Some random thoughts...

Stuart, do you reckon there\'s any way to \'open source\' those php caluclations so that people could implement different options if they had an idea rather than having to copy stuff into excel? (or writing some awful scraping script as I did before!)

My feeling is that, as Paul points out, the league suffers a lot from who shows up on a given day. Depending on which event you go, to the results of others and who is there can have a massive effect. I also like the idea of assigning classes to courses for the purpose of prizes, with one caveat, there should be an avenue for beginners to work their way up while competing for prizes the whole time.

I toyed before with the idea of a national ranking, which would address some of the \"strength of field\" issues, but that\'s quite a heavyweight solution and would only reward people with a good ranking in the first place.

Having a glance at this document from 2004 when this current system became all the rage:  https://www.orienteering.ie/results/leagues/resources/IOLeagueScoringSystem.pdf
One sentence about the number of competitors jumped out: (paraphrsaing due to weird copying problems) \"With fewer than 10 competitors the standard deviation is taken as 16% of the mean\"

Another system that might be worth considering is the DVOA/USOF/Attackpoint ranking system or some variation. This uses an iterative normalisation approach, giving the top 3 runners an average score of 100. The 2013 section on this page ( http://www.dvoa.org/rank/calcrank2013.php ) gives an overview. It address some of the strength of field issues by recalculating all races being considered. There are for sure some disadvantages (such as all the scores being changeable until after the final race!!) but it would certainly be interesting to apply it to the league and see what happens!

Hi all,

Firstly Ruairi, yes of course - if you could write a PHP function which takes in a PHP array of times and outputs an array of points then it\'d be very easily to plug in. I can give you the exact array definition. That said, I\'ve provided the full data extract (anonymised) in csv format above for the 2019 Leinster League. I think it would be far better to analyse this first and agree changes, if necessary, rather than to modify the live system.

Paul, just to make a minor point, it\'s best 6 of 13 races to score. This means that the range is actually 73, not 195. At 6% of the average score, I don\'t feel that this indicates a significant difference between the events (though we could probably do a proper Chi-Sq test or similar to know for sure). On top of this, the competitor was first female in only half her scoring events so we\'d expect her points to vary. That said, you are perfectly right in that the current system is opaque and contingent on who turns up for each event.

I think we should analyse the data to see how different methods compare before proposing a change. Personally, I don\'t feel we need to get too carried away but I\'m happy to be corrected!

Stuart

How did you arrive at that value of 73 Stuart?

Paul

Quote from psmythirl on January 3, 2020, 11:31
How did you arrive at that value of 73 Stuart?

Paul

Hi Paul. Of the six events to score, it\'s just the highest less the lowest: 1214 - 1141

Yes, I know that only six events count. I gave a sample with 12 data points. These are spread across 3 or 4 different competitors. The range for a sample is highest - lowest which is 1312-1117 = 195. Your range is for a sample of 6 data values. How are you deciding which six values to use?

The point I\'m trying to make is that I could have won one of these events last year and got 1312 points and another competitor could have won another event and got 1117 points. That\'s a difference of 195 points awarded for similar results.

Let\'s take the best case scenario. Suppose the points I gave are for two competitors and each won exactly six events. Since the points are varying from event to event the fairest result will occur if one competitor wins the event with the highest points, the other wins the event with the next highest points, the first with the next highest and so on. This gives us

Competitor A
1312
1214
1210
1174
1141
1122
Total = 7173

Competitor B
1250
1212
1191
1152
1134
1117
Total = 7056

That\'s a difference of 117 points resulting in a league victory for competitor A. There are two points to be made about this.

First, this is the best case scenario and in general the chance of this scenario occurring would be low, so the actual difference will be larger. In the worst case scenario, where one competitor is unlucky enough to get all the low scores, the difference is 549 points. In any given case the difference will be somewhere between the two.

Second, this is entirely an artefact of the scoring system allowing variability in the maximum points available for an event. If the system always assigns the same maximum number of points this issue disappears.