Myth-Busting: The Cheltenham Gold Cup

Bobs Worth is favourite for this year's Gold Cup

In the latest of his series in the run up to the Festival, Simon Rowlands supplies us with some perhaps surprising results on the Gold Cup.

One of my books of the year in 2012 was "Bad Pharma", written by Ben Goldacre, who was also winner of the unofficial Simon Rowlands' Blog Book of The Year in 2009 with "Bad Science". 

Neither book is about horseracing, but both are very concerned with scrupulous analysis and reporting of data. In that respect, they are must-reads for anyone interested in tackling racing by similar means.

In Bad Pharma, Goldacre makes a particular point of identifying so-called "publication bias" as a problem. This is the process whereby only certain types of research - such as those favourable to those connected with that research - make it to publication.  

Publication bias is rife and gives a distorted view of the subject being analysed, which can be a matter of life and death if, as in Goldacre's case, your subject is medicine.

Goldacre has been a driving force behind the "All Trials" public petition, which asks that all clinical trial results be reported, regardless of their conclusions. I - perhaps getting a bit carried away with my title of Head of R&D at Timeform - was one of the early signatories of the petition. 

All Trials probably could not give a fig about publication bias in horseracing, where the consequences are trivial by comparison. But publication bias can still exist in the sport, and so it is that this week's "Myth-busting" blog will in fact be a blog that does not bust a myth and which could have easily therefore not been published.

There are many shortcomings of so-called "Trends Analysis", several of which have been tackled on these pages already. But "trends" do sometimes stand up to closer scrutiny and possibly even shed light on deeper truths. 

Where the Cheltenham Gold Cup is concerned, "trends" have plenty to say, including that "younger is better" and that "a lack of a recent run is not necessarily a disadvantage". Both are probably correct. 

Horses less than 10 years old have beaten 53.1% of their rivals in the Cheltenham Gold Cup this century. Horses aged six (a very small contingent) and seven have done even better, accounting for 57.6% of their rivals and posting five wins where only 1.35 wins could be expected by chance. 

Horses that had their most recent race before the Gold Cup in December outperformed all other months by a long way, beating 61.3% of their rivals and winning eight races when 2.2 might be expected. 

That figure would have been even better had the King George VI Chase not been postponed in December 2010 to January 2011: Long Run won that race and the 2011 Gold Cup, while Kauto Star finished third in both races. 

And that gives us this week's two caveats about trends analysis, regardless of how skilled that trends analysis might be done.
 
Firstly, trends analysis require that we put performances into strict categories, and those categories can misrepresent the underlying truth. This is particularly so when, as with the delayed 2010 King George VI Chase, a performance indicator shifts subtly from one category to an adjacent one (this has been termed a "boundary problem" in science). 

A more granular measure, such as specific "days since last run", would be better, but would then need to be aggregated and recategorised again in order for any meaningful conclusions to be drawn.

Secondly, perhaps what is really being identified here is that the very best horses often go from the second-best chase of the year (The King George VI Chase, usually run in December) to the best chase of the year (the Cheltenham Gold Cup, run in March) without a race in between. 

That is, an absence since December introduces a positive selection bias, rather than identifies an advantage of an absence itself. 

Another trend that stands up is that a horse should ideally have had five previous runs in the current season (63.9% rivals beaten, impact value of 2.58). But there is little between the individual categories of horses with fewer than five runs, which may be of comfort to those who have supported once-raced ante-post favourite Bobs Worth. 

Further encouragement may be gleaned from finishing position on previous run, with last-time winners (63.2% rivals beaten and IV of 2.78) far outperforming alternatives that have more than a handful of qualifiers. 

Among the ante-post market leaders for the Cheltenham Gold Cup, nearly all "qualify" on a number of the aforementioned trends (Imperial Commander is a notable exception), not least Long Run, last seen when winning the King George VI Chase, again, in December. 

I think that's worth publishing, even though it busts no myths, and I hope you do too.