2016 PY Predictor

Maybe clubs could usefully take the mean of the PY and SJ numbers?

Originally posted by pompeysailor "Bottom line is the SJ series and other similar large handicap events that use the GL numbers do attract many of the best across virtually all classes.  They also sail good examples of each class (and not some tired old clunkers) - their results at these events should get much closer to the true POTENTIAL of their class relative to that of other classes is the argument."   totally agree - the SJ series is the closest to a uniformed level of ability (and quality of boat/sails etc) we're likely to get across multiple classes

It is also a tiny sample compared to all the racing that goes on, week in, week out across the country.
Luckily for the vast majority, yardsticks are taken and used week in, week out and provide a way of different classes to race against each other. Because of the way it works, numbers will change, and again, the vast majority of people using the system understand that, will laugh when it goes well for them, whinge a bit when it goes badly, and then just get on with things.

In my totally view, based on my own experiences, the system is closer to being accurate now than it was 5 years ago.

edited to add : Steve posted while I was typing my response. Yes, that is the theory. Great Lakes tries to tweak the numbers to the fastest of each class.

... and that weights it towards the POTENTIAL of that class does it not ?  What handicap systems are trying to do is remove the class effect on results leaving 'crew ability' to determine who prevails at major handicap events.   Is this not the aim of both the PN system and the GL variant implied or explicit as well ?  

Bottom line is the SJ series and other similar large handicap events that use the GL numbers do attract many of the best across virtualy all classes.  They also sail good examples of each class (and not some tired old clunkers) - their results at these events should get much closer to the true POTENTIAL of their class relative to that of other classes is the argument. 

Just because some see the GL system as challenging the status quo (the nerve !) that does not mean their sample selection or weighting is 'guesswork', biased or whatever anymore than the traditional approach for PN is ....  It is simply naive to suggest both systems do not have the same objective - or either could not be improved for that matter.  Many of us consider the GL numbers are currently getting much closer to achieving that common goal - reducing or eliminating the effect on handicap results of the 'hardware' employed.    

You are all trying to do the same thing - select the sample that 'best' achieves the same goal.   PN and GL just have  different views on what are the most efficient samples to use and how to weight them.   Is this not true ?

How about a probabilistic PY, ie not a single PY for a class but a range of PYs with probabilities e.g. 10% chance a Laser PY is 1070, 30% it is 1090 etc. then calculate the results to show there was, say, a 5% chance you won, 27% chance you were 2nd etc. then add up the cumulative scores to get an aggregate score for the race. ;-)

Edited by Peaky - 04 Feb 16 at 7:08am

Originally posted by Jeepers TBH, the Laser PY has always been an issue.

Not an issue, its a mystery. There is so much data for Laser and Solo handicaps that there is no chance of a problem with the statistics per se. Once the online system has settled down and the more major improvements on the jobs list have been made then for the first time it may be possible to do some detailed analysis of how performance varies across classes.

I have done some sudying at my club, which may or may not be representative. I had enough data to get an idea of trends, but not enough to reach firm conclusions. Also the detailed numerical analysis is beyond my pay grade. It needs much more sophisticated analysis than I was getting from Excel.

My club Laser and Solo fleets are reasonably equally strong, and when people swap between the two classes either way they seem to end up in much the same place in the pecking order. But the time difference between mid Laser fleet and top of Laser fleet is much less than top of Solo fleet and mid Solo fleet. Similarly the time difference between say middle third and bottom third Laser fleet is greater than matching Solo fleet. And note that I have pretty good confidence that crew skill is pretty even. To put it in statistical terms the Laser results have a much greater skew factor than the Solo results at my club, and the evidence suggests, not very conclusively due to not enough data, that this is not due to crew skill variations.

The effect is pretty visible though. If I sit down with a spreadsheet and a bunch of my club series results, then if I pick a handicap that brings the top Laser sailors what I reckon would be their fair share of race wins then the top quarter of the results are simply stuffed with Lasers: the Laser fleet as a whole is quite comically advantaged. Its clearly and obviously wrong. But if I pick a handicap in which say the top halves of the two fleets are evenly mixed up together then the top Lasers don't win as many races as I think they ought. This is just my club though. It may not be reflected nationally.

If it is though the implications are challenging. If a handicap can only be "right" for part of the fleet, where do you pitch it? At the average fleet, greatest good for the greatest number? At the top half of the fleet who perhaps care more? Do you try and pitch it at race winners which is firstly a statistical nonsense and secondly impossible for smaller classes? The "fair" solution, if "fair" meant that sailors of equal ability got equal results right through the fleets, would actually be to have variable handicaps through the fleets so that the top Solo sailors have a harsher number than the mid feet Solo sailors. Don't fancy trying to explain *why* that would be fair on this or any other forum...

Edited by JimC - 04 Feb 16 at 4:02am