A possible improvement to grading?

[Geoff Chandler]

If you took away the points you gained, your grade would go up?

Please explain.

"Yes, I am one of those sad losers who keeps track of it."

A girlfriend. You need a girlfriend.
(and on the first date don't start talking about your grade).

Joking aside Seb.
Time would be better spent studying why you did not win the game
you drew.
Forget the numbers, they may be influencing how you play.

[Sebastian Stone]

If you took away the points you gained, your grade would go up?

Please explain.

Erm...Ok.

My grading performance for games I've either won or loss is grade A

My grading performance for games I've drawn is grade B

Grade A is bigger than Grade B.

"Yes, I am one of those sad losers who keeps track of it."
A girlfriend. You need a girlfriend.
(and on the first date don't start talking about your grade).

So THAT's what I've been doing wrong.

Joking aside Seb.
Time would be better spent studying why you did not win the game
you drew.
Forget the numbers, they may be influencing how you play.

There are many things that influence how I play.

If you want an essay on why I suck as a chess player and will never get any better, well to be honest I can't be bothered, but needless to say I have many many flaws.

[Geoff Chandler]

"....well to be honest I can't be bothered."

What a refreshing and honest answer.

You clearly enjoy your chess and play for the love of the game. Well done.

(not sure sure about grade A and grade B do you have two grades?)

Don't bother answering. I'm off out now to meet some mates to perhaps play some chess,
batter out Beatles and Bonzo Dog songs on my guitar, tell some bawdy jokes and get drunk.

[timlawson]

It's clearly opened a can of worms!

I try to make a point of not knowing my opponents grade prior to kick off... I try to play the position and if I know the player, I play the player. Obviously, this will get more difficult for me. My circumstances are that I started playing league chess again in an area where I didn't know my opponents and didn't know their ratings. Two seasons down the line (well, almost), I am now more familiar with some of my opponents as they are doubtless familiar with me.

Yes, it's possible to draw too many games and lose rating points... it's also possible to draw games and gain rating points. My personal view is that the ECF ratings are shot to pieces and part of this is due to the "adjustment". The "adjustment" in itself is an admittance that the existing doesn't work. It's obvious it's not a sustainable mathematical model but as Geoff points out, it's chess not rocket science.

Personally, I don't take any pride or shame in having the ECF rating I do - I'm there at the moment because that's where it puts me (including errors and a rather erroneous system which takes into accounts results from five years ago). Has this affected my own rating adversely? I dunno is the answer and to be honest, I'm not really that bothered. If I do well, the chances are my rating will improve and if I carry on playing the way I have since I took up competitive chess again, things can (and will!) only get worse for me. Eventually I will find my level and that will be my grade. Until the ECF undertake another "adjustment".

It's definitely an interesting subject though and perhaps someone with an understanding of maths could provide a better formula? I cannot possibly subscribe to the argument that the existing system is in any way accurate - it's a year out of date when it gets published. This alone is enough to debunk it as a worthwhile effort. I appreciate that teams need to determine board order etc but the leagues that I play in, you can chop and change to a certain extent anyway..... I'd rather people just got on with playing chess and didn't worry so much about a number after their name!

[Roger de Coverly]

I'm there at the moment because that's where it puts me (including errors and a rather erroneous system which takes into accounts results from five years ago).

You've been misinformed. Unlike Elo and probably Glicko based systems, the ECF system is quite ruthless in discarding old results. Everything from more than three years ago is ignored, and if you play enough games the calculation period shortens to one year.

You may have noticed that tournament and league controllers will be aware of your previous grade. This is only used for board orders, pairing orders and tournament eligibility. The calculation of your next grade has no knowledge of previous results if they are more than three seasons ago.

[timlawson]

I'm there at the moment because that's where it puts me (including errors and a rather erroneous system which takes into accounts results from five years ago).

You've been misinformed. Unlike Elo and probably Glicko based systems, the ECF system is quite ruthless in discarding old results. Everything from more than three years ago is ignored, and if you play enough games the calculation period shortens to one year.

You may have noticed that tournament and league controllers will be aware of your previous grade. This is only used for board orders, pairing orders and tournament eligibility. The calculation of your next grade has no knowledge of previous results if they are more than three seasons ago.

Hi Roger - I don't think I have been misinformed as some results from the last season I played (2005) WERE included with my rating that was issued in 2010. I checked this with Richard Haddrell. You can check here: http://grading.bcfservices.org.uk/getre ... ef=163276A and if you look at the data it states 21 games counted latest season, 25 games counted total. My last chess was played in 2005 which is a five year gap.

This gives me a D rating as well so it would appear that results older than three years ARE used. I'm not complaining though because as I say, it's all rather relative. Had the results from 2005 not been included, I'd have been (in my opinion) overrated. I'm probably overrated at 142, hence the frequent nosebleeds!!

[Roger de Coverly]

Hi Roger - I don't think I have been misinformed as some results from the last season I played (2005) WERE included with my rating that was issued in 2010. I checked this with Richard Haddrell. You can check here: http://grading.bcfservices.org.uk/getre ... ef=163276A and if you look at the data it states 21 games counted latest season, 25 games counted total. My last chess was played in 2005 which is a five year gap.

I spotted that as well, but I assumed you must have played 4 games which fell into the previous season. If true that the 2005 games were included, the system is not performing according to the generally understood specification.

So 25 games is 21 in 2009-10 and 4 in 2008-9. Cut off dates should have been May 31st 2009. Anything before that would have been 2008-9.

[timlawson]

Hi Roger - I don't think I have been misinformed as some results from the last season I played (2005) WERE included with my rating that was issued in 2010. I checked this with Richard Haddrell. You can check here: http://grading.bcfservices.org.uk/getre ... ef=163276A and if you look at the data it states 21 games counted latest season, 25 games counted total. My last chess was played in 2005 which is a five year gap.

I spotted that as well, but I assumed you must have played 4 games which fell into the previous season. If true that the 2005 games were included, the system is not performing according to the generally understood specification.

So 25 games is 21 in 2009-10 and 4 in 2008-9. Cut off dates should have been May 31st 2009. Anything before that would have been 2008-9.

I've checked the email Richard sent me - apparently it's four games from 2007-2008. I think they must have been carried forward from a previous year though. The odd thing is that I definitely didn't play any chess from 2005 onwards and it's been noted as Metropoliatan CC Championship (which sounds like club championship games rather than league games).

I was listed under Metropolitan after moving to Milton Keynes in 2004. I remember playing a handful of games towards the end of that season (I arrived in MK in March 2004) so perhaps they were held over and graded in 2005. Still doesn't make sense though as from what Richard has written, they were four games, average rating performance of 45. I remember losing a couple and getting a draw but mostly against higher rated players (I think the draw was against a 130 odd player) so perhaps it's just some games someone else has played that have been allocated against me. It doesn't make sense though, as that means I should have still got a grade in 2008 by my reckoning?!

It's a bit late to ask for things to be adjusted now and it's actually a good example of the "human error" element I've pointed out previously! For the record, it appears my rating would have been around 160 had those four results not been included - that would be an over estimation of my playing strength in my opinion. I had a lucky run when I started playing again! The same as this year's results which will be a bit skewed for me as I've had a few losses against players rated 40+ points above me so I think my rating will fall a short of what it is realistically (which I reckon is about 150 but not more). I think it's going to stay around the same or drop slightly. However, who knows?! I've not played 30+ games so some results from last season will be counted but again, I don't know for sure which results will be counted. I do know that at present it will be a B rating (instead of D) with 21 results from this season plus 9 from last season.

Looking back, I appeared to have one game missed off from last season but this was played on 27/4/2010 (after the grading cut off perhaps?) That would mean 22 to be counted from "this" season and then 8 from last season - but which 8 games from last season? Which of my opponents grades would be used to calculate these 8 results? This seasons or last seasons grades?

I can certainly see where a lot of the confusion arises! I therefore stand by my statement that although the rating system provides an indication of relative playing strength, it's easy to skew the figures and data is often out of date!

[Roger de Coverly]

It's a bit late to ask for things to be adjusted now and it's actually a good example of the "human error" element I've pointed out previously!

I don't think you should tolerate the error. If there's another T Lawson or similar, then your games could continue to be confused unless it's made completely clear that there are two of you. Considerable strides have been made in improving the reliability of the data capture for grading and training people when inputting results to get the grading code right.

That would mean 22 to be counted from "this" season and then 8 from last season - but which 8 games from last season? Which of my opponents grades would be used to calculate these 8 results? This seasons or last seasons grades?

Usually it's done by counting backwards from the 31st May. League games aren't always accurately dated. At any one time, an individual should only have one grade, so it's the grade in existence at the time the game was played.

A grade is an estimate of strength. For adult players it's usually quite reasonable to assert that their strength changes slowly over time or not at all. The notion of blending historic and recent results is present in some form in almost all rating systems.

No grading or rating system can cope with games being allocated to the wrong player. The ECF (or those working on its behalf) are working on enhancements which will put detail results on-line. This hopefully will help pin down results incorrectly reported or credited to the wrong people.

[timlawson]

It's a bit late to ask for things to be adjusted now and it's actually a good example of the "human error" element I've pointed out previously!

I don't think you should tolerate the error. If there's another T Lawson or similar, then your games could continue to be confused unless it's made completely clear that there are two of you. Considerable strides have been made in improving the reliability of the data capture for grading and training people when inputting results to get the grading code right.

That would mean 22 to be counted from "this" season and then 8 from last season - but which 8 games from last season? Which of my opponents grades would be used to calculate these 8 results? This seasons or last seasons grades?

Usually it's done by counting backwards from the 31st May. League games aren't always accurately dated. At any one time, an individual should only have one grade, so it's the grade in existence at the time the game was played.

A grade is an estimate of strength. For adult players it's usually quite reasonable to assert that their strength changes slowly over time or not at all. The notion of blending historic and recent results is present in some form in almost all rating systems.

No grading or rating system can cope with games being allocated to the wrong player. The ECF (or those working on its behalf) are working on enhancements which will put detail results on-line. This hopefully will help pin down results incorrectly reported or credited to the wrong people.

I raised this with Richard Haddrell, more out of morbid curiosity rather than a wish to have my rating amended (with two games left in my season!). I got a reply stating that I had been misinformed and he was wondering why I hadn't raised the issue over the 4 erroneous rated games when he sent his original reply. My response?! Human error on my part - I just didn't read much past the " 4 games played at average rating of 45". I didn't read that this was from the 2007-8 season (I didn't play any chess between May 2004 and the start of the 2009 season). Obviously there has been an input error at some stage and all I was doing was highlighting my prior theory that, when you introduce humans into the equation, errors get made! (And this is errors at both ends, the ECF/Grading Officers for messing things up and my good self for not scanning an email correctly!).

As Richard was kind enough to tell me, there is little point in investigating now and I have not asked him to. The simple fact is that had those games NOT been included, I would, in my opinion, have been overrated. As it stands, my grade is what it is. I don't enter loads of congresses and have not been played out of grading order for either team I play for due to league rules so no harm has been done. The "dodgy" results will drop off at the end of this season and I'll get a new grade which will be roughly the same as my existing rating (give or take a few points).

I'll be examining my rating very closely next time round and will notify Mr Haddrell of any potential discrepancies - unfortunately, my email decided it wouldn't send my reply to him for some reason (issues at my end most likely).

[Roger de Coverly]

I'll be examining my rating very closely next time round and will notify Mr Haddrell of any potential discrepancies

The validation approach in recent years is that the ECF will send out a list of games to be included in the grading calculations to everyone where it has a grading code, an email address and implicit permission to send. This gives a wider set of eyes who can notice elementary things like events missing, events counted twice or phantom results. The second validation is that a web-only list will be published at the end of July. This gives a second chance for players to check their game count, club designation etc. An additional validation was that a list of possible duplicates was prepared. You were invited to confirm or deny that these were separate people.

Relatively few data collection errors have been detected.