Reforming the Portsmouth Yardstick System

I feel for you Jim C.  I spent a week of evenings on Excel post-processing some old race series at my club using Great Lakes, standard RYA and locally adjusted PYs.  It made zero difference, bar so very minor anomalies in individual races for lower placings.  Verdict: PY ranks skill and performance very well.

This probabilistic PY proposal is really, I submit, about the scoring system rather than the handicaps. Whilst I'm not sure that the specifics of Peaky's system would be readily explained to the average competitor, I think there may be a case to consider radically different scoring systems for handicap racing on these general lines.

Essentially with handicap racing we are racing against the clock rather than against individual boats. At the moment if the first ten boats finish within 3 seconds on corrected time they might be scored say 2, 2, 2, 5.5, 5.5, 5.5, 5.5, 9, 9, 9. This isn't really an accurate representation of the actual performance, especially when one considers the likelihood of minor timing errors and rounding making the last second something of a lottery. A better representation of the race results might be scores like 5, 5, 5, 5.5, 5.5,5.5, 5.5, 6, 6, 6.

So I submit that there's an argument that points should be based on time rather than finishing order. What you don't want to do, however, in most cases, is to give points simply by corrected time, in which case a 45 minute race would give better or worse scores than a 60 minute one. So there's a need for some kind of calculation to normalise scores between different length races. Race management being the art of the possible, such a scoring method would be impractical for paper scoring, but perhaps its worth considering for these days with IT readily available.

There's one other issue though. Some years ago I dreamed up a new scoring system that I reckoned combined the best features of high point and low point scoring. I wrote a custom spreadsheet that did all the calcs, and got ready to present my idea to the club sailing committee. I then took some old series results, ran them through my sheet in order to be able to demonstrate exactly what difference it would make and why the revised results were superior. To my horror the revised series results were almost exactly the same, and it certainly wasn't possible to state that the few place changes that did happen were other than essentially random. My new scoring system died there!

The corrected time and rank columns are the results using the usual method. The points column is the new bit, so the winning Finn scores 1.5 points because it didn’t win by much and might actually have been second. Similarly 7th, 8th and 9th are only separated by 17 seconds on traditional handicap, which is almost a dead heat given any uncertainty in the yardsticks to use, so score 7.7, 7.8 and 8.3 respectively.

It’s quite straight forward in practise. In addition to the PY each class has a second number, standard deviation, which represents the uncertainty in the PY number. As discussed ad nauseum that uncertainty arises due to low number of returns and the wind, tide and course on the day. The upshot is that no one knows for certain what the PY of the Finn should be relative that the Vareo. So with this method the race results get calculated across a range of PY for each class, and the end result is a likelihood of where you came.

I think that I would need to understand it in order to comment