|
Joined on 20 Ιαν 2003
Total posts: 28
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 4:38 πμ
Zbysek (Admin),
I must applaud your efforts at trying to introduce a ‘worldwide ranking system’ but I think at best it will never be anymore than of academic interest to those who visit this excellent web site. The ‘real’ ranking systems will continue to be the IDSF ranking system and those of national governing bides e.g. the EADA charts in England. From what I can see the main purpose the web site’s ranking system is to encourage people (who are interested in it) to visit your web site and in that respect I applaud you for introducing it. However visitors should ‘take it with a large pinch of salt’.
On the subject of how you obtain results/maintain information on whether or not couples have split or not, I can tell you that from personal experience there is only one way in which these can be made to work. In the former case (obtaining results) it is high time that a standard system for inter-changing results was set up between promoters/scrutineers and webmasters. As some will know I tried to do this within EADA by getting EADA to standardise on the use of Scrutelle for all EADA NSCs. While Scrutelle can not be described as a ‘standard’ for competition results inter-change it does represent an example of a piece of software which supports the ‘direct’ publishing of results on the Internet (as do ProDance and Scrutiny etc also). This attempt at ‘standardisation’ was blocked by EADA’s new chairman and was one of the many reason’s why I resigned as EADA’s webmaster. I had hoped that by standardising on Scrutelle for all EADA NSCs that it would then quickly be accepted as the de-facto standard for all dancesport competitions in the UK. I hadn’t however bargained on the ‘lack of vision’ that would be displayed by EADA’s new chairman. Incidentally gathering of ‘digital’ results in this way would very quickly have built up a large database of results, which could have been analysed for judging bias. The absence of such a large volume of data precludes the possibility of any analysis.
On the latter point i.e. maintaining records of who has or hasn’t split, I was far more successful. After taking over as EADA webmaster one of the first things I did was to set up the web site so that it was fully database driven and was more dynamic. By that I mean content (e.g. news items etc) could be added to and ‘maintained’ via the web site itself. To this end, I set up login IDs and passwords for all registered EADA members, which enabled them to log into the web site and update their personal, and partnership details. This meant that they could specify by themselves (following an initial authentication process which was fully logged) whether or not their partnership had split and if they had formed a new partner and if so who with. On doing this their partnership was marked a ‘split’ within the database and any reference to their partnership e.g. in the EADA charts, photo gallery etc) was automatically no longer displayed on the web site. At the same time an email was sent both to myself and EADA membership secretary (so that he could update the ‘off-line’ membership database). In this way it was possible to keep the web site and in particular the EADA charts ‘bang’ up to date (as required by EADA’s own rulebook). By the way for those ‘morons’ out their who post adverts for members of partnerships which have not split, should be aware that ALL postings on the EADA web site are logged as I’m sure is also the case on this discussion forum. I had intended to use this ‘authentication’ system to provide a facility for registered EADA members to ‘pre-enter’ EADA NSCs. This implementation of this facility was also blocked by EADA’s new chairman. It was also my suggestion that this system could also be used to enable EADA members to vote ‘electronically’ in Council elections and so avoid the need to have to attend the AGM in order to vote EADA Council representatives. Indeed I wanted to take this one step further and provide a means by which EADA members could vote ‘directly’ on key EADA Council officer posts e.g. Chairman, Vice-chairman, and Treasurer etc. This attempt at making EADA a far wider enfranchised, more democratic organisation was also blocked.
I apologise if I am being 'disruptive' (as I've been referred to in certain emails between EADA Council members) in posting these details but I feel that all EADA members have a right to know what has happened in the past and is likely (or should that be not likely) to happen in the future.
Regards
Kevin
|
|
Joined on 14 Φεβ 2001
Total posts: 1.974
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 6:57 πμ
Kevin - I appreciate your problems, but could you move these issues to the English Dancesport area?
As for our rating system. We have stated quite clearly that it is not ranking, but rating. It's like in chess - nobody is saying 'i'm 5237th in ranking' but they say - 'my current rating is around 1600'. And also we have said that it is not supposed to replace any ranking system. As for IDSF ranking - id does not include Professionals (by definition) and does not include couple who do not travel around. It also does not take into consideration many best comps in the world. Local (like EADA) systems do not care for lower level dancers (Intermediate, Pre-Champs etc in England and classes B and below in other countries)
|
|
Joined on 12 Αυγ 2003
Total posts: 84
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 7:05 πμ
Ooops, I just noticed that my posting further up didn't work...
What I was trying to say was in response to the requests of an indication of whether people did better or worse than expected:
The ELO system is designed to do exactly that. The mathematical basis for it is that it calculates your expected result based on your rating and that of the other competitors. If your actual result is better, your score increases, if it is worse, it decreases.
The reason that does not work at the moment is that the system has not reached an equilibrium. The differences are not big enough. Ie a difference of 200 points is indicative of it being very likely that one person beats the other. Yet in the pro modern for example, the top 6 are separated by a mere 60 points, in terms of ELO this means that *any* of them have a good chance of winning, which is clearly not true.
This arises as the system was not correctly primed, it's still in the process of equilibrating. That's why all finalists in big comps always improve their scores at the moment. If the ratings had sufficiently diverged, their scores would not change much, unless they beat someone they were not expected to....
Admin says that the priming didn't produce sensible results, but I thought the last time i saw results from the prototype, they were actually much more sensible...
|
|
Joined on 13 Ιαν 2005
Total posts: 468
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 7:25 πμ
Regarding priming:
the problem now is that if it takes a long time for the equilibrium to be reached, by that time the couple will have mived on i.e. either they are already professionals or are no longer at the level that the quilibrium calibrates them to be. I mean that couples do improve all the time normally - so the question is if anybody actually OVERTAKES someone or not. This is what the rating should reflect in my opinion.
For example, if You take a look at Soale's rating, it has been steadily improving for 7 years (those of us lucky to see Admin's test graphical chart page). Although I agree that his dancing have been continuously improving during this time, his POSITION and expected position, is definately not this obvious - it should at least have been constant in the last year - although still at they should be the highest rated couple, but since it is now "form result" that they win the amateur comps, I am not sure that their "rating" should continue to improve all the time.
Even if we expect that this is so because of the calculations, there should then be a way (IMHO) to display the result vs. expected result differences, as this is the real interest. So either prime the calculation so the change in rating is actually highly corresponding with the difference of real vs. expected result, OR display this simply based on the order in the ranking before the comp vs the result of the comp.
Just my 2 cents.
|
|
Joined on 14 Φεβ 2001
Total posts: 1.974
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 7:54 πμ
As for being lucky :) to see graphs - they are now available to all (on couple result page and couple profile page you can click apropriate link). If you do couple comparison and click on compare by rating - you will see that they DO overtake each other sometimes - compare for example Soale and Bizokas.
I think there are several reasons for this behaviour - priming is not the most importan in my opinion. First reason is incomplete data - if we have only final for a competition it works badly on 5th and 6th couple. Second reason (and probably the most important) are coefficients in calculations (ELO function and cut-off point for selection of couples for competition). I would really appreciate discussion of these coefficients with some mathematical minds :)
As for priming - first rating is given as average between people who won with this couple and people who lost to them. Only if we have nobody who won or nobody who lost - we assign default value.
I think also that initial value and further distribution of rating values is tightly connected to the coefficients mentioned above.
|
|
Joined on 14 Ιαν 2005
Total posts: 5
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 9:21 πμ
quote:
Originally posted by Elaine
Mo is right in that some of the couples which she mentions have changed status but not all of them. For example Jack Beale is not yet 16 and his partner Joanne is younger than him so I don't think they will have gone up to youth yet. So of course any information for changes really has to come from the dancers themselves and not via third parties. If admin accepted information from anyone there would be a great deal of room for error and also for mischief making by the odd one or two morons who would think it was funny to invent a split partnership or two.
As a third party, in answer to Mo and you, I can certainly say that most of the couples were at Champions (eg Karl and Drew, Nikita with a new partner), have moved to youth (David and Lucy danced youth at Champions), have split (Leon split with Rebecca after the nationals, Ashley and Lucy split about 2 years ago, and Ashley now dances youth with Kerri-Ann Hallbrooke(?spelling)). The rest looks OK to me!
|
|
Joined on 13 Ιαν 2005
Total posts: 468
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 12:20 μμ
Zbyszek,
I think the rating, the top chart, the comparison was very well implemented with the overall site. I am continuously being impressed by how well You manage to implement the power of database system into a usable interface ![Smile [:)]](/emoticons/emotion-1.gif)
My only suggestion remains the expected result vs real result comparison. Am I alone?
|
|
Joined on 12 Αυγ 2003
Total posts: 84
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 5:20 μμ
I don't think priming of inidividuals is the problem, the average of the round they were knocked out seems a sensible and simple approach. I think the problem is priming of the whole system. Everyone starting at the same point 5 years ago is just not enough time for the scores to diverge sufficiently...
Which, in my opinion, is the reason that some scores are still continually increasing...
as for cut-offs etc. Is there a page with the cut-offs you currently use displayed on it?
cheers
Robin
|
|
Joined on 14 Φεβ 2001
Total posts: 1.974
|
Rating system
Παρασκευή, 14 Ιανουαρίου 2005 6:08 μμ
Unfortunately there is no such page.
At the moment couples which have difference over 400 in rating are not compared.
Coefficient C in 1/(1+10^(C*dR)) is set to -400, while multiplier is set to 32,000/R, where R is an old rating
I would rather discuss it over let say MSN (or chat) one day - something like a brain storm, because it would be much easier
|
|
Joined on 08 Ιουλ 2004
Total posts: 8
|
Rating system
Σάββατο, 15 Ιανουαρίου 2005 1:26 μμ
I can confirm that Karl Drew and Joanne Barnett move to Youth on the 29th of January (Jo's 16th birhday).
Karls dad
|