|
Joined on 14 helmi 2001
Total posts: 1 975
|
Couple ranking system
10. maaliskuuta 2004 19:19
First version did not produce very good results :( We are looking for flaws in the procedure.
Regarding sending results - ideal would be file in the following structure:
Couple1 Name;Status;Country;Round;Place#
Couple2 Name;Status;Country;Round;Place#
etc.
For example:
John Smith & Anna Brown;S;EN;48;35#
Where John Smith & Anna Brown is the name as displayed on our pages, S stands for senior (A for Amateur, Y=Youth, J=Junior, N=Juvenile, O=Over 50, P=Professional), EN is code for England, 35 is real placing (or 0 if unknown)
48 - needs further explanation - in the final positions is 1 to 8 in semifinal it's 12 in quarter it's 24 in previous round is 48,96,192,384,768 etc, even if number of couples in the round was smaller.
We are also able to allow remote access to our database via Internet under some conditions.
|
|
Joined on 03 maalis 2004
Total posts: 5
|
Couple ranking system
10. maaliskuuta 2004 21:35
If couples compete on the Sunday Circuit, they need to be registered with EADA (this is the case in the UK anyway). Presumably, a similar rule applies elsewhere. Why not ask for the couples' registration numbers instead of their names for each event (after their first score, it might save time and improve accuracy) ?
It might be an idea, also, to check with existing ranking bodies, out of courtesy, before going ahead.
|
|
Joined on 12 elo 2003
Total posts: 84
|
Couple ranking system
10. maaliskuuta 2004 22:02
What sort of results did it produce?
I could imagine with the way dancing results work (e.g. results are very similar from comp to comp, Warren and Kristy always beat all other UK amateurs) you could get run-away ratings, so maybe you need to define a cut-off (if the difference is more than 200 points the comparison does not add extra points).
Also the inclusion of very large competitions might skew results as they suddenly include 100s of comparisons, especially when you start everyone off with the same score. To alleviate this you could scale the final adjustment by 1/sqrt(n) where n is the number of entries. This would mean that beating 16 people is worth only twice as much as beating 4 people rather than 16 times as much...
If you could post some results sets that came out, maybe Alun or I could have a look as well and see what went wrong. I can also ask some statistician friends of mine for ideas...
As for results sets, i really think that once the system works you can make it a condition for competition organisers that to be included in your ranking they must submit results in a machine readable form. Alternatively you could offer to type them up if they paid for it. Once a rating system is established it will be of immense benefit for any organiser to be included (see the current wrangling over who gets to organise the ranking comps), so they might be prpared to spend an hour typing up results in return for the free publicity.
I look forward to seeing some attempts at rankings!
Robin
|
|
Joined on 14 helmi 2001
Total posts: 1 975
|
Couple ranking system
11. maaliskuuta 2004 0:00
I am afraid you are living on different world :) Only about 1% of our results is from organisers!
They are NOT willing to give results - usual excuse is that they are too busy. As best they MAY give printed results and only if they are in the good mood :)
I also had dreams like you have at the moment, but I am cured now :) EADA is trying to get ranking competitions results from organisers and is failing! I am not joking!
Forget about IDs, codes and electronic form in England! We get them sometimes from US, Canada, Russia, Poland, Iceland and Finland, but not from England.
Back to ranking - working with organisations is out of the question - they just do not wish to talk.
We are not part of anything and out of all loops, so we can do our ranking.
I have modified code (and fixed bugs :)) and now works a bit better, but still not perfect.
I just wonder how can I post you results? Code is written in T-SQL (for SQL Server), hence is far from being elegant. I have splitted ranking into Pro, Amateur, Youth, Junior, Juveniles, Senior and Over 50 and additionally Ballroom, Latin and 10 dance. For example the best Pro Latin are:
4134,Slavik Kryklyvyy & Karina Smirnoff
4073,Bryan Watson & Carmen Vincelj
4012,Paul Killick & Hanna Karttunen
3913,Dmitri Timokhin & Anna Bezikova
3872,Michael Wentink & Beata Onefater
3819,Andrej Skufca & Katarina Venturini
3755,Michal Malitowski & Joanna Leunis
and Slavik being on top is rather disputable, as he never won with Bryan
|
|
Joined on 12 elo 2003
Total posts: 84
|
Couple ranking system
11. maaliskuuta 2004 9:27
Those results are interesting... wonder how that happens. Do Slavik and Karina just enter many more events? I guess there have always been ranking systems where the undisputed number one is not placed 1st.. I guess you could try introducing a steeper gradient in the probability (e.g. 200 rather than 400) or maybe defining a cut-off whereby the probability is taken as 1 (or 0) if the difference is more than 100 points or so...
Anyway, at least it does produce some results, and requires no extra work, so maybe with some tweaking of the formula it could be a good start?
As for dance teachers/comp organisers... I was afraid this was the way it was. There certainly seems to be an "old guard" that wants to do things how they've been done centuries ago. I suppose you could try approaching scrutineers themselves, eg Estelle Grasby or Jacky Conway (spelling?). A brilliant solution would be to produce a new scrutineering software (not hard to make it better and cheaper than the current ones), that would not only produce printed results and an html version but also a version that is automatically submissable to your database! Incidentally that could also be a new revenue stream...
I heard you were in contact with one of the university student comp organisers... i can almost guarantee they will provide you with results! Students get very stroppy if they don't see not only results but also judges marks up on the web within a week or so! But you're right, it's a differnt world!
Robin
|
|
Joined on 12 elo 2003
Total posts: 84
|
Couple ranking system
11. maaliskuuta 2004 12:10
just 3 more thoughts:
1) If you have the algorithm set up, why don't you try running it through the uk adult results (including all levels). These are much more variable and will probably give a better indication as to whether the system copes with that! For the top 6 pros the order never really changes so a ranking system is a bit unnecessary...
2) If a new dancer is introduced to the system they should not all start with the same score. This will seriously bias the event this happens. I would suggest leaving a dancer out of the calculation in the first event they appear in and give them an initial score based on how they did in the event. (ie give them a rating such that their expected score for the event is the same as the score they received).
3) There should be a penalty for not entering any events for some time to stop someone sitting on a high rating forever... maybe take away 20 points if someone has no rated event for 6 months... or something like that...
Anyway, great work i say!
Robin
|
|
Joined on 14 helmi 2001
Total posts: 1 975
|
Couple ranking system
11. maaliskuuta 2004 12:21
Funny enough after applying this routine 5 times (i.e. changing starting point from 'everybody 1500' to 'previous result') we got quite good results. We still have to flatten big comps as they change too much, but I hope that VERY soon we will publish experimental rating. That will cause a stir! :)
As for scrutineer program - it's not a good revenue road - we have tried it. First of all, not many scrutinners in the country, secondly most of them do not wish to change what they use and some of them are still using DOS based program. Also, program would have to work in all even most exotic competitions (like disco, medal etc) and cater for all possible cases, like repecharge (spelling?), rule 10, 11 etc. Possible to be written of course, but would not be popular as we would step on some people toes.
It would be easier to write conversion program which would convert data from existing programs to our format. We have already done it for one of the scrutinners programs, but ... scrutinners do not really have time to send us results and we would have to pay them for their service, and ... we are back to the point 3 in one of my previous postings. Publishing results on the net is also a competition for them - some of them are selling printed copy of results.
Publishing marks is beyond our scope at them moment. Too much time and too few results.
Robin, AlunB, Janine - when I will be doing rating page - I would like to mention you as people who encouraged me to write this and helped me with finding some solution (even if not perfect :)).
Can I publish your names? Or just nick names you are using here?
|
|
Joined on 19 marras 2002
Total posts: 76
|
Couple ranking system
11. maaliskuuta 2004 13:16
A few quick points:
1. I'm happy to help with typing in both results and event details - I already try to update information on universitydancesport.com so keeping your site up to date with sunday circuit shouldn't be a major problem.
2. Estelle Grasby has apparently updated Scrutelle, her software, so that she can put results online the day after the event. Have a look at www.scrutelle.com as it has some results from previous competitions already. She has been willing to work with the university circuit to try out a completely different scoring system in our team match so I'm sure she is flexible.
3. I think a revenue stream could come from the development of a printed magazine version of Dancesportinfo.net, with some extra articles etc. Again, I'd be happy to help out with that if I could. To be honest it wouldn't be hard to keep it more up to date than Dance News...
|
|
Joined on 12 elo 2003
Total posts: 84
|
Couple ranking system
11. maaliskuuta 2004 13:44
Excellent news! Sounds great! I expected that you might have to scale down the results from larger comps (see my post above), and that it takes a few iterations to settle into a sensible order... This could be especially true with groups with little overlap (e.g. intermediates in russia and uk).
I would of course be honoured to have my name (Robin Schlinkert) mentioned in connection with this. Be sure to let us know when you publish the rankings so we can help spread the news!
Anyway, well done! I'm greatly looking forward to it and i think a lot of people will be interested. Once dancers start asking organisers before the event whether it's included in your ranking, who knows, the world might change more quickly than you expect...
Robin
|
|
Joined on 12 marras 2003
Total posts: 41
|
Couple ranking system
11. maaliskuuta 2004 15:05
My goodness, what have we started?! Well, if Robin is willing to go public about his involvement I am also happy to be mentioned. We are going to have to be very proactive to get this started as I think there could well be resistance to anything new, particularly since it requires effort to run! I am hopeful that we will be able to win over organisers as it increases in popularity. Zbyszek, let me know what you would like me to do next and I'll get on with it!
I like Dave's idea about the magazine version which may fit in with Virginia's suggestion of a few days ago. However we shouldn't under-estimate the achievements of Dance News. Apart from their own production schedule they are also prey to the other priorities of their contributors who tend to have heavy adjudicating commitments. I think it may be worth exploring the idea, but it would be a huge task - I know, since I was formerly a PR Manager involved in just this type of thing! If anyone wants to follow this up perhaps they would like to e-mail me.
Janine Denman
|