New grades (split from Editorial thread)

[John Upham]

Secondly, no evidence has been presented that ELO is actually better than the Clarke system.

As a past academic I'm never that happy seeing assertions displaying this level of confidence.

Maybe "I'm not aware of any evidence has been presented that ELO is actually better than the Clarke system." would be a more credible statement.

I could have presented such evidence to the Surrey County Chess Association, for example.

During my DPhil (on Monte Carlo numericlal methods as applied to Molecular Quantum Mechanics) I felt I had devised a new method for sampling randomly from a weighted histogrammic distribution. I prepared a paper for publication and at the last minute uncovered similar work in an obscure journal from Rumania.

My thought that nothing had been published was demolished.

[Ian Thompson]

Regarding junior increments. Its worth remembering that the junior increment is supposed to reflect the average expected improvement in the junior in the next 12 months. So a juniors grade just after the publication of the grading list should appear inflated. Wait until next April though. If these junior grades still appear over graded the increment was too high. If however they are kicking our butts, then it was probably about right!! Given that Howard says the new grades give a good fit to expected scores, I fear we may still hate playing these kids come the end of the season.

If this is what has been done, isn't that a mistake? Surely the junior increment should estimate the expected improvement of a junior mid-season, not end of season. If the increment estimates end of season strength then all the junior's opponents, except those played at the end of the season, will be playing someone whose true strength is less than their grade. If the increment estimated mid-season strength then the junior's opponents in the first half of the season would gain from playing someone whose true strength was less than their grade, and the junior's opponents in the second half of the season would lose from playing someone whose true strength was more than their grade. Overall, gains and losses would cancel out, which is what you want.

Ian Thompson

[Sean Hewitt]

Secondly, no evidence has been presented that ELO is actually better than the Clarke system.

As a past academic I'm never that happy seeing assertions displaying this level of confidence.

Academics are never happy, period!! I think this statement holds good though.

I could have presented such evidence to the Surrey County Chess Association, for example.

And I could have been to the moon. It's semnatics. The pertinent question is, have you? Have you presented such evidence to the SCCA John?!!!

[John Upham]

The pertinent question is, have you? Have you presented such evidence to the SCCA John?!!!

No, but its a good idea. Thanks!

[Sean Hewitt]

Regarding junior increments. Its worth remembering that the junior increment is supposed to reflect the average expected improvement in the junior in the next 12 months. So a juniors grade just after the publication of the grading list should appear inflated. Wait until next April though. If these junior grades still appear over graded the increment was too high. If however they are kicking our butts, then it was probably about right!! Given that Howard says the new grades give a good fit to expected scores, I fear we may still hate playing these kids come the end of the season.

If this is what has been done, isn't that a mistake? Surely the junior increment should estimate the expected improvement of a junior mid-season, not end of season. If the increment estimates end of season strength then all the junior's opponents, except those played at the end of the season, will be playing someone whose true strength is less than their grade. If the increment estimated mid-season strength then the junior's opponents in the first half of the season would gain from playing someone whose true strength was less than their grade, and the junior's opponents in the second half of the season would lose from playing someone whose true strength was more than their grade. Overall, gains and losses would cancel out, which is what you want.

Ian Thompson

Ian, you've misunderstood me so I apologise for not being clearer in m original post!

When I say the increment intends to reflect the average increase in a juniors performance over 12 months what I mean is this. Say a player is 100 strength at the start of the season, and is expected to increase to 130 lineally by the end of the season. Then his average playing strength over the course of the season will have been 115, 15 points above his actual starting strength. Therefore his increment should be 15. I think this is what you were saying should happen - and indeed it does.

This is why I said juniors should appear over graded at the start of the season but still kick our butts (ie appear under graded) at the end of the season!

[John Upham]

I assume that it now occurring to an increasing number of persons that a major step forward would be to increase the frequency of the ECF Rating List updates / publications.

Call me old fashioned but this will reduce the need for silly fudge factors and bizarre future predictions of Junior Ratings.

It is now 2008 and the technical means of achieving this are with us.

I'm not going to debate if League Rating Officers prefer to store all the results up for a frenzy of activity in May /June.

We can't go on looking backward (not backwards).

John

[Roger de Coverly]

Say a player is 100 strength at the start of the season, and is expected to increase to 130 lineally by the end of the season.

Working backwards his strength at the start of the previous season would have been 70 and thus his grade would have been 85. So the addition to the grade to estimate the "mid season" strength as 115 is 30.

I agree with Ian and others that you don't get many players improving by 30 points every season. In practice what happens is a large leap followed by a period of consolidation as they adjust to the higher standard of play at the higher level.

On a different topic, I've got part of the 1981 grading list (175 and above) in front of me ( it's in the 81-82 year book).

Some observations:-

The 1981 top 100 went down to 202 as does the normal version of the 2008 list ( ENG players only)

A 1981 grade of 178 was just outside the top 500. A 2008 grade of 178 would be about 400th. At 172 my ranking is at about the same as it was in 1981.

In the 1981 list, it's a struggle to find players who were born before (say) 1945. Many of the names in 1981 are also in 2008.

John Nunn was top at 245.

The rest of the top 10 were
Miles 243
Stean 243
Short 239
Chandler (Murray) 238
Mestel 238
Speelman 238
Keene 234
Hartston 230
Littlewood (Paul) 230





;

[Sean Hewitt]

I assume that it now occuring to an increasing number of persons that a major step forward would be to increase the frequency of the ECF Rating List updates / publications.

Call me old fashioned but this will reduce the need for silly fudge factors and bizarre future predictions of Junior Ratings.

It is now 2008 and the techincal means of achieving this are with us.

I'm not going to debate if League Rating Officers prefer to store all the results up for a frenzy of activity in May /June.

We can't go on looking backward (not backwards).

John

Absolutely right. 6 monthly lists are a minimum requirement, and frankly quarterly is where we should be.

My view is if the graders dont like it, then leagues need to get new graders!

[Sean Hewitt]

Say a player is 100 strength at the start of the season, and is expected to increase to 130 lineally by the end of the season.

Working backwards his strength at the start of the previous season would have been 70 and thus his grade would have been 85. So the addition to the grade to estimate the "mid season" strength as 115 is 30.

??? Why ??? Just because a player improves 30 points in one season does not mean that you can deduce that he improved 30 points in the previous season!!

On a different topic, I've got part of the 1981 grading list (175 and above) in front of me ( it's in the 81-82 year book).

Some observations:-

The 1981 top 100 went down to 202 as does the normal version of the 2008 list ( ENG players only)

A 1981 grade of 178 was just outside the top 500. A 2008 grade of 178 would be about 400th. At 172 my ranking is at about the same as it was in 1981.

I don't think this tells us anything. However, does the yearbook give the quartiles or quintiles? That would be useful!

[Martin Carpenter]

Anyone worried about the new grades in general should look at http://grading.bcfservices.org.uk/newgrades.php - lots of useful information including how they're being calculated, where the %iles are moving to etc. It's really very helpful.

To go back to the point that Richard Palliser raised earlier about junior grades - this isn't related to any form of junior increment. They've still got that and the level seems sensible enough to me (+15/year ago 0-11, +10/year 12-14 and +5/year 15-17 - I don't know what it currently is.).

However what they're also doing is giving every junior in the country a considerable boost in relation to the rest of the grading list.

Here are their (guideline!) formulae: Adult New Grade = 0.79 x Old Grade + 45; Junior New Grade = 0.76 x Old Grade + 64 . So fundamentally they're seemingly contending that - independent of their year on year improvement - juniors are currently undergraded by about 15-20 points across the board. In fact - as with the two people Richard quoted - they've been a little more generous than this in places.
(there's one York junior gone from 32(old) to 96(new) )

This is really quite a big thing to be doing I think. Is it justified? I've really not noticed it playing in the North - the stronger juniors here seem to play enough older players to keep their overall grades in line with the rest of us. Certainly the two that Richard was talking about have.

Perhaps it is justified in the south where - I must assume - all the juniors end up mostly playing each other and so can create their own grading pool.

I'll be interested to hear the reasons behind this move.

[Roger de Coverly]

Why ??? Just because a player improves 30 points in one season does not mean that you can deduce that he improved 30 points in the previous season!

I presumed you were just invoking a 30 points a season model of junior improvement. I think it's actually 15/10/5 but it's still greater than the previous scale. The reality check is that if you have a player with a current grade of 120 and you give them a new grade of 170, do you think they can score 50% against an average field of 170?

I don't think this tells us anything. However, does the yearbook give the quartiles or quintiles? That would be useful!

I don't actually think we have an agreed definition of the terms inflation and deflation.

To me they relate to a time series which shows an upward or a downward trend. For example the price of a pint was 13p in 1971 and it's now £ 3 or thereabouts - That's inflation and you don't need the average price of a pint to notice the effect.

If you have a group of players with grades of 175 and above in 1981 and many of them still have grades of 175 and above in 2008, that's not inflation or deflation. If we all had grades in the 220's that would be inflation, similarly if we had grades in the 125's that's deflation.

Your definition please.

[David Robertson]

If the price of pint in 1971 was 13p, and now it is 300p, that indeed is an example of the effects of inflation. But it's not the whole story by any means. What we need to know is the relativity of pint prices over the period - that is, has a pint become more or less expensive relative to other prices, incomes and so forth. To get at that answer, economists apply a deflator, thereby rendering the 2008 price of a pint in 1971 terms. I've no idea what that might be, but an intelligent guess might give us a figure of 10p - that is, the current price of a pint is in fact relatively cheaper now than in 1971. Similar, revealingly instructive analysis can be (and is) undertaken for cars, computers etc (or indeed, anything).

Something similar is being attempted with the grading system. But there's a difference. Whereas we can see clearly that the price of a pint has inflated over the period, we can be surprised to learn that it has in fact become less expensive to us (ie 13p in 1971 is 'more than' 300p in 2008 where money is used to price pints). That's because inflation acts on money, not on the pint; and we can print more money, hence inflation. But we don't generate more grade points to the same degree; while the win-draw-loss machinery tends to stabilise the grading currency by sloshing points backwards and forwards in a more or less regular manner.

To muddy the water further, there's evidence of grade inflation at the international level. I know this because, with John Saunders help, I've been comparing the mean grades of the strongest British tournaments since the war. On that basis, the forthcoming event in Liverpool next week is vastly the strongest ever held in Britain. The mean grade of the top 14 players for Liverpool (2008) is 2628. Philips & Drew (1984) in comparison was a mere 2591; and the GLC Masters (1986), a piddling 2561. Nothing else comes close.

But look further. As John Saunders explains to me: in 1988, the world's 30th strongest player was Polugaevsky (2590); in 2008, it's Rublevsky (2699). And there were only two 2700+ in 1988; today, there are 29.

Alas, not even I can fly in the face of the evidence, and claim Liverpool (2008) is the strongest-ever. It's certainly the strongest Open tournament ever held in Britain; that's something to be pleased about. But the inherent inflationary tendencies in the grading system, like those in degree classifications, make comparisons over time rather awkward.

David
Atticus CC

[Sean Hewitt]

Alas, not even I can fly in the face of the evidence, and claim Liverpool (2008) is the strongest-ever. It's certainly the strongest Open tournament ever held in Britain; that's something to be pleased about. But the inherent inflationary tendencies in the grading system, like those in degree classifications, make comparisons over time rather awkward.

I agree with everything that you have said David. Indeed, the goal of correcting grades to pre-inflation levels is / was probably unobtainable - even if the grades could and have been corrected relative to each other. It probably would have been best to add a constant so that the new grades do not look like the old grades.

However, hopefully we will move to ELO in the future and render that unnecessary anyway!

[Matthew Turner]

Inflation is measured by, New Price - Old Price/Old Price * 100.
If the price of a pint goes from 13p to £3 then this is inflation, GDP deflators, wage deflators or big mac indices do not effect this fact. The only way that it would not be inflation is if the quality of a pint had improved by 2200% (which I think is unlikely).
I certainly think the top chess players have improved enormously over the past twenty years and probably higher FIDE ratings reflect this. It is therefore, probably true that in absolute terms the Liverpool EU Championships is the strongest event ever held in Britain.
I think Roger's observations of the 1981 and 2008 grading lists are somewhat misleading, the 100th highest player might still be 202, but when you adjust for quality then I would suggest that this is actually evidence of deflation. When you consider this it seems strange that the grading corrections tends to lead to the top players losing points.