New Event Structure

[Harley]

..... Young Neville ....... is resolved to get out and do more events (at the shorter distance

Bugger! I'll have to step down to the string course then!

[swat]

will any colour coded events run in the transition year up to the date the new list comes out be taken into account

The current ranking list has been updated to allow colour coded events to be submitted (more details). Results from the Aire Adel & Meanwood new-style regional event are up there already.

[greywolf]

The current ranking list has been updated to allow colour coded events to be submitted (more details). Results from the Aire Adel & Meanwood new-style regional event are up there already.

Seems somehow fitting that the brave new world is ushered in with a "new-style regional event" where the top course is a short Short Brown at TD4 ....

[Scott]

An awful lot of people at Adel & Meanwood are either "unranked in this class" (ie are "running up" on what the class assignments tell them they should be running) or have an "invalid class" (ie "running down"). It seems that people do want freedom of choice after all

[greywolf]

An awful lot of people at Adel & Meanwood are either "unranked in this class" (ie are "running up" on what the class assignments tell them they should be running) or have an "invalid class" (ie "running down"). It seems that people do want freedom of choice after all

Perhaps, but I think you need to look a bit closer - something has gone a bit awry in the attribution of runners to classes - e.g. I don't know them personally but I'm not surprised that Alice Leake & Chloe Haines aren't previously ranked at M21L...

Likewise M40L, where it appears that almost everyone is previously unranked, but working down the list, Andrew Thorpe (EPOC) doesn't appear to have any previous ranking at all, Steve Watkins (AIRE) has just moved up from M35, so not surprisingly this is his first score at M40...Dean Field is an M45 who has entered what the guidelines say is the most appropriate course for his age class - i.e. one where the estimated elite winning time is just over 46 mins
(see: http://www.britishorienteering.org.uk/d ... 9DRAFT.pdf ) but his score has been attributed to the "wrong" age class.

Similarly, Chris Mitchell - Number 1 ranked M21S, should also, according to the guidelines be running an EWT course of 46 minutes - he did so - but his score is attributed to M21L, where he doesn't have a previous ranking...

So there's a lot more to it than "freedom of choice"

Looking at the AIRE website there's nothing in the event flyer or final details to match courses & classes, so I'm not sure how that was done, and the colours used don't really match the new system either (short brown, blueish & short green might've been more appropriate).

Btw - i'm sure the event was greatly enjoyed, and I'm not trying to have a dig at AIRE, but it seems indicative of how much confusion there is over the new structure...

[Edit - having looked more closely at the link in Swat's post above it seems part of the problem at least is the mis-labelling of courses - e.g. had the "Brown" been called "Short Brown", which is what it was, then some of the above problems would've been avoided. Still would have had M21L & M21S on the same course tho... Roll on the single list]

[Scott]

Yes, I did spot those - it looks, if you don't run your specified class, the system is assigning you to the "highest" class on the course, and the "mapping down" of absent courses doesn't help. So Miss Haines and Miss Leake, as W18s, are "supposed" to be running Short Blue, but as there isn't a Short Blue they should (supposedly) be running Green. As they were in fact on Brown, perhaps it would have made more sense to rank them in W21L... In any case, they weren't running their age/sex designated course.

Similar problems apply to the M40 list - the M45s should have been on a course designated Short Brown, as should M21S runners. It would be interesting to see what would happen if the AIRE were to resubmit the results using the course designations that you suggest (Short Brown/Blue Short/Green rather than Brown/Blue/Green).

[Scott]

Perhaps a better solution would be to allow clubs to specify what age class-to-course designation they had used when submitting the results to the ranking system. It would mean more work for all involved, though.

[EddieH]

Quote "Perhaps a better solution would be to allow clubs to specify what age class-to-course designation they had used when submitting the results to the ranking system."

Pointless effort. Greywolf's figures eloquently show the futility of trying to mix the system. The WHOLE purpose of the changes is to allow people to run anything.

For the rankings to have ANY meaning whatever, there has to be a single list, unless of course rankings are only done on Level 1 events which is not what most seem to want.

[Scott]

For the rankings to have ANY meaning whatever, there has to be a single list

I'm not disputing that, but such a system is not going to materialise in the immediate future. In which case, if an interim ranking list is necessary (and personally I'm not convinced that it is), it does make sense to try to make as good a job of it as possible in the circumstances.

Hopefully, when the new ranking scheme is finally agreed upon it should be possible to backdate it to the start of 2009, so this event will be "reranked" under the new system anyway.

[roadrunner]

For the rankings to have ANY meaning whatever, there has to be a single list

I'm not disputing that, but such a system is not going to materialise in the immediate future.

But Graeme said that the algorithm for the single list was designed two years ago, had been tested and was ready to run a year ago, so surely there isn't any real reason why it can't go "live" pretty quickly?

[Arnold]

How about only counting National Events for the ranking list? They're all using age classes, so you get rid of all the confusion in one fell swoop. You'd have a lot fewer events to count (10 in total including JK & L/M/S champs) but it would at least mean that they will be true National Events. That's how it's done in many other countries.

You can then use the Regional results to do a simple ranking by course, and do that on a regional level only (which is already done today in some regions like the SCOA galoppen).

It feels like the current solution is a compromise between (i) simpler, more open events using colour coding and (ii) the perceived need for ranking lists - resulting in a complicated muddle. If we think (i) is the main purpose of all the changes then you might as well get rid of (ii).

Do people care that much about ranking lists anyway?

[Scott]

But Graeme said that the algorithm for the single list was designed two years ago, had been tested and was ready to run a year ago, so surely there isn't any real reason why it can't go "live" pretty quickly?

Well, nothing is going to happen until the Review Group actually reports. And even then, assuming they (we) recommend an approach compatible with Graeme's algorithm, it's going to take quite a while to write an operational system capable of dealing with myriad different results formats (the main issue), decide on and set up the initial seeding, test it with real-world data, integrate the whole thing with the new BOF fixtures/entries/results database, and get it on the website. And even then there may be teething problems (see the saga of the new online fixtures system). Don't hold your breath.

Arnold - restricting it to just National Events is one of the options under consideration, but my personal feeling is that we'd be better off going in the other direction - aiming to include more events, not fewer. Please do email your views to rankings@britishorienteering.org.uk

[graeme]

Well, Mike Cumpstey managed to

write an operational system capable of dealing with myriad different results formats (the main issue), decide on and set up the initial seeding, test it with real-world data, integrate the whole thing with the new BOF fixtures/entries/results database, and get it on the website.

We exchanged four emails to get the teething problems sorted

It was available in November 2007, see

http://ccgi.orienteering.plus.com/ranki ... /news.html

But its gone now.

All I can remember was that it looked perfectly sensible, and the highest rated run of the year (2007) was GG at Culbin.

For anyone feeling patronised by the lack of maths , times are heteroskedastic so the idea is to use median and interquartile range for average and spread in the distribution, by course, assuming a linear relation between time and ability. Initial rankings were taken from the BOF list for 2006, scaled by BOF age-speed ratios, and run through a year of old results to become stable. All races count towards each persons "ability rating", weighted for statistical significance by the number of people on the course, and again using medians to account for heteroskedasticity. In principle individual's spreads would tell you not only how good you think someone is, but also how confident you are in that estimate. "Ranking Lists" can then generated for best-6 runs to encourage attendance, and/or filtered by age-group, club, association whatever.

And to keep me happy, the definitive version lists athletes in alphabetical order.

Graeme (who doesn't like ranking lists, but does like statistics)

[swat]

It would be interesting to see what would happen if the AIRE were to resubmit the results using the course designations that you suggest (Short Brown/Blue Short/Green rather than Brown/Blue/Green).

No problem, you can see what happens if we do that here. Finding the anomalies is left as an exercise for the reader.
Seems to me that if we submit each course as a separate event with just one class (M21), we'll get a single ranking list. But it won't use Graeme's algorithm. And it would currently be more work.

Don't hold your breath.

I don't think it needs to take too long:

it's going to take quite a while to write an operational system capable of dealing with myriad different results formats

The current system does that pretty well by accepting results in SportIdent format which is output by all the main results software.

decide on and set up the initial seeding

Quite a few single ranking lists were presented in an earlier discussion on Nopesport.

test it with real-world data

Put a test system online and I think you'll get people submitting results to it pretty quickly.

integrate the whole thing with the new BOF fixtures/entries/results database, and get it on the website

Now that might take a while.

[Scott]

No problem, you can see what happens if we do that here. Finding the anomalies is left as an exercise for the reader.

Thanks! Not sure whether it's any better - the main effect seems to be to give a lot of people their first short course ranking - but at least there seems to be less wrong-gender ranking going on (aside from a few men who ran down too far to be eligible for points)...
By the way, should Long Orange be being ranked at all, or is that an error?

integrate the whole thing with the new BOF fixtures/entries/results database, and get it on the website

Now that might take a while.

Yes, but BOF are apparently determined that the ranking system should be brought in-house and integrated with their website as part of this all-encompassing (yet strangely elusive) new database, and as such the new system will be written from scratch.