ranking list

[AAH]

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

Under this system, the reliance on outliers means that if someone at the bottom end of the results (but not so far down that their time is not counted) makes a mistake they will have as dramatic an effect on scores as if one of the leaders made the same mistake…

When anyone makes a mistake they slightly increase the average time and everyone else scores a little bit more, irrespective of where they finish. If there are a lot of competitors that effect is pretty small.

If a mistake is made by someone who is faster than the mean time, and particularly if they are at the top of the field, the mistake reduces the standard deviation of finishing times. This increases the number of points awarded for every minute faster than average (and lost for every minute slower than average) so people at the top of the field benefit and those at the bottom lose out.

However, if the person who makes a mistake is someone who is slower than the average speed, they increase the standard deviation of finishing times – this has the opposite effect – the slowest runners benefit and the fastest runners lose points.

So, if someone makes a major error and finishes well below average, but is not quite so slow that their time won’t count, it might be good etiquette for them to mispunch to avoid reducing the score that the winner gets…

[DaveK]

Nice to get some clarity in postings for a change?

[Blonde bombshell]

And I thought it was just me

[Adventure Racer]

So, if someone makes a major error and finishes well below average, but is not quite so slow that their time won’t count, it might be good etiquette for them to mispunch to avoid reducing the score that the winner gets…

So a good score will also depend on the morals of the other runners?

[AAH]

Nice to get some clarity in postings for a change?

This is the thought sport… we must be allowed to post maths! And there wasn’t even one equation…

My point was intended in reply to the comments on the previous page about highly rated runners who have had a bad run mispunching and avoiding giving away ranking points to competitors – what I was trying to say is that errors by slower runners close to the cut off time above which their result won’t be counted for the rankings have a surprisingly large effect on scores at the sharp end. This is in contrast to the ranking system in the 90s where only the times of the fastest runners affected scores.

I think the new one list ranking system is a great development for the sport which will improve as it settles but in some future iteration some years hence it might be best to use an algorithm that is not so sensitive to errors made by runners who come near the top and (particularly) near the bottom – Snail’s suggestion of removing times that clearly arise from errors would help a lot, for instance.

[Spookster]

As of yesterday, Club Results Secretaries are able to upload their own clubs' results. It's all working nicely, I just uploaded SYO's Treeton Local event.

[graeme]

… we must be allowed to post maths!

Absolutely (and your analysis is spot on)!
I don't care about the ranking list at all, it's only the maths that I'm interested in.
Except my purist friends tell me its not maths at all, just statistics

Quiz question: In the limit of an infinitely large event with a normal distribution of times, how many points does the winner get?

Spoiler alert...
The correct answer is "nobody scores any ranking points".

[King Penguin]

Not sure I agree with part of what AAH said, unless I misread it.
Standard deviation is a measure of dispersal from the mean, and most appropriately applies to normal distributions.
In a normal distribution I'm sure a big mistake by someone x% above the mean would have the same effect as the same size of mistake made by someone x% below the mean.

However, before Graeme jumps in, results are usually skewed with a long tail, so the effect of an equal size mistake above and below the mean may well be different. My brain's too rusty to work it out...

[AAH]

Standard deviation is a measure of dispersal from the mean, and most appropriately applies to normal distributions. In a normal distribution I'm sure a big mistake by someone x% above the mean would have the same effect as the same size of mistake made by someone x% below the mean.

King Penguin – I think/hope that we are both right - the magnitude of the effect of an error by a runner x% above the mean and a runner x% below the mean should be the same (where x is small). So, yes, I agree that the effect of a mistake by a fast runner and a mistake by a correspondingly slow runner should be the same.

But the point I was trying to highlight is that these effects have opposite senses – a mistake by a fast runner reduces the dispersal from the mean, a mistake by a slow runner increases it. Thus, errors by the leaders give other fast runners more points. That seems perfectly fair. But errors by slower runners near the cut-off (beyond the errors you would expect from their existing score) give the fastest runners fewer points… That seems less fair...

In saying this, I haven’t looked at how far the winner’s time + 100% cut-off usually is from the mean or relative number of runners at either end of the distribution, but the importance of slower runners does seem exaggerated.

[Scott]

In saying this, I haven’t looked at how far the winner’s time + 100% cut-off usually is from the mean or relative number of runners at either end of the distribution, but the importance of slower runners does seem exaggerated.

As I understand it (and I'm sure graeme or greywolf will correct me if I'm wrong), one of the weaknesses in the current system is that winner's time + 100% can exclude far too many of the slower runners - eg. on that course where Cat Taylor picked up all those points at Graythwaite, when 48 out of 72 finishers were below the cut-off, thus inflating the mean score on that course.

[greywolf]

Re: the debate about deliberately mispunching, and whether bad runs gift points, and whether the algorithm should eliminate those running below par:

Firstly, the logical conclusion of systematically removing times affected by errors is that you end up only counting scores from people who had a perfectly clean run…so it's a bad idea, and the only exception should be the trimming of genuine outliers from the tail (because whether the last finisher took 2 hours or 4 hours including a picnic shouldn't affect the points scored by everyone else)

Deliberately mis-punching because you’ve made mistakes and you don’t want your fellow competitors to benefit in any way is at best poor sportsmanship and at worse pure spite.

On the other hand, if you know before you start that e.g. you’re injured or hungover and you intend to walk or jog round, then you have the option to declare yourself non-competitive before you start and your time won’t be included in the calculation of rankings points

But in either case you need to get a sense of proportion – the effect of your performance is indirectly proportional to the size of the field (or maybe the sq root of the size of the field?) and in a large field the effect is trivial. If you have an average points score of 1250, and the average points score of the rest of the competitors is 1150, then in a field of 100 your inclusion raises the mean points score of the field by 1. The difference between a 1250 point performance and an 1150 point performance is usually about 20% (judging from a few M45L results I’ve just looked at) which is a pretty major differential.

Lets say you’re a reasonable M45L with an average point score of 1250. If you turn up at BOC and run to form (c. 68 mins) you’ll be in the top 10 and score 1250 points. If you jog round or have a complete 'mare you run 16 minutes slower than expected, end up say 45th pretty much in the middle of the field (and score 1150 points)…or you could mispunch and remove yourself from the results.

So what effect do these 3 options (good run, crap run, deliberate mispunch) have on your fellow competitors? Well judging from last years’ M45L at BOC:
For the winner (<60 mins) the respective scores are 1304, 1306, 1304
For a “bit better than average” competitor (c. 80 mins): 1173, 1174, 1173
For a competent tailender (c. 100 mins): 1044, 1044, 1044

so what you do has a 2 point impact on the winner, a 1 point impact on the bloke in the middle and no effect at all on the tailender

But that was a big field (>90), what happens if you strip out 2/3rds of the competitors to give a field of 30? The effect is bigger...but still barely noticeable:
For the winner (<60 mins) the respective scores are 1304, 1311, 1304
For a “bit better than average” competitor (c. 80 mins): 1173, 1177, 1173
For a competent tailender (c. 100 mins): 1044, 1045, 1044

Basically the extent to which your crap run gifts points to your fellow competitors is pretty trivial – if you’re sad enough to be bothered about a couple of points here or there then you’re even sadder than I am for working this out

[greywolf]

one of the weaknesses in the current system is that winner's time + 100% can exclude far too many of the slower runners - eg. on that course where Cat Taylor picked up all those points at Graythwaite, when 48 out of 72 finishers were below the cut-off, thus inflating the mean score on that course.

Yes - you end up with one huge outlier and the rest all close together - something like the left half of a bell curve*. The mean time increases a bit but the standard deviation becomes very small, and the outlier can be 3 or 4 SDs from the mean....which gives huge points.

*please don't lets argue about normal distributions...

[Knee Deep Mud!]

Whilst I can just about follow all the statistical discussions, after looking at my position in the list this week, compared to last week, I am wondering how accurate the basic maths is.
I say this because the column stating change in position since last week says I have gone up twenty something places during the week, yet my actual position has moved up by well in excess of a thousand places. So what are they using to generate the change in position figure, as they are obviously not simply taking last week's position away from this weeks position to give the value?
I just hope this rate of improvement continues, as it means I should be No.1 by next week ! (and all without doing any form of orienteering or running in the last three weeks!!!)

[andypat]

Somethings not quite right there I think - everyone in Clyde went up this week and I'm sure most of us havent run. Unless a lot of people down south have had an event drop off?

[Snail]

Even the people at the end of the list have gone up 21 places.

Various possible reasons: events 12 months ago dropping out (which could remove some people from the list completely), teething problems with individuals possibly duplicated / moving from family to individual membership etc.

(But agree an increase of 1000 places look odd, unless perhaps someone had only 3-4 counting events and missing ones have been added to increase to 6.)