Monthly Grading Lists

General discussions about ratings.
User avatar
IM Jack Rudd
Posts: 4818
Joined: Tue Apr 17, 2007 1:13 am
Location: Bideford
Contact:

Re: Monthly Grading Lists

Post by IM Jack Rudd » Wed Jul 22, 2009 2:12 am

David Haydon wrote:Why don't we just have a ranking system like they do in tennis?
Tennis doesn't "just" have a ranking system any more than chess does. It has a rating system, and the rankings are calculated from that using a descending sort. If you want to come up with a ranking system that doesn't reflect an underlying rating system, you're more than welcome to do so - but be warned, it's not going to be easy.

The real difference between chess and tennis is the relative focus - tennis players are described in terms of their ranking, chess players in terms of their rating. And that stems from the difference in the rating systems - tennis uses a simple accumulation system, in which the values of your points totals are mostly meaningless as a predictor. Chess uses statistically-based systems, designed with the prediction of future results in mind.

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

Re: Monthly Grading Lists

Post by Alex Holowczak » Wed Jul 22, 2009 8:59 am

FM Jack Rudd wrote:
David Haydon wrote:Why don't we just have a ranking system like they do in tennis?
Tennis doesn't "just" have a ranking system any more than chess does. It has a rating system, and the rankings are calculated from that using a descending sort. If you want to come up with a ranking system that doesn't reflect an underlying rating system, you're more than welcome to do so - but be warned, it's not going to be easy.

The real difference between chess and tennis is the relative focus - tennis players are described in terms of their ranking, chess players in terms of their rating. And that stems from the difference in the rating systems - tennis uses a simple accumulation system, in which the values of your points totals are mostly meaningless as a predictor. Chess uses statistically-based systems, designed with the prediction of future results in mind.
Furthermore, tennis and golf give certain points to particular events. Taking golf as an example, you get 100 points for winning one of the four Grand Slam events, then 80 for winning the Players Championship and each tour around the world has events worth x points with people who don't win scoring fewer points. The number of points awarded to an event is based on prestige, prize money and the quality of the entry. Also, tennis and golf players play about 20 events per season. In chess, the average player in the UK will probably only play 5 at the most. So the data you generate isn't likely to be very indicative. This is why snooker - which uses an identical system to tennis in all but numbers - has to use a 2-year ranking system, it simply doesn't have enough tournaments for 1-year to be a very good indicator. As a result, the Snooker world rankings are often a bit laggy, and on good tournament over a 2-year span is often good enough to keep you in the top 16. Steve Davis reached one semi-final per year for about three years recently, and despite losing in the first round of all of his other events, he was still ranked near the top of the rankings. So the only way for a similar system to work in chess is if players enter far more tournaments.

Although, it might be interesting to see how other sports would work with an Elo-style ranking system.

JamesMurphy
Posts: 56
Joined: Tue May 12, 2009 11:54 am
Location: Manchester, UK
Contact:

Re: Monthly Grading Lists

Post by JamesMurphy » Wed Jul 22, 2009 12:31 pm

Alex Holowczak wrote:
Furthermore, tennis and golf give certain points to particular events. Taking golf as an example, you get 100 points for winning one of the four Grand Slam events, then 80 for winning the Players Championship and each tour around the world has events worth x points with people who don't win scoring fewer points. The number of points awarded to an event is based on prestige, prize money and the quality of the entry. Also, tennis and golf players play about 20 events per season. In chess, the average player in the UK will probably only play 5 at the most. So the data you generate isn't likely to be very indicative. This is why snooker - which uses an identical system to tennis in all but numbers - has to use a 2-year ranking system, it simply doesn't have enough tournaments for 1-year to be a very good indicator. As a result, the Snooker world rankings are often a bit laggy, and on good tournament over a 2-year span is often good enough to keep you in the top 16. Steve Davis reached one semi-final per year for about three years recently, and despite losing in the first round of all of his other events, he was still ranked near the top of the rankings. So the only way for a similar system to work in chess is if players enter far more tournaments.

Although, it might be interesting to see how other sports would work with an Elo-style ranking system.
I totally agree with your appraisal here. Each sport has it's own different ranking system primarily because they need to. I prefer the ELO-style ranking system to any of the others because they provide a far more accurate prediction of your current performance. An example being this year's Women's World No. 1 in Tennis, Dinara Safina. She enters far more competitions that the Williams sisters do and as a result ranks higher - but performs worse when involved in the same competition. Clearly the Williams sisters should be ranked higher because they perform better when they turn up?! Just because you can attend more competitions and consistently placed 2nd/3rd doesn't make you a better player.

As far as integrating the ECF rating system is concerned (I have a background in Application programming so have some experience here) and making it more streamlined and reactive to results, I can well understand Carl Hibbard's grief. It is an enormous task to try and make an old-archaic system integrate with a newer system (sometimes this can create more problems than it solves). It's also a significant task to write one from scratch, often requiring additional man-power which unfortunately people = £££. Chess is somewhat lacking in that. What the ECF would need to do to build such a system would probably be to hire an external contractor to build the system for them - if the ECF have that sort of money lying around - you're probably looking at £100k + easily if you want my professional opinion - then sure it's feasible. Such a system would need to be able to process a significant number of results in real-time.

Company's like the one I work for could probably do it (we deal with some of the largest UK retailers and look after their systems - littlewoods anyone?) so know what an undertaking it would be. I would predict with a team of half a dozen developers it would take the best part of a year to come up with something - but that's just a finger in the air estimate without knowing any of the other details.
http://www.jmblogger.com - My blog, chess, film and life
http://www.chorltonchess.com - the Home of Chorlton Chess Club

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Monthly Grading Lists

Post by E Michael White » Fri Jul 24, 2009 2:28 pm

Roger de Coverly wrote:
Rob Thompson wrote:Would it be possible to do the Standard play grades after 6 months, like the rapidplay ones? This is beginning to move towards the idea of monthly grades, but without being a massively drastic change.

This was proposed (and rejected) about eighteen months ago. There would be a problem with this where players didn't play in one of the six month periods. You see this in the rapid grades. In the latest list, I was among a number of players whose grade went up by a decent amount despite not playing a single game in the "most recent" six month period.
I don’t see it as a disadvantage if a players grade increases when he plays zero or very few games; it is a result of early out of date lower rated results dropping out. This can already happen in the current long play grades eg.

A new player in the year 2000 plays 10 games at 70 for a grade of 70) – (2000,10,70, 70) followed by
(2001,20,85,86.6) (2002,7,115,97.2) (2003,8,130,109) (2004,1,140,124.1) so by playing 1 game in 2004 his

grade would have increased by 15 points.

Or this example,
A new player in the year 2000 plays 12 games and is badly treated by the new starter process being given a grade of 40 -(2000,12,40,40) followed by (2001,4,115, 58.8 ) (2002,7,130,80.4) (2003,1,140,125.83) in this case

increasing by 45 points in year 2003

However if he didn’t want to play the game in 2003 but just wanted to remain graded he could just turn up play one move and resign. Whatever his oppos grade the 40 point rule would ensure he got at least -9.57 for that game. (2003,1,-9.57,113.4)

giving an increase of 33 points.

The last example makes me think during the first few years a player plays, games should not go back to 3 years. Maybe that is the new basis for juniors.

Paul McKeown
Posts: 3732
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)

Re: Monthly Grading Lists

Post by Paul McKeown » Fri Jul 24, 2009 3:23 pm

E Michael White wrote:I don’t see it as a disadvantage if a players grade increases when he plays zero or very few games; it is a result of early out of date lower rated results dropping out.
It isn't rational to change the grade, you don't have any new information.

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

Re: Monthly Grading Lists

Post by Rob Thompson » Fri Jul 24, 2009 3:54 pm

even having no new information can lead to an improivement if you are losing erroneous information
True glory lies in doing what deserves to be written; in writing what deserves to be read.

Paul McKeown
Posts: 3732
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)

Re: Monthly Grading Lists

Post by Paul McKeown » Fri Jul 24, 2009 4:07 pm

Rob Thompson wrote:even having no new information can lead to an improivement if you are losing erroneous information
Who's to say that it is inaccurate? It happened.

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

Re: Monthly Grading Lists

Post by Alex Holowczak » Fri Jul 24, 2009 4:14 pm

E Michael White wrote: The last example makes me think during the first few years a player plays, games should not go back to 3 years. Maybe that is the new basis for juniors.
It would be interesting to know a few things:
a) What percentage of juniors play 30 games per season? How does this vary with the equivalent percentage for adult players?
b) Do the players who play 30 games per season more accurately reflect their grade than people who have played fewer games?
c) Do juniors have stronger results at the end of the season than they do at the start of a season?

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Monthly Grading Lists

Post by E Michael White » Fri Jul 24, 2009 4:23 pm

Paul McKeown wrote:
E Michael White wrote:I don’t see it as a disadvantage if a players grade increases when he plays zero or very few games; it is a result of early out of date lower rated results dropping out.
It isn't rational to change the grade, you don't have any new information.
It is rational as the same absolute piece of information can have a different value at a different time. This applies to all forms of information evaluation otherwise there wouldnt be a law against insider trading.

In the case of the grading examples quoted you have new information, in that the performances in each of the 3 years from the evaluation point ie the date of the grading list, have all changed.

Paul McKeown
Posts: 3732
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)

Re: Monthly Grading Lists

Post by Paul McKeown » Fri Jul 24, 2009 4:32 pm

E Michael White wrote:It is rational as the same absolute piece of information can have a different value at a different time. This applies to all forms of information evaluation otherwise there wouldnt be a law against insider trading.
"Insider trading" - smells like a strawman - and non sequitur

As for grades based on insufficient information. Well, they simply shouldn't be published. They are unreliable whatever you do with them. If you don't have sufficient information, you don't have sufficient information. Anything else is simply fiddling about or guessing.

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Monthly Grading Lists

Post by E Michael White » Fri Jul 24, 2009 4:37 pm

The value of information for decision making purposes is clearly time dependent. You need more than a few snorted phrases to dispute that.

Paul McKeown
Posts: 3732
Joined: Thu Apr 12, 2007 3:01 pm
Location: Hayes (Middx)

Re: Monthly Grading Lists

Post by Paul McKeown » Fri Jul 24, 2009 4:49 pm

E Michael White wrote:The value of information for decision making purposes is clearly time dependent. You need more than a few snorted phrases to dispute that.
Insider trading is wrong because you are privy to secrets that the general investor doesn't have access to. In that sense you are defrauding investors. It is nothing to do with timed dependency of information in the sense of a grading system, where all information is open to all. So, I'll stand by my "snorted" remarks of "strawman" and "non sequitur".

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

Re: Monthly Grading Lists

Post by Rob Thompson » Fri Jul 24, 2009 5:22 pm

Paul McKeown wrote:
Rob Thompson wrote:even having no new information can lead to an improivement if you are losing erroneous information
Who's to say that it is inaccurate? It happened.
Note the "can". I'm not saying this is true all the time
True glory lies in doing what deserves to be written; in writing what deserves to be read.

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Monthly Grading Lists

Post by E Michael White » Fri Jul 24, 2009 7:42 pm

Paul McKeown wrote:Insider trading is wrong because you are privy to secrets that the general investor doesn't have access to
In nearly all cases, if not actually all, the information becomes available to the public later which shows the time dependency of its value. If you knew about RBS's USA subprime securitised debt two or three years ago it would have been worth something; the same information now is worthless.

Brent Smith
Posts: 36
Joined: Fri Oct 24, 2008 10:58 pm
Location: Reading
Contact:

Re: Monthly Grading Lists

Post by Brent Smith » Fri Jul 24, 2009 9:06 pm

JamesMurphy wrote: What the ECF would need to do to build such a system would probably be to hire an external contractor to build the system for them - if the ECF have that sort of money lying around - you're probably looking at £100k + easily if you want my professional opinion - then sure it's feasible.
I'll do it for £90k :D

Post Reply