Monday, April 18, 2011

Letter to the editor II

I had originally intended to post Part II of my Anti Chess trilogy today (see Invisible Anti Chess last Monday for Part I), but I'm afraid I have to admit I'm running rather late.

Chessing - six games in the last two weeks - has rather been getting in the way of blogging of late. I wonder if there is a record for this sort of thing: 6 games for 6 teams in 4 separate leagues played under 3½ sets of rules for 2 different clubs - all inside 10 days. Fortunately I was in reasonable form and managed to score 50% which doesn't sound that good, I know, but by my standards the opposition was pretty decent.

As it happens five of those six opponents were graded ten points or more higher than me, my grading performance from the +1 =2 -2 I scored working out to 171.4. In comparison, my previous five games had been against folk graded at least 20-30 points below me so while my score, +3 =2 -0, was much better my performance rating, 168.8*, was a tad lower. Much of a muchness, then, although it could easily have been very different. Had I won the clearly winning position I had against Mr 179, or even only drawn it, there would have been a notable contrast between the outcomes of those two batches of games.

After my experience of the last couple of weeks, I'm not surprised that people say that playing higher rather than lower-graded opposition will boost your grade regardless of whether or not you actually get better. The idea isn't new, of course. Actually, I doubt it was particularly fresh thirty years ago when CHESS (May 1979 vol 44, #815-816) published this letter:-


GRADINGS DISTORTED!

The interesting letter from J. Anstead (CHESS 809-10) seemed to reflect a weakness in the grading system.

Mr. Anstead's theory that grading depends upon strength of opposition (and not simply one's own ability) is fully borne out by analysis of my own results:-

1965-1970
Opponents >180:270  Grading Result:192
Opponents <180:252  Grading Result:181

1971-1975
Opponents >180:255  Grading Result:215
Opponents <180:215  Grading Result:184

1976-1979
Opponents >180:113  Grading Result:205
Opponents <180:87   Grading Result:175

This sample of games, 1217, is much larger than Mr. Anstead's (92) and the results too significant to ignore. It would be interesting to hear form other players, particularly those in the National List. (I remember Ray Keene telling me on one occasion that he hasn't played anyone under 190 for years).

We all know that young players are undergraded (does this phenomenon explain my poor results?) and that adjudicated games should never be graded (weak players hang on to sacrificed material; adjudicators hang on to sacrificed material; adjudicators go by counting pieces etc. - the financial rewards don't justify really hard-working analysis).

Perhaps the answer is to grade only those games are played to a finish (excluding fast play-offs in weekend tournaments) and (a radical step) to include for grading a median of results (as in ice-skating) e.g.

Fred Pawn-Snatch Season 1979-1980
Games played 30
Average strength of opposition 155
Results W.20 D.5 L.5
Grading result 180
i.e. 155 times 30 plus 50 times 15, all divided by 30.

Exclude best 10% and worst 10% (i.e. 6 games). Say 3 at 210 and 3 at 90.

Games played 24
Averages strength of opposition 155
Results W. 17 D.5 L.2
Grading result 186
i.e. 5400 minus 630 minus 270, all divided by 24.

Imagine the differences to be obtained on players recording over a hundred gradeable games in one season! Food for thought anyway.


RON HARMAN
Waltham Abbey, 22 March 1979


Ron Harman, eh? Whatever happened to him?








Letter to the editor







* This includes estimating one opponent's grade based on those of the boards around him.



11 comments:

Tom Chivers said...

It's trivially the case that the grading system is imperfect in this way. If your season consists of losing 30 rated games to Michael Adams, your grade will be 217. But beat Ron Harman 30 times, and your grade will be 203. Presumably, the more opponents you play the more reliable and consistent your grade, provided you make exceptions for fast improvers such as juniors.

Anonymous said...

Style must come into it. There are players whose likely result is a draw against players up to and perhaps beyond 25 points higher. So you have Mr London System with a 150 grade who is more than capable of drawing with 175 opposition but incapable of putting away 125 opposition.

My guess on Ron Harman's results was that he suffered at the hands of then young aspiring players (as he himself suggests). He was after all, a well known personality with frequent letters and disputes to his name and there were a lot of aspiring players around in the 60s and 70s.

Jonathan B said...

I think that's not quite right Tom - because of the rule that counts people higher graded than you as being max 40 points above you.

So if lost 30 times to Michael Adams my grade would be 153 (effectively 30 losses against a 203).

Jonathan B said...

Presumably, the more opponents you play the more reliable and consistent your grade, provided you make exceptions for fast improvers such as juniors.

I agree with you about the exceptions... but I think more games will only bring a more 'reliable' grade if the people you play are spread evenly across the range of playing strengths.

In my case - mostly due to playing in teams with lots of people above me in the batting order who are only a few points stronger - the largest group of people I play is the "10 points or more less than me". In fact ten of my 26 opponents this season come into this category and the average difference of between my grade and these ten is 24 points.

Tom Chivers said...

"the rule that counts people higher graded than you as being max 40 points above you."

What rule is that????

Jonathan B said...

There's a rule - at least there was, I assume it's still in place - that for grading purposes the max grading difference is 40 points. This stops you winning points by losing to strong opponents and losing points by beating weaker ones.

It also leads to grade deflation in that you don't get the full points that you're due when you take a big scalp.

Jonathan B said...

@Anonymous.

I'm sure you're right. In Chess for Tigers Simon Webb writes about Uhlmann and how he scored very big against lower-ranked opposition - much higher than his expected score - but did less well against players of his own rating.

I'd have put myself in the opposite camp, but I'm not sure that's the full story. More of that on Easter Monday.

Tom Chivers said...

I knew about the rule for preventing grading gain from losses. But it seems you are right and the rule also limits gains from wins. How bizarre! I think this is by accident:

" There is a proviso that if your opponent's Grade differs from yours by more than 40 points it is assumed to be exactly 40 above (or below) yours. This is to prevent a player increasing his Grade by losing to a much stronger player, or decreasing his Grade by beating a much weaker player. "

I'll ask the ECForum...

Martin Cowley said...

I seem to have had a similar experience to Ron this season. Because I'm on quite a low board for all the teams I play for, only 4 of the 24 people I played were higher graded than me. I scored 148 against them and only 132 against the other 20.

Anonymous said...

So what *did* happen to Ron Harman - anybody know??

(quite possible he is no longer with us, of course......)

Jonathan B said...

Hello Anonymous,

Not that you would be able to tell unless you already knew it, but that 'whatever happened to him' was rhetorical.

RH remains a well-known figure in London League circles. He was playing on the board next to mine the last time I was at Cavendish.

Ron's Grading Card