PDA

View Full Version : Acf Ratings System/format Poll



UELleida
05-10-2009, 10:31 PM
Which of the following system/formats would you prefer

1) current system of quarterly ratings 2) Live ratings via website 3) Automatic Glicko (e.g. FICS) 4) Automatic Elo 5) Abolish ratings and have a ladder system 6) Other (please explain below)

ER
05-10-2009, 11:05 PM
1) current system of quarterly ratings
Yes, it is effective, inclusive and productive. It gives State rating officers time to proovide accurate results of all tournaments and for ACF ratigs officer to process and calculate them as accurately as possible. Additionally it has this special element of suspense we all enjoy!
2) Live ratings via website
No, although attractive particularly for the young players and for some oldies like me, that option requires continuous involvement of DOPs and State and AFC rating officers! They are volunteers and not all of them (thats for DOPs) have the knowledge required to use online methods to feed results in the system. It also deprives the special element of suspence from the whole thing!
3) Automatic Glicko (e.g. FICS)
No for the same reasons as in 2
4) Automatic Elo
No for the same reasons as in 2 plus I am convinced that our present system works better!
5) Abolish ratings and have a ladder system
No, I think that can only be a part of a Kaitlin's plot to advertise her ladder! :P

Davidflude
06-10-2009, 08:11 AM
1) Tournaments and rating fees submitted directly to Chess Australia.

2) Rate the tournaments as results are received in the present way.

3) Keep up to date ratings on a web site.

4) publish them quarterly as at present.

This will make the ratings more up to date and less bureaucratic.

Ian Rout
06-10-2009, 10:11 AM
2) Live ratings via website
No, although attractive particularly for the young players and for some oldies like me, that option requires continuous involvement of DOPs and State and AFC rating officers! They are volunteers and not all of them (thats for DOPs) have the knowledge required to use online methods to feed results in the system. It also deprives the special element of suspence from the whole thing!
I think there is also a more fundamental difficulty. The "live" list is only live if everything on it is live - if your opponents' ratings are behind schedule then this will affect your live rating. That scenario comes about in two ways; first in the real world results are not going to get loaded immediately, secondly many tournaments are run at one round per week in clubs (unless it is envisaged that results will be loaded after every round?).

For those who do have a burning desire to have their rating regenerated every second there are two things you can do. First, get counselling. When this fails, the second approach is to go to Barry Cox's calculator and estimate your latest rating by entering results for the quarter - this will be at least as accurate and involves no work, except by yourself.

lost
06-10-2009, 11:39 AM
I think convert it to the same as FIDE for the publishing of ACF ratings.

lost

Garvinator
06-10-2009, 11:49 AM
I think convert it to the same as FIDE for the publishing of ACF ratings.
Do you mean using the elo rating system, or do you mean publishing new ratings every two months?

UELleida
07-10-2009, 08:13 AM
1) current system of quarterly ratings
Yes, it is effective, inclusive and productive. It gives State rating officers time to proovide accurate results of all tournaments and for ACF ratigs officer to process and calculate them as accurately as possible. Additionally it has this special element of suspense we all enjoy!

Using a live website system, state ratings officers would only have to check if results were uploaded correctly. While it is true the element of surprise is there for some, they can just ignore the website and follow a published rating !


2) Live ratings via website
No, although attractive particularly for the young players and for some oldies like me, that option requires continuous involvement of DOPs and State and AFC rating officers! They are volunteers and not all of them (thats for DOPs) have the knowledge required to use online methods to feed results in the system. It also deprives the special element of suspence from the whole thing!

The same knowledge of posting to a this chat forum would be required... i.e. basic computer skills which i'm sure these volunteers have if they run tournaments. The simple uploading of a swiss perfect file to ftp site would be all that is necessary , once the website is made.


I think there is also a more fundamental difficulty. The "live" list is only live if everything on it is live - if your opponents' ratings are behind schedule then this will affect your live rating. That scenario comes about in two ways; first in the real world results are not going to get loaded immediately, secondly many tournaments are run at one round per week in clubs (unless it is envisaged that results will be loaded after every round?).

What you are saying is true , however for those who like to be up to date as possible can check their current rating far more accurately. Also the whole point of the suggestion is to automate the ratings process and thus make it systematic. This would eliminate human error, and also put less stress on ratings officers etc.


For those who do have a burning desire to have their rating regenerated every second there are two things you can do. First, get counselling. When this fails, the second approach is to go to Barry Cox's calculator and estimate your latest rating by entering results for the quarter - this will be at least as accurate and involves no work, except by yourself.

Yes some people put far too much emphasis on ratings ! This whole thread came about from an innocent question by maddam12 in the shoutbox which was along the lines of 'why doesn't the ACF put the ratings formula (how they are calculated) up on a website?' Is that too much to ask of the ACF? If this were revealed then people could construct their own websites for live ratings should they desire. The current system could also be kept, if that were the more popular option.

What do you say, Bill ?

Desmond
07-10-2009, 09:08 AM
I'm unsure what you mean by "automatic" in options 3 & 4. Is this the same as "live"?

UELleida
07-10-2009, 11:00 AM
I'm unsure what you mean by "automatic" in options 3 & 4. Is this the same as "live"?

Ok, by automatic I mean it is done like how the chess servers do it. Having a set table of possible rating changes after each game, which is dependant on the difference between the rating of the two players in a game. This means other results in the tournament don't count in the calculation of a player's rating because it is done on the spot, hence automatic.

Desmond
07-10-2009, 11:02 AM
Ok, by automatic I mean it is done like how the chess servers do it. Having a set table of possible rating changes after each game, which is dependant on the difference between the rating of the two players in a game. This means other results in the tournament don't count in the calculation of a player's rating because it is done on the spot, hence automatic.OK thanks. Not sure if I like the idea of seedings changing during an event.

UELleida
07-10-2009, 11:10 AM
Well who said that for the automatic version the seeds would change during the event? At the start of the event the seeds are determined and that can be the set seeding for the tournament. Just the ratings change upon each game.

Bill Gletsos
07-10-2009, 12:10 PM
Ok, by automatic I mean it is done like how the chess servers do it. Having a set table of possible rating changes after each game, which is dependant on the difference between the rating of the two players in a game. This means other results in the tournament don't count in the calculation of a player's rating because it is done on the spot, hence automatic.This isnt going to happen.

Chess servers doing it are fine because they have hundreds if not thousands of games rated daily.
This does not work well with only a few games per day rated.

Also rating events as they are completed online would have significant issues with the use of intermediate ratings, not to mention the effect of playing unrated players.

UELleida
07-10-2009, 12:59 PM
This isnt going to happen.

Chess servers doing it are fine because they have hundreds if not thousands of games rated daily.
This does not work well with only a few games per day rated.

Yes I think so also, but I put this option in the poll in case someone had some further input into it.


Also rating events as they are completed online would have significant issues with the use of intermediate ratings, not to mention the effect of playing unrated players.

Well, to a lot of us, the workings to calculate these intermediate ratings are quite unclear. If the formula used was posted we could analyse it and consider any issues.
The bottom line is that if a process is manual, it can be automated with careful planning. So how about post the formula anyways so that the members can examine it... or is the ACF a 'secret society' as Brian Jones once mentioned ?

Bill Gletsos
07-10-2009, 01:09 PM
Well, to a lot of us, the workings to calculate these intermediate ratings are quite unclear. If the formula used was posted we could analyse it and consider any issues.
The bottom line is that if a process is manual, it can be automated with careful planning. So how about post the formula anyways so that the members can examine it... or is the ACF a 'secret society' as Brian Jones once mentioned ?You knowing the formula is irrelevant as you cannot do any calculations as you do not know many other factors.

UELleida
07-10-2009, 01:21 PM
How about post the other factors also then? :)

With all due respect, if the ratings officer happened to pass away unexpectedly would the ACF have to start from scratch??

Kevin Bonham
07-10-2009, 01:23 PM
With all due respect, if the ratings officer happened to pass away unexpectedly would the ACF have to start from scratch??

No. The ACF has two ratings officers who work in tandem.

UELleida
07-10-2009, 01:30 PM
Ok that's good. However what if those officers were voted out of office?

And is this formula needed to be guarded or something? I don't understand why you can't just post it for people at the club to see.

lost
09-10-2009, 11:54 PM
Do you mean using the elo rating system, or do you mean publishing new ratings every two months?


I mean publishing ratings every two months.

lost

Garvinator
10-10-2009, 12:23 AM
I mean publishing ratings every two months.
I do wonder what Bill thinks about this from a rating accuracy point of view for the ACF system. I can imagine from a workload pov that Bill may not be that keen.

lost
10-10-2009, 12:33 AM
I do wonder what Bill thinks about this from a rating accuracy point of view for the ACF system. I can imagine from a workload pov that Bill may not be that keen.

This is just a thought and a suggestion that the ACF may want to take on board. Could Bill Gletsos let us know of his thoughts as well!

lost

Tony Dowden
15-10-2009, 06:58 PM
Admit that Glicko is inferior and convert to Elo

Garvinator
17-10-2009, 11:00 AM
I do wonder would the ACF rating system be 'better' if we also moved to two month rating periods?

CivicChessMan
02-11-2009, 08:05 AM
:) Had to smile at Tony's comment. Mark Glickman, the man behind Glicko, is on the USCF ratings committee. Does USCF use Glicko? I think not!!

Rincewind
02-11-2009, 08:47 AM
:) Had to smile at Tony's comment. Mark Glickman, the man behind Glicko, is on the USCF ratings committee. Does USCF use Glicko? I think not!!

Your statement doesn't lend any support to Tony's assertion. The USCF rating committee have discussed moving to Glicko but have not as yet. However, federations (especially large ones like the USCF) have a substantial investment in their current systems and the cost of moving as well as organisational intertia can prove to be substantial.

If you look at online chess server (where existing system investment and organisational change management are issues of a lower magnitude) many use Glicko. Even though Glicko is a little more difficult to implement, the ability to capture the reliability of a rating and (in the case of Glicko-2) player volatility, are distinct advantages of Glicko over the Elo system.

Kevin Bonham
02-11-2009, 11:14 AM
USCF ratings have had so many problems over the years that the USCF not using a system is almost an argument in its favour. But in any case, the current USCF system (http://math.bu.edu/people/mg/ratings/rating.system.pdf) already accepts basic principles of Glicko in its use of a sliding k-factor; it is not true ELO.

Bill Gletsos
02-11-2009, 01:36 PM
USCF ratings have had so many problems over the years that the USCF not using a system is almost an argument in its favour. But in any case, the current USCF system (http://math.bu.edu/people/mg/ratings/rating.system.pdf) already accepts basic principles of Glicko in its use of a sliding k-factor; it is not true ELO.It also uses the 'intermediate rating' concept.

Thunderspirit
02-11-2009, 05:18 PM
I do wonder would the ACF rating system be 'better' if we also moved to two month rating periods?


Like the rating officers have nothing better to do? How about we have weekly rating lists? Be reasonable, Garvin...

Spiny Norman
02-11-2009, 05:24 PM
The poll above doesn't allow me to express my preferences properly. I vote primarily for:

current system of quarterly ratings

BUT I also would like to see:

presentation of player ratings via ACF website

I see no reason to provide "live ratings" (whatever that means). Quarterly ought to be quite sufficient for anyone without an unhealthy obsession with ratings. ;)

Carl Gorka
02-11-2009, 08:31 PM
Go check out the British Chess Federation's rating system which I grew up with:lol:

Since a young(ish) age I have had a deep suspicion of rating systems, so the ACF shouldn't take offence if I think their system is bad....I'm not too keen on any. I don't have a preferece but would like to see uniformity so I suppose I'm in favour of ELO just because its th internationally accepted system.

Saying that, I'd be just as happy if ratings suddenly disappeared and I had to think to myself things like, West next round...hmmm toughy, or Hamilton but I'm White, better than being Black I now have a chance, or Hacche, I'll play whatever:lol: This is what I do anyway.....

CivicChessMan
03-11-2009, 01:45 AM
In Arpad Elo's book, The Rating of Chessplayers Past and Present, he writes about the use of a graduated K in section 3.73. He also writes about feedback processes in section 3.75 where intermediate ratings may be used. A rating system is still an ELO rating system if one or both of these are applied.

Bill Gletsos
03-11-2009, 04:16 PM
In Arpad Elo's book, The Rating of Chessplayers Past and Present, he writes about the use of a graduated K in section 3.73.I dont have the book handy at the moment but if I recall correctly Elo wasnt suggesting using a different K for every player but instead suggested using different K's for different categories of player where those categories could be based on their rating or where they were within their chess career.

He also writes about feedback processes in section 3.75 where intermediate ratings may be used.Again if I recall correctly Elo wasnt suggesting intermediate ratings be used for all player calculations but only for exceptional performers so this isnt really the same as the USCF or ACF system implementation of intermediate ratings.

A rating system is still an ELO rating system if one or both of these are applied.Not really.

Bill Gletsos
03-11-2009, 04:26 PM
I do wonder would the ACF rating system be 'better' if we also moved to two month rating periods?Possible issues would be:
1) For 'unrated players' who are new to the rating list it could result in very weak players not getting a rating as their chances of scoring 0% in a rating period would be increased due to possibly having a smaller total number of games in the period.
2) Since improving players may not be picked up as quickly due to less games in the period and hence their 'intermediate ratings' being lower what impact this would have on established player ratings.
3) What impact is there to predictive accuracy.

As such I would want to run tests on existing historical data to see how using a 2 month rating period would affect ratings.

Mischa
03-11-2009, 04:58 PM
could players under a certain rating be excluded perhaps?

Bill Gletsos
03-11-2009, 09:23 PM
could players under a certain rating be excluded perhaps?Be excluded from what?

Garvinator
03-11-2009, 09:42 PM
As such I would want to run tests on existing historical data to see how using a 2 month rating period would affect ratings.
Of course I would expect it to be throughly tested before even being trialled. As for the points you raise:

1) For 'unrated players' who are new to the rating list it could result in very weak players not getting a rating as their chances of scoring 0% in a rating period would be increased due to possibly having a smaller total number of games in the period.Would this be such a bad result? It would result in their first rating being produced from more first set of games.

2) Since improving players may not be picked up as quickly due to less games in the period and hence their 'intermediate ratings' being lower what impact this would have on established player ratings.If ratings were produced every two months, the need for use of 'intermediate ratings' may not be required, or at least reduced in number by quite a margin. I would think that this is a good thing, if the use of an intermediate rating can be reduced.

3) What impact is there to predictive accuracy.
And this would be the main issue.

A side issue is that when a one round per week event is seven or nine rounds, it may just miss the cut off, instead of making the cutoff under the current 13 week approx schedule. This could lead clubs/organisers to reduce the number of rounds for a tournament if they wanted to match the reporting times for rating periods.

Bill Gletsos
03-11-2009, 10:19 PM
Would this be such a bad result? It would result in their first rating being produced from more first set of games.No it wouldnt as testing has shown that carrying over such 0% results to future rating periods generally results in lower initial ratings and reduces their predictive accuracy.

If ratings were produced every two months, the need for use of 'intermediate ratings' may not be required, or at least reduced in number by quite a margin. I would think that this is a good thing, if the use of an intermediate rating can be reduced.Actually the use of 'intermediate ratings' improves predictive accuracy across the board.

And this would be the main issue.As it should be.

A side issue is that when a one round per week event is seven or nine rounds, it may just miss the cut off, instead of making the cutoff under the current 13 week approx schedule. This could lead clubs/organisers to reduce the number of rounds for a tournament if they wanted to match the reporting times for rating periods.I am not so sure that organisers pick their finishing dates to match up with the ratings periods so this probably is not an issue.