| Author | Message | ||
     Frances Collings Username: Victory Registered: 11-2009 |
Wish we could keep it simple all too much for us oldies to understand .....co efficients! | ||
| Adrian MacLeod Username: Ademac Registered: 03-2015 |
Thanks Mike. It's useful to have a full explanation. I was getting confused by the standardised scores in last year's results but it's much clearer now. I don't think either Jenny or myself intended to criticise or argue the numbers -- just understand where they come from. Thanks for taking the time to explain. | ||
| Jenny Moon Username: Jennym Registered: 01-2012 |
Mike, thanks for that. I was just trying to understand and your explanation makes a lot of sense. Much appreciated. | ||
| Mike Watts Username: Mikew Registered: 10-2006 |
The idea of showing a "percent" on the results is to show a simple number to try to make qualification more transparent, giving a way of comparing across events, regions and classes. I hope it will all be clear - and as simple as we can make it - when we display the points leagues and as the season results build up. Please give it a chance. The class coefficients are the same for all the events. I have listed them here. They are derived from Keysoe scores where possible, since they are to help determine Keysoe qualifiers, i.e they reflect what the top slice/likely qualifying competitors can achieve in a given class. They don't affect event rosettes. We use the scores for those as before. There is no perfect way to calculate something like this - we've chosen something which is simple, intended to be understandable. It's 'just' another number to give a relative measure. If it does not seem to be giving good results mid season or so, ICD UK can always review. Since ICD reserves the right to invite competitors to Keysoe, there is already a manual mechanism for getting the most appropriate competitors to Keysoe. However, publishing these numbers as we go is intended to give a more transparent mechanism ... let's see how it looks when it has run for a good number of events. We could argue forever about what scheme to use. We've looked at a stack of schemes and you'll see another one on previous results, a 'standardised' score [calculated from median scores] as well as the 'percentage'. I looked at fitting curves [normal or other distribution], ignoring outliers, weighting top few, calibrating by finding an individual at an event with a 'known' record [this is used in setting the French Equivalence ski test]... Each scheme has its merits, but the bottom line is that, especially on smaller events, the number of figures is too low for any method, especially any complex method, to be reliable in all cases. The more complicated the algorithm, the more chance of odd results in unusual cases and the less transparent. The typical score at any one event will depend on the length of the obstacles, going/environment, difficulty of the cones course and the judges' outlook - that's why some normalisation of the numbers is needed. Of course sometimes obstacle designs will favour some types of turnout and other factors can also make it hard to compare classes, but I think this will come out OK over a number of events and over time, judging by previous year's figures. I know you can construct unusual cases where the results may not be how you'd want even with this simple formula [one very exceptional person in an area etc]- however, I think it's actually quite a robust algorithm and gives quite fair results - even where we do have exceptional scores. We wanted to do something, so new areas could still be added and small or new areas compete on an equal basis without messing up the Keysoe qualification nor making it entirely committee/manual - but trying to handle the case where a person would qualify regardless of standard just because they are the only one in their class at an area. They need a way to qualify, but not just automatic as they're the only one. Coefficients and formula: Your normalised score is your score times your class coefficient. We used Keysoe class where we could work out what class that was from the Notes. (Pseudo) Percent score = 100 * (best normalised score)/(your normalised score) Novice Single Pony 0.95 Open Single Pony 1 Novice Single Horse 0.92 Open Single Horse 1 Junior 0.95 Club 0.9 Teams 0.68 Intermediate Pony 0.98 Intermediate Horse 0.96 Extra Young Juniors 0.8 Young Juniors 0.9 Senior Juniors 1 Veterans 0.9 Pony Pairs 0.94 Horse Pairs 0.73 Small Pony 0.9 Small Pony Pairs 0.9 We could argue numbers until the cows come var/www - let's not - everyone may have their own view, and it changes over the years.... | ||
| Adrian MacLeod Username: Ademac Registered: 03-2015 |
I think the coefficients need to be more transparent. They seem to change from event to event. Am I right? Does this mean the number of entries in the class changes the coefficient? I have only looked at Merrist Wood percentages last year but there seems to be a different coefficient applied depending on whether I was down as open or small pony (it was all the same pony just different paperwork) I don't see the sense of this as small ponies are invited rather than qualify for the finals... or has this also changed? Also if my percentage puts me first but my penalties place me third can I have a percentage of a red rosette?  | ||
| Mike Watts Username: Mikew Registered: 10-2006 |
You can see how the percentage scores look on previous years scores. Go to https://www.indoordriving.co.uk/results.php click 'show totals', select the results of an event you are interested in and click the Percent tab. Where it refers to a competitor's 'best percent score ... in that class and region' it means the highest percent they get that season, in that class, in that region. We plan to add a column on the points leagues showing each competitor's best percent score so far in that region and class for the season, so it's easy to see. Still working on how to make the points league pages look as clear as possible - we might display an additional 'local points league' tab showing the results in the 'local classes' - if that doesn't look too confusing, so Intermediate and Open would be mixed in that. ... but championship qualifying points league should separate the championship classes. So as you say and 10.1 says, first Open Pony would get 10, second Open Pony 8 etc, while first Intermediate would get 10, second 8 etc. The coefficients come from analysing previous years' scores, to try to equalise the difficulty across classes. I hope it will all be clear - and as simple as we can make it - when we display the points leagues and as the season results build up. | ||
| Jenny Moon Username: Jennym Registered: 01-2012 |
Please could somebody please explain "the best percent score" and where the "coefficients" that are applied come from. I understand what is trying to be achieved as an end result but don't understand the method. Also, how are points awarded in an Open class, where there may well be Intermediate drivers.... If,say, an intermediate driver is placed second within the competition, are they awarded 10 points as first Intermediate, and then by definition the next open driver given 8 points as second in the open class. |