remove that, you haven't got anything that can properly be described as a rating system.

I don't see why this should not also apply to ECF grades,
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Grading games where the scoreline does not total 1

Post by E Michael White » Wed Oct 14, 2015 11:15 am

Brian Valentine wrote:..... The grading system is based on results. Often underneath such results events distort the performance of the players in a game for instance: not understanding the time limit, letting a mobile phone go off, or occasionally an arbiter’s decision. I see nothing to be gained from revisiting results correctly submitted....
I was suggesting different processing for unusual scorelines rather than changing the result and the gain would be more accurate grades for thousands of others. I raised the 2013 misgrading not to illustrate that graders are mortal or embarrass them but to remind of the consequences of a single wayward result affecting thousands of grades. This is a consequence of using the forward phased Markov steady state solution for junior grades and adult new starters which links the grades of all those players. While the new process instils some consistency it has the drawbacks of spreading junior grades too much and highlighting any mistakes in an unforgiving way.

Mathematical logic dictates to me that by processing these results in two parts some of the problems are alleviated. Taking the Aberystwyth ½-1 result as an example I think you could process it as ½-½ initially, perform the iterative process, calculate all the adult and junior grades and then add back 50/n to the player awarded the bonus where n is the number of games played, although there is a case for not adding anything back. The resultant grade would be discarded at the next grading date except for those approaching 18. An alternative would be to scale the result down to 0.33 : 0.67. I don't see a problem with the scoreline used to determine places and prizes being different from that used by the grader to calculate more accurate grades.

The reasoning behind split processing is that one way of looking at the iterative process is to view the same results as processed 30 or so times until the steady state is arrived at. It is unrealistic to assume that the same rare event of a mobile penalty or arbiter addition or arbiter mistake would occur in the following 29 games. Players would avoid that arbiter or deal with mobiles properly.

Grades are used for junior selection and board orders in matches and any improvement in accuracy would be reassuring.

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

Re: Grading games where the scoreline does not total 1

Post by NickFaulks » Wed Oct 14, 2015 12:11 pm

E Michael White wrote: An alternative would be to scale the result down to 0.33 : 0.67.
I have previously considered that in another context. There is no theoretical objection that I can see, but once you start on this path there is no knowing where it will end. Better avoided.
I don't see a problem with the scoreline used to determine places and prizes being different from that used by the grader to calculate more accurate grades.
Absolutely right, and in fact this should be obvious. The reason that many arbiters take the opposite view is that it allows them surreptitiously to add the docking of FIDE rating points to the already luscious list of penalties available to them under 13.4 when they wish to punish players.

Sorry, I forgot that since I disagree with John Foley on certain issues I must be a member of the controller-arbiter nexus. Will this get me expelled?
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

User avatar
Christopher Kreuzer
Posts: 8806
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Grading games where the scoreline does not total 1

Post by Christopher Kreuzer » Wed Oct 14, 2015 12:39 pm

It happens so rarely, just ignore the results and don't grade them. It may be an incentive for people to ensure sensible results.

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Grading games where the scoreline does not total 1

Post by E Michael White » Wed Oct 14, 2015 2:07 pm

Christopher Kreuzer wrote:It happens so rarely, just ignore the results and don't grade them. It may be an incentive for people to ensure sensible results.
That's ok between already graded adults. Where the game includes two top placed juniors in an event, the ECF grading model can adversely affect the grades of others in the event if the game is not graded.
Last edited by E Michael White on Sun Oct 18, 2015 11:36 am, edited 1 time in total.

User avatar
Michael Farthing
Posts: 2069
Joined: Fri Apr 04, 2014 1:28 pm
Location: Morecambe, Europe

Re: Grading games where the scoreline does not total 1

Post by Michael Farthing » Fri Oct 16, 2015 2:33 pm

Michael, can you say by what amount other players' grades might be affected? If we are talking about barely graded players (say 10 games) then the grade is already very suspect and in any case unlikely to put these players in contention for 'high office'. If we are talking about players with 30 games then I would have guessed that the effect is likely to be within the recognised inaccuracy of the grade. Would this be fair?

Richard Bates
Posts: 3338
Joined: Fri Nov 14, 2008 8:27 pm

Re: Grading games where the scoreline does not total 1

Post by Richard Bates » Sun Oct 18, 2015 10:55 am

For all the talk about 0.5-1 "corrupting" rating/grading systems, it could be pointed out that the same is equally true of 0-0.5 or 0-0. The first is explicitly allowed for within the FIDE laws of chess, possibly even the second. We have an IMO ridiculous situation at the European Club Cup when draws agreements before move 40 are disallowed under penalty of a double default award. This of course is particularly nonsensical in a team event scored by match points...

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

Re: Grading games where the scoreline does not total 1

Post by Roger de Coverly » Sun Oct 18, 2015 11:01 am

Richard Bates wrote: We have an IMO ridiculous situation at the European Club Cup when draws agreements before move 40 are disallowed under penalty of a double default award.
Are you allowed to follow McShane-Nakamura?

Actually you could repeat five times and challenge a watching arbiter to apply the Laws of Chess.

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

Re: Grading games where the scoreline does not total 1

Post by NickFaulks » Sun Oct 18, 2015 11:18 am

Richard Bates wrote: it could be pointed out that the same is equally true of 0-0.5 or 0-0.
I can assure you that it has been, with considerable force. I have already explained why arbiters like rating 0-0. As for 0-0.5, that's just stupid.
If you want a picture of the future, imagine a QR code stamped on a human face — forever.