Are "Black" and "National Rankings" incompatible ?

[Oldman]

I think part of the problem here is the bizarrely skewed distribution of times you get when you arbitrarily discard more than half the runners on the course from the ranking calculations for being under 16.

I can't follow this logic: would you still have had a "bizarrely skewed distribution" if the under-16s hadn't even run!

[Scott]

I'm not disputing that it was a good run (although "fastest woman by some considerable margin" is a little unfair on Aine!), but I'm not convinced it was deserving of quite such a high score. For the sake of completeness, the same applies to this this run (and several other Light Green courses from the CSC first round in addition to the one I linked to above).

This anomaly was very prominent on the M18L and W18L courses from the 2009 Scottish back when it was only top-year 18s who were ranked - Mr Cherry was originally in the top five on the overall rankings list on the basis of some spectacular (1400+) scores which then decreased by around 200 points when they were recalculated with the bottom-year 18s included (graeme even commented on these himself, so I don't think I imagined it).

As Neil says, this doesn't seem to make much sense as the distribution should just be the same as if the younger competitors hadn't run - maybe something has gone wrong in the implementation of the algorithm

[graeme]

I'm suspicious of the algorithm implemention, so when all else fails, I thought I'd look at the appendix K.

The formula uses " MP, the average points for the runners in the race "

However "average points" isn't actually defined anywhere. If the software developers
used for MP the points from the ranking list instead of "current score", this would produce a significant skew towards courses with people who run a lot of races.

Later we find:
"SP, the standard deviation of the ranked runners’ points"

Again, unclear whether this is the sd of their points from individual events (the correct measure), or the sd of their "current score", (wrong) or the sd of their current ranking points (even wronger).

Also, there's a bizarre "small course" formula which, (after some algebra) implies a spread of 20% the average time. Where does this come from? ... History.
So the people who exclude black claiming you can't measure the spread from other courses on the SAME area, believe you can measure it from other courses on DIFFERENT areas.

I hadn't noticed this totally discredited idea creeping back in. (Cue godders "I got the best ever score for night-O on Inshriach")

It will lead to huge scores whenever the actual spread of times far exceeds 20%: which will happen for fields of very diverse abilities: e.g. for CSC when people run down/up and on courses where there many beginners making technical errors.

[Scott]

I think the problem with the small courses formula has been noted, and I was under the impression that at some point that bit was going to be taken out and they were going to be calculated in the same way as other courses, although I'm not sure whether or not this has actually happened (does anybody know - djm, spookster?). The anomalous courses that I was grumbling about above all have fewer than twenty runners, so perhaps it's the small courses thing which is the cause?

[SYO Member]

Graeme, Section 2.2.2 of Appendix K also says that "points awarded to a runners are (as far as possible) only dependent upon the quality of the run and not who turns up at the event".

From what you've discovered today, does this mean that this is not really the case?

I only ask because in the example you gave earlier regarding the Ecclesall event (which
was positively enhanced by all courses have a higher than average number of controls)
and looking at the splits, it seems that both Mrs.Spookster and Barrie Speake had runs
on different courses with no significant errors.

And although there are no splits for Black showing, it looks like Mr. Gallier had some
significant errors as he ended up 10th, 18 minutes behind 9th, and 16 minutes greater
than the average of all the runners. So maybe Homer was correct in his observation.

Even though mins/km are not part of the calcultion and Mrs. Spookster and Barrie did
shorter courses, don't they deserve more points than Mr. Gallier because they had no
errors?

On a personal note, and based on my post in Page 21 of the Results and Rankings thread about having an average counting points of 985, the software developers must have done something right because running Blue with 30 controls at Ecclesall, I ended up a miserable 35th out of 64. This is because although I had been 17th after 17 controls (2nd fastest on Split 6) I then had 3 errors giving a points score of.....982!

Without these errors even I would have beaten Mr. Gallier's total but I didn't because
like him, I didn't have a clean run.

Going to CLARO's Level 2 at Harrogate on Sunday, I'll try and exceed 1000 points for
the first time since April as I'm getting fed up with these 980s.

[graeme]

Graeme, Section 2.2.2 of Appendix K also says that "points awarded to a runners are (as far as possible) only dependent upon the quality of the run and not who turns up at the event".
From what you've discovered today, does this mean that this is not really the case?

quality of the run includes both speed and accuracy. So if I beat you, despite making 15mins errors when you're clean, mine is the "higher quality" run.

That's certainly the intention: that a head to head race should be won by the higher ranked runner. The ranking list does achieve that rather better than its predecessors did, and better than most of the proposed alternatives would.. Of course it depends on how generously you interpret as far as possible. Statistics professionals are likely to be more demanding!

Even though mins/km are not part of the calcultion and Mrs. Spookster and Barrie did
shorter courses, don't they deserve more points than Mr. Gallier because they had no
errors?

No, because he more than made up for it by running faster, and this is a running sport.