Potts and Segal are idiotic dinosaurs. Their methods of jockey assesment are based on their self-acclaimed 'expertise' in race reading, which is frankly laughable. In the computer age any serious method of jockey assesment has to take advantage of the huge amount of data we have on each jockey's riding history and how they have performed in comparison to the opportunities they were given.
"Potts and Segal are idiotic dinosaurs. Their methods of jockey assesment are based on their self-acclaimed 'expertise' in race reading, which is frankly laughable."
Various methods have been used to try to compensate for the differing opportunities jockeys have but they basically fall into two categories: those which compare how jockeys have run relative to their position in the market (used by flatstats, Adrian Massey etc) and those that compare how jockeys have performed on the same horse (as advocated by John Whitley). The former methods show Richard Hills in a particularly good light. However, they don't really tell us whether one jockey is better than another or simply one jockey is better thought of by the punting public than another, so we are left with comparing how jockeys have performed on the same horses - a method we shall now consider.
As far as I'm aware the only published ratings of this sort are produced by John Whitley, who has been producing annual ratings for over two decades. He claims his ratings are 'the result of an exhautsive examination of how each horse has performed under its different jockeys', which is all well and good but there are several drawbacks to his method. Not least of these is that we don't know exactly what his method is - he never spells it out, or at least I've never seen him spell it out, beyond the use of phrases such as 'mathematically correct way' and 'proper mathematical treatment'. This is more than me being picky - the common method of doing this statistical analysis is regression, specifically OLS regression, and the ratings can be meaningless if the analysis isn't done properly.
"The principle of garbage-in garbage-out applies. If the handicap ratings are wrong in a systematic way then the subsequent jockey ratings are likely to be wrong.."
Leaving aside such potential diagnostic problems, there are three main problems with Whitley's method:
1. He assumes that by holding the horses consant he is holding everything constant. It is doubtful whether this is the case. In particular more fashionable jockeys will tend to be put up when the horse is fit and suited by conditions, when a horse is out of form or out for the run less fashionable jockeys are likely to be put up. Thus more accurate ratings are reached by restricting are comparison on a run by run basis, looking at jockey switches only since last time out.
2. The difference between being beaten half a length or winning half a length is given equal weight to being beaten thirty lengths or beaten thirty one lengths. Whitley sees this as a strength of his method, because it picks up these differences nobody sees. In fact it's a weakness - it punishes jockeys for easing down beaten horses and places too much emphasis on meaningless 'noise'.
3. The principle of garbage-in garbage-out applies. If the handicap ratings are wrong in a systematic way then the subsequent jockey ratings are likely to be wrong. In particular getting the lbs per length calculation wrong, as I believe Whitley has, can bias the results in favour of low strike rate jockeys that tend to ride rags. This can have a knock on effect in causing lower class races to be over-rated, which would bias the results against light weight jocks.
Despite these drawbacks, regression analysis is light years ahead of listening to what Potts, Segal or some bloke down the pub thinks. I've included ratings for all jockeys, derived from such analysis, with a reasonable number of rides here.
It looks at all jockey changes from Jan 1st 2000 to mid-October 2004. To avoid accusations of cooking the books, and to make the analysis replicable, I have used raceform ratings. The number you should be interested in is the first number in the B column. This rates how many lbs above or below average the jockey in question performed on average. The other numbers are confidence measures of various hues and are included for any stats buffs. The ratings compare how jockeys have performed including any claims that they had.
The biggest surprise here is that R Hills is only in the top 10% and not number one. Most of those ahead of him owe their positions to the claims they are alllowed to make, the rest can be accounted for when the adjustments are made to a proper lbs per length scale and weighting the analysis to only take into account 'competitive' performances. Rest assured, the figures say it loud and clear: Richard Hills is number one.
BBC
