1580 Ranking Points!!!

[mharky]

Good to see such constructive comments - they really help

Sorry, should I have come up with a solution to the problem? Sadly, I'm not a consulting mathematician, so other than telling you the system is flawed, there isn't much I can do to be constructive.

Graeme, on the other hand, does know what he's talking about, so is probably worth listening too.

[AAH]

If someone suddenly gets thousands more points than they have previously managed just leave them out of the calculations until they are validated. Or is that just too simplistic?...

There might be a lot to be said for automatic removal of particularly drastic outliers e.g. fit straight line, establish if anyone is more than x% faster or slower than predicted from that line, bin their score, repeat.

Alternatively, manual removal of outliers by either the controller of the event or someone at BOF who could cast a quick eye over a graph. Sure, it’s best to have an automatic algorithm without human input but we are dealing with small samples. The algorithm works for the IOF ranking because there are large fields of very consistent orienteers. All that was needed here is for one competitor’s time to be disregarded.

[graeme]

If by this you mean that "Wesley Musall"'s points shouldn't have been included in the points calculation, then you're not correct as it's the last 10% of the ranked runners measured by the time they took who are ignored, not those with the lowest points.

Ah, I see. In that case the coding may be OK and the algorithm at fault.

Treating a very slow person who runs exactly as expected (i.e. very slowly) as an outlier is a statistical nonsense, You should be discarding people who are a lot slower than expected, because they are the ones likely to be injured, pregnant, misidentified etc.

yllacitamotua siht seod RQI ,esruoc fO

[pete.owens]

Another "guest" runner was Nigel Beasant on the blue course - I can provide an alibi.

One way of filtering out anomalous input would be for the algorithm to flag any runner whose ranking points for that event are a long way out of line with their points at other events. The event could be put through the system for a second pass with that runners time not used in the calculation of mean and spread.

[madmike]

Graeme, on the other hand, does know what he's talking about, so is probably worth listening too.

Mharky,
first time I've seen you admit you don't know what your talking about

[Scott]

What can BO do to prevent it? Any cross checking of name and age against BOF number when results are submitted will throw out lots of spelling errors and typos when there is no problem with the BOF number, and give the results official a pile of extra hassle.

I have no idea what the backend to the BOF rankings is, but if it's something standardish I would guess that there's probably an implementation of Levenshtein matching available that could be used to flag up genuine mistakes in name without hassling organisers with minor typos.

[graeme]

Looked at very broadly, the ranking list is based on Normal statistics with a standard deviation of 200. So, 1580 actually has a well defined meaning: 2.9 sd above the mean. That is a run in the top 0.2%.

With 5000 ranked competitors, this is the type of score a top-10* athlete should routinely be getting. Indeed, since
Appendix K2.2.4 The current scores of all ranked runners are rebased after each event to ensure that the mean current score of all ranked runners is 1000 and the standard deviation of their current scores is 200.
This should be happening automatically.

* (i.e. 10=5000x0.002)

[martin]

.... and a bottom-10 athlete should be routinely getting a score of 400 plus.

Just goes to show that the distribution is not 'Normal'

[CS99er]

I think we have to accept that how ever sophisticated the system is there will always be anomalies that sneak through and the easiest way of dealing with it is to avoid using the most extreme scores (i.e. best 6) in the either the ranking list or in the underlying ranking calculations.

[EddieH]

That is thinking outside the box. I like it

[mharky]

Not that much outside the box, artistic olympic events (diving, gymnastics) discard your highest and lowest scores... a very good idea come to think of it.

Many people are a crazy high score that they got from some local-hero event. Chopping your highest few score will get rid of that, and perhaps, show who is more consistently performing at a top level.

Although, not sure if that would get rid of Dave Rollins, other than him the top 20 is pretty much who you'd expect.

[Big Jon]

Interesting that a lot of highest scores for individuals are from sprint races - are these better for getting points from or is there another explanation, like fewer courses therefore greater spreads of times?

[EddieH]

Well looking at WOC results there is little doubt that British Elite are generally better at sprint than middle or long - so perhaps it's a true reflection.

[jac]

I think we have to accept that how ever sophisticated the system is there will always be anomalies that sneak through and the easiest way of dealing with it is to avoid using the most extreme scores (i.e. best 6) in the either the ranking list or in the underlying ranking calculations.

Is there an echo in here?

http://forum.nopesport.com/viewtopic.php?f=1&t=10392&p=113562#p113562

The Swedish elite ranking lists used to do this.

Regards

jac