Results and Rankings.

[martin]

I've just seen the points for the Aire event at Bramham.
Are they a statistical anomaly or a miscalculation?
I like getting lots of points, but ...

Ranked runners getting more than 1000 points ...

Brown: Only 3 out of 22
Blue: Only 3 out of 46
Green: A massive 28 out of 62 !

[Neil M40]

Hmm... If I had done the same time on the Green that I did on the Blue, I would have got the same number of points.

If Andy Thorpe had done the same time on the Green that he did on the Brown, he would have got the same number of points.

In fact wherever you look in the Blue or Brown results, the same time on Green would give you the same number of points.

That's not reasonable.

[greywolf]

looks like an error - which Neil M40 has (almost) identified: my bet is that all the competitors times have been uploaded as if they were on a single course

needs erasing and starting again

[graeme]

some nonsense

[greywolf]

Meanwhile at Chorley, Carl Edmonds of BARRO wasn't given any points in the C final. But if he had been, he'd have got more than GG or anyone else.
Slower runners are still overrated (but greywolf knows why now).

Nope, looks like it was very roughly 1s = 1 point...and so he'd've got 1300 - 1310 pts (cf. GG 1380 points and Ali scoring 1324 in the D final )

So what happened to your cunning plan to score loads of points by "deliberately" only qualifying for the C final?

[Spookster]

looks like an error - which Neil M40 has (almost) identified: my bet is that all the competitors times have been uploaded as if they were on a single course

needs erasing and starting again

Yes. It is no longer possible to score points on any course that has less than 10 ranked runners, but at Bramham two runners on Orange are being given points. That suggests that the system is not recognising the different courses.

The SoL scores uploaded yesterday look fine, so the system is working, but it's probably just the Bramham event that needs to be re-uploaded.

[andy]

It is no longer possible to score points on any course that has less than 10 ranked runners

there goes points from most SOL black courses, then

[DM]

Yes. It is no longer possible to score points on any course that has less than 10 ranked runners,

So, a course run early in the year is won by a long way by an M16 who is less than 16 after a brilliant run, with over 10 ranked runners gets 0 points - and an M16 (over 16 who comes 10th at a race much later in the year after a run full of errors) gets a good number of points. And we think this system is going to motivate our juniors?

Let's cut the nonsense and include all 16s, first and second year, in the rankings. This would be in line with the majority view expressed during the consultation on the rankings system.

Hopefully the full minutes of the meeting will give the exact reasons behind the decision.

[Garnon]

2nd year M/W16's and first year M/W18's are now included in the BOF ranking list. A good decision in my opinion

Link here: http://www.britishorienteering.org.uk/news/view_news.php?id=MTk0MQ==&rtn=aHR0cDovL3d3dy5icml0aXNob3JpZW50ZWVyaW5nLm9yZy51ay8=

[Scott]

A good decision, although it seems to have cause further bewilderment to the algorithm, which now reckons that two first-year M18s running a Light Green course deserve what appear to be the two highest scores currently in the system. This time I'm not sure whether the anomaly can be attributed to the seed data, but I expect that some mathematical type on here will be able to explain what's going on...

[Scott]

Also, there are some peculiarly large scores on M/W16A courses at the Scottish 6 Day. Previously, there were peculiarly large scores on M/W18L courses at the 6 Day, but these now seem to have vanished with the first year 18s being ranked.

I'm therefore guessing that these unexpected scores are down to skewed distributions as a result of around half the runners on those courses being excluded from the ranking calculations. Perhaps, as well as the minimum number of ranked runners on a course, we need a minimum proportion of ranked runners? (I have a feeling that I've suggested this before, but can't remember when or where.)

[DJM]

There appear to be fewer than 10 ranked runners on this Light Green course and it should therefore not have been included for rankings - a good illustration of the anomalies which can occur where there are too few previously ranked runners.

I'll look into it!

[Scott]

I initially thought that as well, but from the results it looks like Light Green Men and Light Green Women were actually on the same course, so presumably the system has (correctly) treated them as the same course for ranking purposes, giving a total of 14 ranked runners.

[DJM]

Well spotted!

I'm still trying to find out why this happened however.

[Adventure Racer]

Combining LGM and LGW they are indeed self consistent. I think the anomoly is due to a combination of a relatively high SD of the ranking points along with the top two being well over 1SD faster than the average time. I was going to be all clever and work out exactly what the means and SDs were, but realised the problem is that I don't know which runners were ranked before this event - less than 10 of them currently have ranking points from events before this one! Fundamentally the issue is a combination of a wide spread of times in a small field (likely to be only 10 or 11 previously ranked runners) combined with some of the runners included in the calcs having inaccurate ranking scores to contribute, which all makes for dodgy stats.