Beware statistics when preparing your betting strategy

Betting Strategy / Jack Houghton / 13 December 2008 / 1 Comments

Numbers can be seductive, says Jack Houghton, but they can also be very misleading.

It worries me when statistics are used to recommend betting strategies.

It's not the statistics themselves - numbers don't kill people, people kill people - but the way they are used is often pretty reckless. And the problem is statistics bring automatic credibility to an argument that might not always be warranted.

A few years ago I got myself into a bit of a frenzy after Hurricane Run was beaten by Pride in the Grand Prix de Saint-Cloud. After all, how could Pride, a horse rated 11 pounds below Hurricane Run in the previous year's World Rankings, have managed to beat him?

I had a theory: that good horses are less able to stamp their authority on inferior horses in small fields.

And so I went off, beaver-like, to find some statistics to support the theory. In 2,151 races run between 2000 and 2005, where a favourite at evens or less competed in a race with six or less runners, the favourite won 60 percent of the time. Backing all of those favourites to a level stake would have lost you five percent of your money over the period. Refining the search to those favourites at [1.5] or less, the picture was worse - showing a seven per cent loss on turnover.

So there we had it. Theory proved. I think I even wrote an article about it at the time: lay horses at evens or less in small fields and you will win money in the long run.

Problem is I've just looked at the spreadsheet again and noticed that favourites actually did pretty well return-wise in two- and three-runner fields. So all I actually proved was that in four-, five- and six-runner races, short-priced favourites had shown a slight tendency to not win quite as often as their price suggested they should have done. My actual hypothesis - about small fields inhibiting the ability of horses to stamp their relative authority on each other - wasn't really proved or disproved at all.

Some might not care. As long as the resulting strategy returns a profit, the thinking that got us there is irrelevant. Leave truth and intellectual rigour to the academics and philosophers - we'll be too busy spending our profits to worry about all that.

The problem with that attitude is that - with the exception of those who have reduced their betting to disciplined following of rigid betting systems - most of us bet on individual races, rather than on blocks of 2,000 historical races over a five-year period. Betting on individual races, how do you apply a slight tendency you think you have discovered for favourites to struggle in small fields when, in actual fact, you're not entirely sure you've discovered anything at all?

That's my problem with statistics.

Take Wayne Bailey's latest article about jockey form. Now this isn't an attack on it. It's a fine article with some good arguments made. It's just that it got me thinking about the whole issue of how to successfully marry statistics and punting again, and I thought I'd share the reasons why.

First, it starts with a hypothesis - that winners breed winners - and seeks to prove it. Generally, in my experience, statistics sought in order to prove a hypothesis tend to me more unreliable than hypotheses generated by happenstance, as a result of broad statistical work.

When you're trying to prove something, it can be easy (and usually quite unintentionally so) to make assumptions that handily provide the answer you are looking for. Scientists call it "writing the result into the research design". And it's something to be wary of.

So, with regards to jockey form, Wayne makes a number of assumptions that include: a two-week period to assess jockey form (why?), 25 rides in that two week period to assess jockey form (why?), an in-form jockey being one who has ridden 40 per cent of winners in that two-week period (why?), and others. Why?

Does it matter? Wayne's statistics show a profit over a five-year period, so shouldn't we just go with them? Well they show a profit, but only just. A one per cent profit on turnover is a profit; but I would argue it is well within a margin for error that gives no guarantee of history repeating itself. And remember, you could have put the money in a savings account in 2003 and made considerably more money in one year than the system did in five.

And even if you don't look to back all "in-form" jockeys blind over a long period, but instead simply choose to give some kind of preference to them in your day to day punting, how do you quantify the value of their in-form-ness? With regards to the form of McCoy mentioned by Wayne, does it mean that his mounts have a one per cent increased chance of winning? A five per cent? Not sure.

So to conclude that jockey confidence is a significant variable is dodgy. The numbers demonstrate that, in a tightly defined instance, a very small profit is shown backing certain runners. It tells us nothing about jockey confidence at all. And there's nothing "significant" about the result. At least not by any measure of statistical significance I'm familiar with.

Comments (1)

1. Steve Nolan | 13 December 2008

   Over the past few months 'articles' have been written under the guise of horse racing advice and observation which appears to me to be nothing more than digs at Wayne Bailey. Labouchere anyone?

   I'm an avid reader of Wayne's articles and within his posts he does mention that none of his findings should be followed blindly as a system and are merely posts showing his findings, significant or otherwise. Regardless, I've always been of the opinion if a punter does follow a system or tipster blindly wothout their own research they deserve everything they get.

   Surely Jack your must've been taught as a youngster the mantra "If you have nothing nice to say, don't say anything". Either that or may I suggest you find Wayne and fight to the death!

   All the best