Amendment

The relationship between the two measures of genoa overlap discussed in the second section of Article 2.9 has been amended slightly.

Model sensitivity

My current search for model improvements (however small) has focussed in part on assessing how a number of statistically equivalent models behave when applied to estimating handicap numbers.  In particular, the sensitivity of results when the input data has some inaccuracy can point to the type of model components contributing to making a model robust. 

Clearly, the effectiveness of any model relies on the accuracy of input data.  Article 2.5 states the accuracy needed in the required ‘brochure’ measurements.  To date, the most efficient models require extra information about foresail area and this may involve user measurements and estimation.  Newly added here is Article 2.9  which discusses this issue, and articles 2.5 onwards have links with 2.9.         

 

 

Autumn update

The most recent articles to be added here are 2.8 and 3.2.  New readers should refer to the last posting for a description of how the articles are organised. Work is ongoing to increase the amount of reliable boat data available for model development and to search for any worthwhile model improvements.

The structure of this website

This site is intended to provide useful background and information on current developments in the estimation of yacht handicaps.

There will be accumulated material of general interest to the casual reader and as guidance for those engaged in handicapping at club level.  There is also for technical readers material relating to the statistical science behind the modelling.

The ‘Articles’ menu has three classes of article. The lead number indicates the following:

1:             General and historical.

2:            Material concerning formula -based handicap models.

3:            The statistics behind the model building.   

 

The beginning

Back in 2006 I did what academics do when they want to see if something is a good idea: they get a student to try it out first!  I offered to undergraduates, as a final year project topic, the problem of finding a simple linear regression model for predicting sailing handicaps.

It was fortuitous that a student stepped forward who was very able and a sailor. He obtained data on 50 boats in the Caribbean Sailing Association (CSA) and showed that a simple formula using waterline length, beam and sail area gave a respectable approximation to the CSA performance ratings.

My target was closer approximation, but the project demonstrated the potential.

(The student, by the way, won a national prize for his work.)

 

Why model sailing handicaps?

It is some years now since I had my arm twisted into looking at the possibility of developing a simple and statistically sound formula for producing a yacht handicap based only on basic boat measurement data.

Why did I need to be persuaded? First of all, I know that this is the stuff of much argument in the common rooms and bars of sailing clubs. Second, how do we  know what is the perfect handicap rating for a boat? The vast majority of handicap setting is based on performance data. The problem with this is that too few data are available for each of the many hundreds of different types of boat out there, and what there is will be contaminated by factors due to crew, weather and sailing environment.

In spite of the reasons for not getting involved, in the end I could not resist the challenge of this enticing, but very frustrating problem.