Chess Scotland Noticeboard

Daniel Rocks

Hi, having hit the 200^ rule in 2 consecutive seasons, I've noticed a possible flaw:

1. Firstly, at the start of this season, my average performance was around 1300 level and so I lost to some 1400 level players, later, as the season progressed, my playing level has increased somewhat and so, as of late, my average performance has been around the 1750 mark (based on last 3 tournaments) - why then do the 1400 players who I lost to at the start of the season get credit for me being a 1650 level player (according to my new rating) when they beat me when I was 1300 level? - This, to me, appears to be quite a flaw and could make some grading calculations in-accurate. Still, it's only a number at the end of the day but I'm intrigued to know why this is the case? Is it just one of those things where nothing sensible could be done to prevent it and it should just be allowed to happen? I suppose it's not a big deal but I'd be thankful if an arbiter or someone who knows lots about gradings would enlighten me on the subject at hand.
_________________
Daniel Rocks

DMB · King Joined: 08 Mar 2007 Posts: 267

A grading is an approximation of an ever changing reality - said Mr Elo. So if someone is improving fast its hard to ever get the grading "accurate". It's much more likely to be right with established players who are stable.

The 200 up was something I used as a quick fix to move improving players ahead quickly and to compensate opponents. One of the implications could be as you described of early season opponent being compensated when he doesn't deserve the full bonus from a season long 200 up performance.

So you then have to think of something better to replace and secondly you need someone to program the changes (NB: the latter is not a trivial issue). At the moment I really don't have the time to search for an improved system - but if someone wants to spell out exactly what would be a better way it will be considered.

Daniel Rocks · Posted: Mon Jun 25, 2007 6:21 pm Post subject:

No, no... I think the current system is very good and the problem of early season opponents being credited for a season long 200^ rule is just one of those things. Smile

_________________
Daniel Rocks

Mike Scott · King Joined: 01 Feb 2007 Posts: 676 Location: Edinburgh

I would concur with DMB's comments - the current system works well and its hard to think of a way to improve it, any sytem is bound to give rise to occasional funnies.

One of these has happened to my son, who has played very well for the first 11 months of the season (scoring 27.5 / 28 vs players graded below 1120 as well as good scores against higher graded players - including the better of an allegro draw against a 2100+ player) and has had a web listed grade of between 1316 - 1340 for the last 6 months of that 11.

In June he lost form and dropped 4 points from 11 games against players graded on the whole below 800, resulting in over 200 grading points being lost. More specifically scoring 0.5 pt from 3 games in the PI this past weekend, caused his grade to drop 184 points by dropping out of 200^.

The impact of these three games compared to the other 67 games played seems out of proportion. I know the grading is determined by the whole season results and it does not matter the order of the games etc (see previous postings from Heth) but in practice you play games in sequence and have the choice to play (or not) these games.

As I said at the start I am not arguing against the grading system, and clearly it is reflecting my son's poor form this month. Afterall there will be more scope next season for grading prizes Very Happy

and grading gains provided the loss of form is just a glitch - its swings and roundabouts.

DMB · King Joined: 08 Mar 2007 Posts: 267

I'm afraid young Jonathan hit one of these arithmetic glitches with the way the 200-up operates. I am sure there could be something better - but there is very little time to think of what it should be. Perhaps some other country has a smart way of dealing with the fast improvers.

Most of the time spent on grading is connected with facilitating the job of the area graders, getting the data into the system, publishing grading output on the website and allocating the correct results to the the correct player. There is constant work done on improving the accuracy of the dbase - first names, spellings, ages, club etc.

There is very little study on the theory of the actual arithmetic used in CS grading. The 200-up was a bit of sellotape to fix the fact that the grading system in the 1990s had been flooded by loads of very low rated juniors, who then soared after they improved. The normal arithmetic which was good enough for the majority wasn't catching up with how well these novices had improved. The ELO book talks about rewarding exceptional performances - 200 is an ELO class interval - so 200-up was classed as the level at which improvement would be deemed worthy of a bonus.

Two grading lists a year might help to reflect improvement quickly - say January 1 in addition to August 1.
ISSUES:
Does that mean the junior additions are halved - it probably would.
What about the Grand Prix - to which section would people be allocated - probably the Jan 1 grade.
What about club data - just take it as and when supplied.
Other issues?

However at the moment it's not possible to make any pronouncements on big changes since we have some programming issues to address. Alex Bisset has indicated that he wont have as much time to support the program as previously. Gordon Rattray of Polytechnic is now looking at making the program more web based. So as is usually the case we're back at dealing with practical issues of how to do data capture/publication rather than the theory of the grading numbers.

Mike Scott · King Joined: 01 Feb 2007 Posts: 676 Location: Edinburgh

One year when Jonny had gone 200+ I did check what would have happened had the gradings been published 6 monthly and found it actually made no significant difference on his grade.

My only thought is that the 200^ type of rule would introduce a smaller step change/discontinuity (in both directions) if it were a 150^ (or lower) rule. There is some logic in reducing it to a level similar to the junior addition.

Daniel Rocks · Posted: Tue Jun 26, 2007 11:39 am Post subject:

Phil Thomas · King Joined: 26 Mar 2007 Posts: 758

I see another problem in the current system. When a rapidly improving junior has got gone 200 up and has played 100 plus games he/she knows that the grade won't change whatever the results for the rest of the season.

So how about doing the same as FIDE. Abandon the 200 up rule and do 4 grading lists per year?

Would that create extra workload for all those who work with gradings?

admin · Site Admin Joined: 09 Jan 2007 Posts: 1386

I think the idea of 2 gradings a year was floated previously, from memory it was rejected as it would make it difficult because of the timings for getting results in (one period would be over Christmas), what the effect on the leagues would be with grades being reset during the year and the confusion for tournament organisers

Alan Tate · King Joined: 01 Feb 2007 Posts: 377 Location: Edinburgh

I think if it's not too much extra work for our volunteers, who i should add, work in a thankless role, then it could be tried?
One aspect of FIDE's system i do not like is that the cut off point for new grades is a month before, and the grades come out every 3 months! It defies logic...

DMB · King Joined: 08 Mar 2007 Posts: 267

Never seen these postings till now - is having to log in a good idea? No matter.

Two updates a year was not rejected. It was mooted and then I pointed out the implications of having two updates. I wont repeat that all again. Four updates - why not monthly? It's easy to make suggestions - working out the practicalities much more difficult.
Anyway CS members get weekly updates at the moment - not sure many other feds offer that service. (Albeit in relation to a once a year start grade).

At the moment programmer Gordon Rattray is creating a new grading application - we are not making any changes to calculation methods.

Our programmer will look at arithmetic once we have a working program. (The existing Alex Bisset prog cannot be amended - so we need to create a new prog which we are then able to change as and when required) It all takes time.

************

Ireland has an interesting system based on continuous ratings where grades change after every event. http://www.icu.ie/ratings/index.php

Grades are published 3 times a year - presumably to create a figure you quote to enter tournaments.

How does it deal with leagues which takes place over several months? How long does an event have to submit data? I wonder how we would work the Grand Prix. Do they use junior additions? I suppose the printed list is finished - 3 pdfs instead maybe? etc

Would that be appropriate for Scotland. I have asked Mark Orr for further information.

admin · Site Admin Joined: 09 Jan 2007 Posts: 1386

Pdf's

We'll pull you into the 21st century yet Wink

Peronally I think it is a fantastic idea and something that we should be looking to work towards, but (and it is a big but), we need to make sure we have it right and have the resources in place for doing it.