New Longplay twice yearly grading...good or bad

[Stewart Reuben]

Responding to Paolo
The grade is a fact. Let us assume we have a player who, in a season, meets 30 or more opponents graded 175 and draws with the lot, his grade will be 175. If he met 30 players graded 175 and beat half, while losing to the other half, his grade would still be 175. This is perfectly accurate, but being once a year, not very useful.

A rated player of 2200 meets 30 opponents in one two month period (that's a lot of chess) all rated 2000 and draws with all of them. Then his new rating can be calculated as follows:
From the table, his expected score was 0.64 for each game. Thus he lost 0.64-0.5= 0.14 points for each game. His k factor is 15. Thus he loses 0.14 x 15 x 30 = 63. His new rating is 2137. Had he won 15 and lost 15, the result would be identical.
The 2137 is as estimate based partly on the old information and partly on the new. By definition it cannot be accurate. But, being every 2 months, it is quite useful.
Had he played 50 games (almost impossible), his new rating would be 2095 and it would be no matter what his starting rating was, provided the k factor was 15.
If the k factor is 10 and he is rated 2400 and he was playing against 30 2200 players drawing 30, then his new rating would be 2158.
If the k factor is 30 (as he is relatively new to the list) and he is 2200 playing against 2000. Then the new rating would be 2074.
If he played 60 (that's 2 a day) games, then his new rating would be 1948. He would have 'overshot' the mark. It is an anomaly in the FIDE System and never been corrected.
k factor 10 turns over in 75 games, 15 if 50 and 30 in 25. It is 750/k = the number of games.

Were the ECF to go to monthly lists, or after each result, they would have to keep a rolling grade or go to the FIDE System.

In response to Roger, Elo (and probably Clarke) made it clear that, for a rating to have statistical validity the player must meet a mix of opponents above and below him in grade. Kasparov could never have an accurate rating. If a player always meets the same opponents, his grade will have less statistical validity relative to the whole list. There is some evidence the players rated over 2700 have somewhat inflated ratings because of this problem.

Stewart Reuben

Gleko takes into account a player's variance.

[Roger de Coverly]

Were the ECF to go to monthly lists, or after each result, they would have to keep a rolling grade or go to the FIDE System.

There is a third way of doing monthly or more frequent grades in an ECF system. You calculate the performance so far in the season and calculate the resulting grade assuming no further games played. If you have more than 30 games, you only need the current season. For less than 30 you have to either roll back or average to make up 30. New and ungraded players are a problem because you need the establish their estimated grade (or leave then out completely).

The ECF would stick with once a year, or twice a year updates of the grades used in the calculations.

It's a traditional method in which you use the column of opponent's grades in your scorebook as the data.

[Brian Valentine]

I think Stewart needs to check his maths though his discussion on grading changes is essentially sound except he uses very extremes cases - a curse of this particular issue.

If a player always meets the same opponents, his grade will have less statistical validity relative to the whole list.

One point is that the winning methods in the recent Kaggle competition to predict chess results actually used the idea that results from a common swiss event were more useful in reducing the prediction error, contrary to the view, shared by Stewart, that I would have expected.

Roger, I think there are a lot more than 3 methods! There is the Yorkshire approach (weighting by a discount factor) and Kalman filters to add for starters.

I'm all in favour of more frequent gradings, but the solution very much depends on the objectives (repeating my contribution on the first page of this thread). From what I have read since, some leagues value "real time" gradings and Stewart asserts such lists would increase games played. Everyone has a different take on the differences between Elo and Clark; from my point of view Elo can work "real time", whereas Clark's approach would fall down. However Elo is very poor at evaluating fast improving players.

I'm sure we will move to much more frequent lists in due course. The six monthly lists will highlight administation issues, but they are a step away from the right path for methodology.

[David Pardoe]

In response to Daniel Young.

David Pardoe still doesn't seem to realise, there is no extra work if the grade at the BEGINNING of the season is used throughout the season for team order and qualification. If the league decides to use a different system, then no doubt it will take into account the extra work involved.

Stewart Reuben

Stewart,
Have you thought about this from the player or club, league, officials perspective, given the various options that might be available.
Here`s an `email discussion` I`ve had recently, but you might like to try it with various permutations of player grading changes for league & county players, where grades fluctuate (including ungraded players)....say for a club with say 4 teams in 2 different leagues, each adopting different policies...
We had one ungraded player last year given E143, who then went on to perform to 185.

Here`s the discussion...

David, and all....
In responce to your email (below)...re Twice Yearly Standardplay grading....
If you look at the discussions on the ECForum http://www.ecforum.org.uk/viewtopic.php?f=25&t=3073 , you`ll see a variety of comments on this...over 100, I think. Some quite interesting points made.....
No, I`m not trying to speak for anybody but myself......others can feel free to express there own views.
I can see the attraction of your arguement, but in practice this could involve whole teams changing round in mid season, with clubs having to review there player lists, and team captains suddenly finding they`ve got a changed team (having spent time trying to get a settled team), which might also mean changed transport arrangements, if they follow your preferance. In most cases I suspect that player grades might actually fluctuate within say 15 grading points, so maybe two players per team might change team (say going from 2nd to 3rd team), and maybe 2 players might just switch board order, in a team of 7 players. This level of minor grading creep at mid season is probably pretty normal and can be ignored for league match purposes I suspect.
I just wonder if all that shuffling around will really achieve much...and for clubs with several teams in various leagues, it could just become a hassle, particularly if different leagues adopt different policies..
If this applied in county matches the situation could be more tricky, with `out of form` players suddenly finding they were `out of the team`...with nowhere to go. Alternatively, players could find themselves occupying strange looking board orders, if the counties teams just decide to ignore the new grades as suggested.
As an illustration, a county player graded say 160, playing in the U180 team on board 14 suddenly drops to grade 135 at mid season. He retains his low board, then meets another 160 player who has gone up to say, 185 at mid term. You suddenly have a curious looking match where some players look commically outgraded, others look as if the board order has been randomised, making the competition look almost farcical/confusing. One well regarded ECF Arbitor & grader has said his thoughts on this are unrepeatable.
But for specific cases, like the one we had last year in the Stockport league, a slight mid term adjustment might have made good sense. But what we dont want is an accelerated `yo-yo` situation, where players are `up` one minute and `down` the next. They might just get fed up playing musical chairs.
Might this not be better dealt with by simply allowing club secretaries, league secretaries, and team captains to exercise and make mid season adjustments where they think fit (within the league rules)...with league secretaries being empowered to issue directives where necessary, rather than altering thousands of player grades.
As another point, this proposal will not just speed up grades, it will alter the actual players grade calculations over time. League stats software may also need changing, along with league rules on eligability.
The one area where these gradings are more important is for Congresses, where `dark horses` might need to be flagged up more promptly, but maybe this could be done directly between Congress organisors and the ECF grader team, applying changes directly to grades in certain cases..
I dont have any consistant `form`, so it doesn`t matter to me what grades apply..ie, I can play badly in one match, then quite well in another.
Curiously, I`ve played better in the 4NCL Northern league, less well in league matches, and dreadfully in county matches....with no Congress form this season.
Incidentally, the ECF are actually anticipating that leagues, and county teams will not use the mid term grades, but that would be down to the individual leagues to decide.
One question I have is `do we really need to regrade all our players in mid season in order to deal with a problem that probably only effects a handful of players in each league`.?
I raised these and other issues because it looked to me that the ECF were pushing this through without much real consultation with various chess bodies.
Whether they take any notice or not remains to be seen...and whether ECF delegates will be prompted to ask more detailed questions...we`ll see.
I certainly think that people should discuss these issues at forthcoming AGMs.

This was in responce to the following....

----- Original Message -----
From:
Sent: Wednesday, June 08, 2011 3:49 PM
Subject: RE: gradings


Hi Peter
I think it’s great we are having 2 snapshots each year !!
The new grades will be used in our Summer league !!
Which starts each June – Sept !
Plus at the turn of each January we can regroup our winter league team registrations to reflect who is doing well, or not as the case may be !!


PS It could be that people think this will add some `spark` to things, and I can empathise with such views......

[Roger de Coverly]

Plus at the turn of each January we can regroup our winter league team registrations to reflect who is doing well, or not as the case may be !!

It's the likely choice of most leagues that they won't do this. Board orders within teams may change but not teams within clubs. Personally I would advocate the ECF publishing up to date performance information on a monthly or continuous basis. You would always have the begin year or half year grade used in calculations but you and the world would see all the processed results for the season. As Stewart Reuben suggested, the implementation of this would probably require match captains to report results direct to a central system in a manner whereby the central system received the results without doing anything. One way to make this work would be that the online system had to be loaded with every league's fixture list and club player list. This makes the input just a few mouse clicks.

[Alex Holowczak]

Personally I would advocate the ECF publishing up to date performance information on a monthly or continuous basis.

You can't, because that'd be a pain with juniors, who are treated as ungraded until the end of the season. However, I believe it is Richard's attention to update the website every month or so, which may not have any grading calculations on it, but it will have the results of any games you've played in congresses - or anything else that's submitted its results - on it. At least, that's the impression I got from the thing he wrote on the SCCU website.

[Adam Raoof]

I would like to have my grade updated during each game, preferably after each move. Wouldn't it be great?

[Roger de Coverly]

You can't, because that'd be a pain with juniors, who are treated as ungraded until the end of the season.

It's chicken and egg. If you have an objective of publishing more frequent grades, it doesn't make sense to increase the number of ungraded players you have to deal with. It was in only in 2009, less than two years ago, that the grading team introduced the juniors rule. That juniors rule has all the signs of a last minute panic, a move played without proper analysis with the flag about to fall.

New players could be included on a provisional estimate basis or left out entirely so there are feasible solutions to the problem of new players. The grades calculated are indicative only, so there the consequences of them being slightly wrong have no long term effects.

FIDE are going monthly, and effectively continuous by tournament, where results and rating changes are published on the next day after a tournament.

[Stewart Reuben]

Alex >You can't, because that'd be a pain with juniors, who are treated as ungraded until the end of the season.<

I don't think anybody is suggesting that changes would be made without there being consequent changes. I think this particular matter is solved by changing season to grading period.

David Pardoe >Stewart,
Have you thought about this from the player or club, league, officials perspective, given the various options that might be available.<

Yes thank you. Over the past 45 years since starting to advocate such changes.

Adam >I would like to have my grade updated during each game, preferably after each move. Wouldn't it be great?<

An excellent idea, but unfortunately against the current Laws. There would be little point unless you could see your change in grade during play. Clearly that would be advice. Also computers are not to be trusted. By chance I discovered last night that Fritz 10 recognises rook's pawn + bishop of the wrong colour as drawn. BUT assesses rook's pawn on the seventh + 4 bishops of the wrong colour as +13, until over 40 moves without progress. But don't panic. FIDE VP Ignatius Leong is proposing we change the Laws to create what I would call 'So long Chess'.

Roger >FIDE are going monthly, and effectively continuous by tournament, where results and rating changes are published on the next day after a tournament.<
That isn't expressed quite correctly and may confuse some. Although the information regarding ratings are available immediately after an event is submitted and the work has been done, the rating remains the same for the rating period. Thus a player is 2200 on 1 July. On 15 July he goes to 2250. Until 1 September his rating remains at 2200 for rating and title purposes. From 1 July 2012 change that last sentence to 1 August.

Stewart Reuben

[Alex Holowczak]

FIDE VP Ignatius Leong is proposing we change the Laws to create what I would call 'So long Chess'.

For example?

[Stewart Reuben]

Ignatius is proposing:
To make it illegal to repeat positions.
Not to allow more than 5 checks in a row with the same piece.
This has been reported elsewhere on this forum. Ignatius has gone quite mad, his balance of mind has become disturbed by the number of draws in the Candidates.
Otherwise surely he would have suggested stalemate be a loss for the player either stalemated or making the stalemate. That certainly would reduce the number of draws.
The ban on repetitious exists by the way in Shogi.

Stewart Reuben

[Simon Ansell]

Otherwise surely he would have suggested stalemate be a loss for the player either stalemated or making the stalemate.

I always thought this was a reasonable idea if it people really think it necessary to reduce the number of draws in chess, but you never see it mentioned when these discussions arise - I think this is the first time I've heard it from someone other than myself. The aim in chess is to take the opponents king - if you stalemate your opponent you can take the king next move and therefore you should win.

Sorry to go off-topic (the more grading lists the better imo)!

[Roger de Coverly]

Ignatius has gone quite mad, his balance of mind has become disturbed by the number of draws in the Candidates.

The worrying aspect is that he is on the new Commission for Modernisation of FIDE and that he had a considerable influence over the introduction of instant defaults for late arrivals from the Dresden Olympiad onwards.

[Roger de Coverly]

I always thought this was a reasonable idea if it people really think it necessary to reduce the number of draws in chess!

It does, of course, radically change endgame theory, which in turn may radically change opening theory and the balance between white and black. Many lines where Black gives up a pawn for initiative rely on being able to reach a drawn ending. If fewer endings are drawn, then these lines lose viability.

[Stewart Reuben]

There is nothing wrong with suggesting variants on chess. Recently I introduced a variant on a variant. Circular Exchange chess. But each one is a different game. Tht was why I was against the Laws of Chess960 appearing in the Laws of Chess.
Wrenching us back to grading. Since it is a different game, it needs a separate possibly parallel grading list. I have long thought 3/1/0 should not be rated in the same list as 2/1/0.
Throwing away virtually all endgame theory, with repetition and/or stalemate rules would surely be vandalism?

Stewart Reuben