ranking list

[mike g]

... and the third best score of the year, and by far the highest by a woman, was Cat Taylor as the only runner on W20S at Graythwaite.

[Spookster]

... and the third best score of the year, and by far the highest by a woman, was Cat Taylor as the only runner on W20S at Graythwaite.

Yep, that score of 1423 is an anomaly. But she got it by thrashing a lot of other people (who had a reasonable ranking position) on the same course - the fact she was the only W20S is irrelevant.

[andypat]

Man, this is complicated. I think I understand the principle, but as a mediocre orienteer trying hard to improve I'd like to understand more about how the ranking scores work in practice.

I think I am right in supposing that the fastest and best orienteers (for the sake of argument I'll assume they are M21E but feel free to disagree!) will always be at the top because they get more points for beating the other highly ranked M21Es on the same course.

WHat I would like to understand is whether (given a run of the same standard) does it make any fundamental difference whether I run my normal course, up one or down one, to the points I receive?

OR does it simply depend on who I beat (or who beats me) on the day.

and finally - since this has happened to me before! If I beat someone better than me (but they mispunch) is that taken account of in the new ranking scheme or does their reslut get taken out of the equation?

[graeme]

Yep, that score of 1423 is an anomaly. But she got it by thrashing...

Thrashing being the operative word. By so much that the mean and median runners were over 100% behind her, and therefore cut out of the ranking analysis. Leaving an insanely skewed distribution which was then treated as a bell curve.

Which emphasizes why no aspect of any statistical analysis should be based on outliers (i.e. the winner).

Is there an echo in here?

[jac]

no aspect of any statistical analysis should be based on outliers

And furthermore... some ranking lists insist that you throw away your best scores for just this reason. So for example if you want to have a ranking list based on six "representative" scores, the rule might be that the 3rd to 8th highest scores for each athlete would count.

[Spookster]

Man, this is complicated.

Well no, not really. The stats might seem complicated (only people like Graeme and Scott can explain them ) but the principles are fairly simple.

Yes, you could expect the best M21s to be top of the single list, because they are the fastest. In any race, the points you score depends on (a) how well you run against the people on your course, and (b) how many points those people had before the race.

Here's a scenario:
Let's say you turn up at a Level 2 event, and it has colour courses on offer. You can choose to run any course.

If you choose to run Green, you reckon you could win the course. Say the people who typically run Green have an average ranking score of about 900. If you then beat them, you'll be awarded more than their average, so let's say you get 1100 points.

If you choose to run Black, you'll probably be up against a higher quality field. Let's say that field of runners already have an average of 1100 points. Your result is that you finish about in the middle of the group, and you therefore score the average 1100 points.

So, the result is that your ability should get you about the same number of points, regardless of which course you chose to run.

The same logic applies in theory to an age-class courses event. You can choose to run Long or Short, and should score about the same number of points whatever your decision. You'll do relatively better on Short, but you'll be running against people with fewer points.

Of course practice and theory aren't the same, and right now (with the system only just launched) there are anomalies that may make it seem that you can score more points by, for example, running Short classes.

If a highly ranked runner mis-punches, you get no credit for beating him, because he is excluded from any ranking calculation. Good runners who have a bad run and then choose to deliberately mis-punch to stop others scoring well are not acting fairly.

[Mrs H]

Good runners who have a bad run and then choose to deliberately mis-punch to stop others scoring well are not acting fairly.

That's a nice point Spookster - and yet I personally have heard half a dozen "good runners" admit to doing just that - one most pre-eminent orienteer on the grounds that it wouldn't "be fair" to give other runners the impression they had beaten the "good runner" in true terms.

Is anyone on the board prepared to admit having done such a thing and explain the thinking behind it?

[mike g]

Yep, that score of 1423 is an anomaly.

Seems to me it's not so much an anomaly as evidence that the scoring is massively biased in favour of people running shorter courses.

Take the City of London Race scores as another example (the results have now had ranking points added, though a couple of courses are currently missing). You could get 1000 points by running 8.4k in 1:22 on M Open or by running 5.4k in 1:07 on M55+/W40+. A person who scored 1000 points on M55+ would only have scored about 700 points if he had run M Open at the same speed.

[graeme]

Seems to me it's not so much an anomaly as evidence that the scoring is massively biased in favour of people running shorter courses.

That result is primarily due to the penultimate sentence of Appendix K, section 2.2.4.
Obviously

There does also seem to be a short-course bias, I'm not sure why. There's no information about how the algorithm was seeded, so that might be the problem, and if so it will fix itself. There's also a systematic problem that people expecting to run badly by their own standards (ill, unfit, tired, injured, whatever) will tend to run shorter rather than longer courses, pushing up the perceived quality. If you want to rack up the points, pick a course with lots of pregnant women.

[Spookster]

Good runners who have a bad run and then choose to deliberately mis-punch to stop others scoring well are not acting fairly.

Of course there are also good runners who are having a bad run, and retire from the race. Maybe they have a niggle that is hurting. Maybe their mental focus is all wrong. In those cases, retiring is the right thing to do in order to avoid injury or demoralisation. It's supposed to be enjoyable, isn't it! But you couldn't then accuse them of retiring to stop others scoring good ranking points.

So it's not black and white. Thankfully most people in this sport demonstrate excellent sportsmanship.

[EddieH]

I thought it was an automatic rolling year. How come Jan 25th results are still included?

[distracted]

The single ranking list will only work well if there are enough people running across courses to be able to provide an accurate comparator. If everyone rigidly sticks to one course then this problem will most likely persist - this is probably the cause of the apparent short-course bias. Events with fewer courses, like the city races, old-style colour-coded events, middle distance courses etc will probably give more runners per class and help stabilise the system over time.

Eddie - the ranking list updates once a week on a Thursday, so expect those results to disappear then...

[Scott]

This comes direct from an email I sent to Mike Hamilton a few weeks ago attempting to explain where the cross-course benchmarking comes from - forgive any typos etc.

A quick glance at Guideline B will show where the cross-course benchmarking on TD5 comes from. For instance, assuming a significant number of people continue to run the "recommended" course for their age class (which has been the case so far):

At a "medium-sized" Long event, Short Blue will feature M65, W45, W50, M45S, M50S, M18S, M20S and W21S, all being ranked against one another.

At a "small" Long event, M65, W50 and M50S will now be on Green, scoring ranking points against M70-M85, W55-W85, M55S, M60S, W18S, W20S and W35S-W45S. Meanwhile, W45, M45S, M18S, M20S and W21S will now be on Blue, scoring ranking points against M50-M60, W18, W20, W35, W40, M35S and M45S. Of those, M50 and M21S would have been on Short Brown at a "medium-sized" event, sharing a course with M18, M20, M45 and W21.

And so on. And that's not even considering the different combinations of age-classes on courses at Middle, Urban and Sprint events, nor the daring people who don't actually run the same course every week.

[The Loofa]

Good runners who have a bad run and then choose to deliberately mis-punch to stop others scoring well are not acting fairly.

whereas good runners who are pottering around because they have an injury may "choose to deliberately mis-punch to stop others scoring well are" acting scrupulously fairly.

as I have done on one occasion in the past
and anticipate having to do rather more often under the new system.

[Snail]

Is it possible for the algorithm to recognise runners that have performed particularly "poorly" - perhaps because they are ill / injured, or walking around with a novice - and not use them when setting the 'baseline' for an event? eg ignore anyone performing more than say 20% below "expected".

They can still be allocated points for the event - but the "best 6" process would tend to discard them.