2010 Grading list

General discussions about ratings.
Sean Hewitt

Re: 2010 Grading list

Post by Sean Hewitt » Fri Aug 27, 2010 10:00 pm

Paul Bielby wrote: But one feature of this tournament interests me with regard to ratings. As well as all the normal 1st, 2nd, 3rd prizes, etc. they offer quite substantial prizes for PERFORMANCE - RATING (i.e a players rating performance in the tournament minus their initial rating) This is an idea I have not seen before. Does anybody know of this happening in any tournaments in this country?
Errrm. There's a certain series of FIDE rated events that awards all of it's rating prizes based on performance relative to rating (W-We).

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

Re: 2010 Grading list

Post by Richard Bates » Fri Aug 27, 2010 10:00 pm

Paul Bielby wrote:I have been following the Heraklion Open Tournament in Crete this week. (Two NE players, Paul Townsend (York) and Tim Adams (Jesmond) are taking part, wiith Paul sharing the lead on 5/6 at the moment).

But one feature of this tournament interests me with regard to ratings. As well as all the normal 1st, 2nd, 3rd prizes, etc. they offer quite substantial prizes for PERFORMANCE - RATING (i.e a players rating performance in the tournament minus their initial rating) This is an idea I have not seen before. Does anybody know of this happening in any tournaments in this country? It would work perfectly well with gradings, I think. In the Cretan tournament it seems to take the place of our hierarchy of Major, Intermediate and Minor Tournaments - everybody plays in one big competition. This could well have some advantages.

E.g. Player A of grade 151 might have to play in an Major and player B of 149 be able to play in the Intermediate in one of our tournaments - A would have a much harder task and yet B would be far more likely to win a prize. The Cretan system does away with such artificial boundaries. What do forum members think of this?
It works much better in long tournaments. I think it might work less fairly for UK weekenders, because there wouldn't be sufficient rounds for the result to be determined by genuine performance rather than luck of the draw (eg.one player performing better than the other simply be being paired against (and losing to) a stronger player).

Paul Bielby
Posts: 154
Joined: Tue Dec 22, 2009 4:14 pm
Location: South Shields

Re: 2010 Grading list

Post by Paul Bielby » Fri Aug 27, 2010 10:07 pm

No, Alex, I'd not be surprised - I know the grumbles we had from a 'Minor' player when the new gradings came out and he found he might have to play in a Major!

The Cretan tournament was over 9 rounds and the first round was very much a 'slaughter of the innocents'. An accelerated Swiss system might help to avoid this - or even hyper-accelerated. It doesn't seem to put off the Greeks - I suppose its very much a case of people wanting the system they are used to.

Alex Holowczak
Posts: 9085
Joined: Sat May 30, 2009 5:18 pm
Location: Oldbury, Worcestershire

Re: 2010 Grading list

Post by Alex Holowczak » Fri Aug 27, 2010 10:10 pm

Paul Bielby wrote:No, Alex, I'd not be surprised - I know the grumbles we had from a 'Minor' player when the new gradings came out and he found he might have to play in a Major!

The Cretan tournament was over 9 rounds and the first round was very much a 'slaughter of the innocents'. An accelerated Swiss system might help to avoid this - or even hyper-accelerated. It doesn't seem to put off the Greeks - I suppose its very much a case of people wanting the system they are used to.
I think they (continental Europeans) have a different mentality to us when it comes to most things, including chess. I think England one of the few countries that runs tournaments this way; and we have to for it to be financially viable, because some of the weaker players wouldn't play because they can't win anything.

I wish you wouldn't call it the Cretan tournament; I keep reading it as "cretin" tournament... :?

Sean Hewitt

Re: 2010 Grading list

Post by Sean Hewitt » Fri Aug 27, 2010 10:15 pm

Alex Holowczak wrote:I think England one of the few countries that runs tournaments this way; and we have to for it to be financially viable, because some of the weaker players wouldn't play because they can't win anything.
For us, it's the opposite. We would be better off financially with two sections (the open and the major) only.

Alex Holowczak
Posts: 9085
Joined: Sat May 30, 2009 5:18 pm
Location: Oldbury, Worcestershire

Re: 2010 Grading list

Post by Alex Holowczak » Fri Aug 27, 2010 10:23 pm

Sean Hewitt wrote:
Alex Holowczak wrote:I think England one of the few countries that runs tournaments this way; and we have to for it to be financially viable, because some of the weaker players wouldn't play because they can't win anything.
For us, it's the opposite. We would be better off financially with two sections (the open and the major) only.
Then why do you run the Minor? :wink:

Sean Hewitt

Re: 2010 Grading list

Post by Sean Hewitt » Sat Aug 28, 2010 10:34 am

Alex Holowczak wrote:
Sean Hewitt wrote:
Alex Holowczak wrote:I think England one of the few countries that runs tournaments this way; and we have to for it to be financially viable, because some of the weaker players wouldn't play because they can't win anything.
For us, it's the opposite. We would be better off financially with two sections (the open and the major) only.
Then why do you run the Minor? :wink:
Because we get 20 odd players who want to play in it and I'd rather give them the opportunity to play amongst their peers than tell them they have to play in an U2000 or bugger off.

Alex Holowczak
Posts: 9085
Joined: Sat May 30, 2009 5:18 pm
Location: Oldbury, Worcestershire

Re: 2010 Grading list

Post by Alex Holowczak » Sat Aug 28, 2010 11:08 am

Sean Hewitt wrote:
Alex Holowczak wrote:
Sean Hewitt wrote: For us, it's the opposite. We would be better off financially with two sections (the open and the major) only.
Then why do you run the Minor? :wink:
Because we get 20 odd players who want to play in it and I'd rather give them the opportunity to play amongst their peers than tell them they have to play in an U2000 or bugger off.
Fair enough! :)

User avatar
Adam Raoof
Posts: 2720
Joined: Sat Oct 04, 2008 4:16 pm
Location: NW4 4UY

Re: 2010 Grading list

Post by Adam Raoof » Mon Aug 30, 2010 7:49 pm

Given the number of players (parents / children / partners / clubmates) who travel together to weekend tournaments, it is always a good thing to offer a range of events to make sure everyone's occupied.

How many more entries could you accommodate comfortably, Sean? The Minor would normally be the largest section.
Adam Raoof IA, IO
Chess England Events - https://chessengland.com/
The Chess Circuit - https://chesscircuit.substack.com/
Don’t stop playing chess!

Sean Hewitt

Re: 2010 Grading list

Post by Sean Hewitt » Fri Sep 03, 2010 12:14 pm

Adam Raoof wrote:Given the number of players (parents / children / partners / clubmates) who travel together to weekend tournaments, it is always a good thing to offer a range of events to make sure everyone's occupied.

How many more entries could you accommodate comfortably, Sean? The Minor would normally be the largest section.
We could accommodate 20 more players comfortably. Beyond that we would just hire more playing space, which would not be an issue.

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

Re: 2010 Grading list

Post by Brian Valentine » Wed Nov 24, 2010 9:25 am

In an earlier note on page 5 above, I pointed out how the standard deviation of grades had been steadily increasing since the re-calibration in 2009. I have been looking at this, to me, unexpected result further.

The re-calibration was justified because of a stretching in grades. This had lead to stronger players under-performing against weaker players and vice versa. We can see the emergence of a new stretching, but data is not available to see if the relative under-performance is also emerging.

The variance of all grades increased from 1,372.47 in 2009 to 1,403.20 in 2010. An increase of 30.73.

The reason this situation is unexpected is that there are reasons for thinking the variance might fall. The main driver would be that newer weaker players would be expected to improve quite quickly, whereas the effects of ageing would be expected to be slower. This effect would push grades closer together. However this effect is overwhelmed by the effects of new entrants and leavers. The new entrants (particularly juniors) are relatively weak and the leavers are pretty average overall. This difference in averages of different subgroups is quite important when looking at the overall variance.

The data needs to be split in fairly homogenous groups. I have chosen to split into 72 groups by category. There are 5 categories (A-E) for each year 2009 and 2010 and furthermore there are those uncategorised in 2009 (new entrants) and those uncategorised in 2010 (leavers). This gives 6 categories in each year combined into 36 groups (AA, AB etc). Then Juniors in 2010 are split from the others (“seniors”).

The contribution to change in variance can be allocated 3 sources. The change in the variance of within that group, the change arising from the change in average of that group (squared) and finally the change due to the change in the population .

The sum of (grades – global average)2 / total population =

The sum of (grades –group average)2 / group population +

The sum of (group average – global average)2

And when considering the difference over the year each of the above needs to be adjusted for the change in population (the denominator of the variance calculation).

I have called these sources internal variance, averages and drift. In this summary all those categorised in both years are combined as survivors.

The change in variance of 30.73 is allocated as follows:
.............Survivors.............. New Entrants..........Leavers
............ Senior..... Junior..... Senior..... Junior..... Senior......Junior
Int. Var.__ 22.08____ -2.57_____244.30___ 46.79____ -259.33___-27.88
Averages_ 4.52_____ -14.46______1.25___ 53.47____ -3.73_____-44.23
Drift_____ 8.98______1.53


Comment:
Drift attempts to isolate the changing nature of the population. The number of AAs is the same in each year by definition but their contribution changes because their proportion of the whole changes.

The fact that new juniors are on average weaker adds to total variance.

New entrants internal variance is offset by that of leavers. The balance for seniors reduces variance while that of juniors adds to variance. It is not clear why this is. My guess is that new juniors play more amongst themselves and the bi-product of the new entrant algorithm significantly widens variance. Whereas if new seniors play more against survivors then this effect is dampened.

The contribution of survivor’s internal variance should concern the graders. It may be that there are that the groups are not sufficiently homogenous and so movements in such groupings is not being isolated. However if this result can be taken at face value then the following could contribute:
1) the recalibration was overcooked and the system is still unwinding to a more natural state
2) The wide variance due to the new entrant algorithm is contaminating survivor statistics
3) The result may be an artefact of the effects of different groups playing different numbers of games per year.

We can expect this situation to continue in the current year. The effect of ageing is dampened. Those in categories C-E have a material proportion of their graded games based on prior years. This can be illuminated by looking at the two biggest groups, AA and CC, which are of almost identical size (1196 v 1200). The AA internal variance source contributes 3.54 to the 22.08 above, and the change in average contributes 2.46 of the 4.52. The respective CC numbers are 6.18 and 0.05. My reading of this is that most of the overcooking came out of the AAs in 2009 since their games in that year immediately contributed, whereas this effect on the CCs is still working through. The relative contrbutions to the averages source suggest that secondary effects are working through the system.

There are some sources of stretch (e.g. junior averages) that should not effect the matching expected score issue. There are others that are likely to be demonstating that the issue is emerging again. Only updated “Grist Graphs” will answer this.

The graders thought that the stretch issue needed to be fixed two years ago. If they still think this is the key priority (I don’t) then they need to consider modifications to avoid a future sudden jump that occurred last time. What appears to be happening is that the justifiable desire to include grades for new entrants as early as possible, on a calculation with a wide margin of error, is weakening the overall integrity of the system. There may also be contributions based on the the way prior year performance is included, but the data is not available to investigate this.

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

Re: 2010 Grading list

Post by Roger de Coverly » Wed Nov 24, 2010 10:33 am

Brian Valentine wrote:The balance for seniors reduces variance while that of juniors adds to variance. It is not clear why this is.
Is it worth observing that new entrants amongst the adult players, those above beginner level anyway, will be experienced players? As a consequence their standard of play will be consistent even on their initial games. I'm suggesting this may apply to returnees (players who are re-entering the grading list), but it certainly applies to foreigners with well-established international ratings.

Perhaps unusually for a domestic rating system, the ECF now makes no attempt to remove or not grade players who are "just visiting" for the 4NCL, Hastings, the London Chess Classic etc. In days gone by, such players would be ignored unless the graders knew that the players were locally resident. Prior to the mid eighties, the grading system was region based, so the SCCU had a list, the MCCU similarly etc. Playing in a out of area Congress was hit or miss as to whether your results got consolidated into your "home" area. So graders had to identify "home area" players and this would have excluded visiting players from domestic grading.

If you look at the top end of a list of new entrants, it's dominated by players playing for the first time in an English tournament or players who didn't play last year. Similarly leavers are players who didn't play last year.

Whilst events like Hastings and Gibraltar take place every year, there may be an impact at the top end with events joining and leaving the calendar. Perhaps it's been relative stable, with the loss of the Isle of Man offset by the three years of Liverpool events in turn replaced by the London Chess Classic. Gibraltar has been growing as well.

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

Re: 2010 Grading list

Post by Roger de Coverly » Wed Nov 24, 2010 10:55 am

Brian Valentine wrote:There may also be contributions based on the the way prior year performance is included, but the data is not available to investigate this.
I believe there's been a change in the method used. As an example consider a player who plays a 5 round May congress every year (Frome say) and 20 other games.

Previously his 2010 grade would have been (25/30) * 2010 performance plus (5/30) * 2009 performance.

It now becomes (25/30) * 2010 performance + (5/30) * Frome 2009 performance.

Expressed another way

(20/30) * 2010 performance to end April + (5/30) * Frome 2010 + (5/30) * Frome 2009

Running forward a year the 2011 grade is

(20/30) * 2011 performance to end April + (5/30) * Frome 2010 + (5/30) * Frome 2011

So the Frome performance is double counted and has a disproportionate effect on his grade. Across the whole population, you might expect the means to be unaffected but it might increase the variance.

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

Re: 2010 Grading list

Post by Brian Valentine » Wed Nov 24, 2010 12:25 pm

Roger, I can't disagree with either of your posts. The guy with the lowest reference number on my list plays just 5 games a year (not sure its always at Frome though :) ) and I think he is likely to be the most extreme example of this situation, but I assume his activity is unusual.

Sean Hewitt

Re: 2010 Grading list

Post by Sean Hewitt » Wed Nov 24, 2010 12:32 pm

Roger de Coverly wrote:I believe there's been a change in the method used. As an example consider a player who plays a 5 round May congress every year (Frome say) and 20 other games.

Previously his 2010 grade would have been (25/30) * 2010 performance plus (5/30) * 2009 performance.

It now becomes (25/30) * 2010 performance + (5/30) * Frome 2009 performance.

Expressed another way

(20/30) * 2010 performance to end April + (5/30) * Frome 2010 + (5/30) * Frome 2009

Running forward a year the 2011 grade is

(20/30) * 2011 performance to end April + (5/30) * Frome 2010 + (5/30) * Frome 2011

So the Frome performance is double counted and has a disproportionate effect on his grade. Across the whole population, you might expect the means to be unaffected but it might increase the variance.
@Roger : Is that right? I haven't seen it announced, though to be fair I haven't looked either! How do they handle the many games submitted for grading that don't have a date?