Rise Of The Ok etc

Originally posted by iGRF Ask them about the L3ks fitted out with all the V3ks running gear(bigger sails, gnav, bigger foils) and raced off the L3ks handicap, I very nearly did it as a gag to 'get' the merlin boys at their own game this summer.

Originally posted by Medway Maniac Originally posted by iGRF Ask them about the L3ks fitted out with all the V3ks running gear(bigger sails, gnav, bigger foils) and raced off the L3ks handicap, I very nearly did it as a gag to 'get' the merlin boys at their own game this summer. Do it.  Let's see how you get on.

You already know the answer

 

The system has been in use since Noah decided paddling around was a bit of a drag.   It was not intended to be anything other than a way of running the occasional 'fun' races for a wide variety of classes who normally raced in their own fleets.   Times however have changed.

Nobody was too worried then that it was a crude and often highly limited approach.  It was ‘fun’ occasionally and most seriously believed then, as some amazingly still do, that it measures 'boat potential'.  The trouble today is two-fold 1) Handicap racing is today 'big business' and has become much more important and is taken very seriously.   2) What our current system measures is not 'boat potential' except very obliquely.  There is a correlation but there are a lot of 'extra' factors that should, many of us will argue, be taken into account.  Some of these factors are very influential indeed and can make the 'number' produced rather less than reliable.

The standard defence is "look at the sheer volume of numbers"  .... However the 'data' has to be looked at very carefully indeed if  to be included is the gist of the counter argument.  In summary that can be expressed as ... a) why on earth are we considering non-competitive examples in any class at all ? - Surely they merely distort the resulting EN/PN numbers ?  b) Why not just consider the front 10% (and you might equally even argue the front 25%)  ... that is if you really want to assess 'boat potential' in a more pure form.... ‘Drop the Grot’  if you like.     

The GL approach goes some of the way to eliminating the sample problems of the core PN system.  (a big subject in itself of course)  However why though do the GL developers still feel the needs to anchor itself to the 'core' PN database in any way ?   Few who has raced in the winter Sailjuice Series or in the growing number of other events that use it think it anything less than a step in the right direction.  The racing is very close as the developers hoped and these events are well run and well supported.    The sheer quality of which is the best indication that change really was and is still needed. 

Come on 'PN' - look more at the indicative quality of the raw data and less at the sheer quantity of it !   It is the front of each class fleet who should be handicapped appropriately ... after all real elephant in the room is that 'bandits' will always arise when you just look at averages and sheer data volume.  Loose correlation, however large your sample, does not necessarily indicate a clear causative link.  

 What we all want to know is whether the BEST examples of 'class A' class are faster or slower than the BEST examples of 'class B' etc with their respective handicap numbers.    A few of you know must know how to shift the core formula and methodology on to make PN  more relevant and more ‘fit for purpose’ ... surely.   

If necessary give the problem, data examples and objectives to any decent graduate statistician and ask them to report back in a few months..  You don't even have to adopt any recommendations,  but you might just want to !  

Mike L.

I guess I probably count as graduate statistician (I'll leave judgement to others on "decent").  Yes of course it would be possible to come up with a scheme such as you suggest if a) the PY scheme was much better funded and/or b) we could insist that clubs provided more extensive and higher quality data and/or c) the clubs would be happy for us ignore minority interest classes. 

I have shown that this is possible at our club and can fix the relative performance at any decile that you fancy as you describe (as I said above, my mucking about with the data shows that it looks to have the best repeatability using the 2nd and 3rd deciles).

The problem comes in devising a "practical" scheme - i.e. not requiring a boat load of cash or a massive step change in the club's commitment to data provision.  I can think of several things that could be done to the PYS to make incremental changes, but the first problem to fix is data quality going into the PYS system.  Big improvements have been made over the past 3 years and I think further changes are being considered to improve it further.  When this is done it should be possible to add some semi-automated skill factor analysis in the data with only small changes in what the clubs have to provide, . .  and then you would be able to do performance potential comparisons.

Not sure why you think the GL scheme gives different numbers to the PN scheme if you only analyse large open waters - do you have any analysis that shows that difference.  Not just one class - but a structural difference.


MM - responding to your 95th percentile.  Yup for the big classes - doesn't work so well for the smaller classes like the 3000 etc.



Edited by marke - 07 Oct 14 at 1:41pm

Originally posted by winging it passes iiitick a custard cream

Can I join in if I bring some chocolate biscuits?