new grades wow brandon clarke shot up 206

General discussions about ratings.
LozCooper

Re: new grades wow brandon clarke shot up 206

Post by LozCooper » Thu Aug 13, 2009 11:39 am

michele clack wrote:
That's about a par result for him as his true strength is probably about 190 at the moment.

By the way he's leading the European Union Youth Championships at the moment with 6/7
I must have a look at the European reports. I noticed there is a prominent link on the ECF homepage, which is excellent.

If there is a 10 point junior uplift for a 13 year old, which I think is the figure, then your estimate of 190 +10=200 is just 6 points shy of his published grade. Since he played 142 games I'd have said that his grade is probably pretty accurate. A 3% variance could just be sampling error as you won't have seen all his games.

Ian (my husband) thinks Brandon was a little unlucky at Bromsgrove as he played Lawrence Cooper as black in the first round, which was his only loss. I was tied up in a long battle in my first game but Ian finished early and watched a lot of it and he was impresed with Brandon's play even though he lost in the end. It would be interesting to hear what Lawrence thinks his current strength is likely to be.

To me the inference seems to be that the new grades are OK for juniors who have played a fair few games.
Hi all,

To be fair to Brandon, I don't think he played as well as he can in that game. I won an exchange fairly early on but let him back into the game and was relieved (owing to my time shortage)to be able to sac it back into a winning ending. Having lost to a junior in rd1 of my previous weekend tournament (Dominic Rabbitte at Heywood) I was somewhat apprehensive about playing Brandon and so I concentrated harder than maybe I would have against an adult opponent! His only other dropped half was in a game that he was winning from the opening. I did ask him what he expected his grade to be and he said about 180 (he was 148 I think on the wallchart at Worcs). I did sense he was a dangerous opponent who could take advantage of any tactical chances.

He's clearly very promising and is performing well at the European as we speak. I'm sure it won't be long before he overtakes my grade and I wish him well for the future.

Nick Thomas
Posts: 456
Joined: Sat Oct 11, 2008 9:56 pm

Re: new grades wow brandon clarke shot up 206

Post by Nick Thomas » Thu Aug 13, 2009 6:46 pm

He ended up =2nd with 7/9 and the only unbeaten player in the tournament having drawn the last round with black against the French star Gary Giroyan.

He scored a staggering 7.5/8 for Warwickshire select last season in the 4NCL (dropping a draw to Alan Walton after missing a mate in 6!!) so we might have to consider putting him on board 1 from now on (I think he's the highest graded player in our squad!!).

Sean Hewitt

Re: new grades wow brandon clarke shot up 206

Post by Sean Hewitt » Fri Aug 14, 2009 12:30 am

Brandon is the reigning Leicestershire County Champion and regularly beat 180 opposition in the league last year. We expected him to be circa 200 this season and I think his grade is about right, give or take 5 points.

Of all the juniors whose grades are off the scale, Brandon is not one of them.

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

Re: new grades wow brandon clarke shot up 206

Post by Roger de Coverly » Fri Aug 14, 2009 12:43 am

Nick Thomas wrote:
He scored a staggering 7.5/8 for Warwickshire select last season in the 4NCL
Sean Hewitt wrote:Of all the juniors whose grades are off the scale, Brandon is not one of them.
Danger signals for the out of control new player estimation recursion

(a) very high (or very low) percentage score
(b) played against unrated opposition

I'd suggest an audit might struggle to reproduce the exact calculation but I'd agree it's within the bounds of plausibility.

Sean Hewitt

Re: new grades wow brandon clarke shot up 206

Post by Sean Hewitt » Fri Aug 14, 2009 12:27 pm

Roger de Coverly wrote:
Nick Thomas wrote:
He scored a staggering 7.5/8 for Warwickshire select last season in the 4NCL
Sean Hewitt wrote:Of all the juniors whose grades are off the scale, Brandon is not one of them.
Danger signals for the out of control new player estimation recursion

(a) very high (or very low) percentage score
(b) played against unrated opposition

I'd suggest an audit might struggle to reproduce the exact calculation but I'd agree it's within the bounds of plausibility.
To be fair Roger, he played 20 odd games in the Leicestershire league and scored much closer to 70% I think. I don't know his overall % but he played 142 games - I don't think he will have scored above 80% overall.

Nick Thomas
Posts: 456
Joined: Sat Oct 11, 2008 9:56 pm

Re: new grades wow brandon clarke shot up 206

Post by Nick Thomas » Fri Aug 14, 2009 4:46 pm

Of the 133 games I know about he scored almost 82%.

Nick Thomas
Posts: 456
Joined: Sat Oct 11, 2008 9:56 pm

Re: new grades wow brandon clarke shot up 206

Post by Nick Thomas » Sun Aug 30, 2009 5:28 pm

Nick Thomas wrote:There are some strange looking junior grades including callum killpatrick 234 which has been mentioned and Richard Weaving top of the junior rapidplay list (228). Richard's true rapidplay strength is about 170 -180 I would guess.

Brandon is probably not the best example of a strange rise as he was certainly underrated last season and is improving very rapidly. I would estimate his strength (compared to myself and others who have ended up around 200 this year) to be about 190. When you take into account the junior increment he's not too far off.

I was more surprised that Felix Ynojosa rose above Yang-Fan Zhou by such a large margin.
Well Brandon is down to 201 which is now where it should be allowing for the fact that it includes a junior increment. Richard Weaving's rapidplay grade is now a much more sensible 188.
Last edited by Nick Thomas on Sun Aug 30, 2009 8:00 pm, edited 1 time in total.

Neill Cooper
Posts: 1298
Joined: Sun Feb 24, 2008 4:43 pm
Location: Cumbria

Re: new grades wow brandon clarke shot up 206

Post by Neill Cooper » Sun Aug 30, 2009 6:07 pm

I've noticed that lots of strong (grade over 100), active (Category A), older (11 to 17) juniors have gone down by exactly 5 points. Quite a few others by 6 or 7 points. Coincidence, or a change to the + 5 point increment? [Out of 32 Surrey Juniors in that category, 10 players went down by 5 points and 7 players went down by 6 points.]

User avatar
Rob Thompson
Posts: 757
Joined: Wed Jul 15, 2009 12:03 pm
Location: Behind you

Re: new grades wow brandon clarke shot up 206

Post by Rob Thompson » Tue Sep 01, 2009 12:35 am

As an U-17 this year (so 16 on the grading lists) i have moved from 171 to 170. Idk how this compares with the South-East grades, but of the south-west juniors most maven't moved much, and not by a regular 5 points every time.
True glory lies in doing what deserves to be written; in writing what deserves to be read.

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

Re: new grades wow brandon clarke shot up 206

Post by Brian Valentine » Tue Sep 01, 2009 9:20 am

The grading team have clearly listened to feedback on the "junior issue", for which they should be commended. However the August solution can only be considered, as best, as a patch. Playing around with the grading list, there seem to me to be two fundamental themes:
1. The best indicator, from the information available on the download, for a junior's grade is their prior year grade. It seems to be perverse to (generally) discard this valuable information each year to determine the latest grade.
2. Despite Callum Kilpatrick and a few others, the method throws up the very problem the revamp of the system was trying to fix. Relatively speaking, stronger junior players are underperforming relative to weaker players. For instance the best fit for juniors who were category A or B in both years (and therefore those that will have their old grade discarded) is
new grade = 91% old grade + 22.

Five points (usually) are explained by the junior adjustment. There may be some underlying inflation - above the junior adjustment -in the method as well.

Playing around with the numbers, I note that games played in the year is a more significant guide to junior improvement than age. This theory has been espoused by others, but is now fuzzily visible in the data.

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

Re: new grades wow brandon clarke shot up 206

Post by Roger de Coverly » Tue Sep 01, 2009 10:36 am

Brian Valentine wrote:Despite Callum Kilpatrick and a few others,
That's the 40 point rule in one of its manifestations. If you score near 100% against players you are 40 points above, the new grade would be "last year" grade + about 10. In some cases this is a deflationary force because it restricts the available gain of points. On the new system, the "last year" grade is discarded so the new grade comes out at "estimate" + about 10. It even compounds on itself because the higher the "estimate", the more games are caught by the 40 points rule.

I think we finally know where the grading team are coming from. They have plotted S shaped graphs showing the percent score against grading difference and concluded that because they don't follow the pattern of the ECF or Elo expected scores that this implies that the distance between "strong" and "weak" players needed to be reduced in order to improve the fit. To my mind that's just one possible interpretation. The graph is also measuring the incidence of randomness (fluctuations in form) and ratings lag and possibly even non-linearity. With juniors they run lag and inconsistency so that whatever weird and wonderful fixes they concoct the graphs just don't fit. They've now just gone self-fulfilling - so you improve the "reliability" of the start season grade as a predictor of the season's results by waiting until the end of the season to calculate it.

The whole debate would have been a lot more informed if the s shaped graphs had been published as soon as they had become available and published as part of the annual review of the grading system. This could have sparked a wider debate about what conclusions to draw from them and whether it indicated the need for any corrective action.

If I go a bit mathematical, I could describe the list of prior year grades as a vector and the results as a results matrix. The results matrix is then applied to the prior year grades to get the end year grades. Empirical observation suggests that the results matrix doesn't do much to the shape of the grades, so the statement that the grades are spreading is difficult to validate. Certainly the graph could suggest that spread may have happened in the past but it doesn't tell you that it's continuing. Other factors aside, you would expect supposed underperformance by higher rated players to shrink not expand the grading range.

On junior increments, could we not go back to the system prior to about 1990? This is that the "published" list measures performance only but events handle eligibility issues by specifying a lower grade for juniors. There's a factor in the next calculation to combat deflation. You could make the deflation and eligibility factor time dependent - say 5 points for events up to Christmas and 10 points after that. I'm not even sure I know what the junior increments are anymore.

Ian Lamb
Posts: 496
Joined: Tue Jul 14, 2009 1:45 pm

Re: new grades wow brandon clarke shot up 206

Post by Ian Lamb » Tue Sep 01, 2009 5:43 pm

i see brandon dropped to 201 on the second list out so he down 5 points on first list.http://grading.bcfservices.org.uk/getpl ... ers=clarke

Ian Lamb
Posts: 496
Joined: Tue Jul 14, 2009 1:45 pm

brandon clarke down to 201 on second list !!!

Post by Ian Lamb » Tue Sep 01, 2009 5:51 pm

for those who saw him at 206 he now down to 201 in second list what a mess.

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

Re: new grades wow brandon clarke shot up 206

Post by Brian Valentine » Wed Sep 09, 2009 10:07 am

This note could go on one of several other threads to the grading debate, but I've choosen here.

I've been playing around with a matrix version of the ecf grading method. The final equations support many of the points made in this and other threads.

I have written it up as a 7 page word document, so:
1. It is too long to post here and won't copy well (loads ofsuperscripts)
2. It probably has a very small potential audience!

There is a supporting spreadsheet that demonstrates that the two approaches (iteration and matrices) give the same answer on a 10 player population. It can also be used to model some of the situations observed.

Clearly a matrix approach to a 10,000ish player population is not sensible, but I was hoping it could show the various influences in the system. Frankly, they are not obvious to me!

Nonetheless it does lead me to another line of enquiry, which I am just beginning. If anyone is interested drop me a private mail and I'll send copies.

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

Re: new grades wow brandon clarke shot up 206

Post by Alex Holowczak » Wed Sep 09, 2009 10:12 am

Brian Valentine wrote: I have written it up as a 7 page word document, so:
1. It is too long to post here and won't copy well (loads ofsuperscripts)
2. It probably has a very small potential audience!
You could save it on http://pastebin.com and leave the link to the pastebin.

Or, you could save it as a google doc, and "share" it with everyone, so that they can view it. Then just leave the link to it in here.