Horseracing Betting: Occam's razor and finding a betting strategy that works for me

Betting Strategy / Jack Houghton / 06 December 2008 / Leave a Comment

Jack Houghton tells us how he's finally developed a betting formula that allows him to show a consistent profit and why he's happy to keep it relatively simple.

It's not the most interesting of conversations to find yourself trapped in, but if someone insists on asking about how I punt, I'm happy to tell them. It's pretty simple really. Speed ratings. Tissue prices. Find discrepancies. Punt.

Okay, it's a bit more complicated than that. Not all speed ratings are created equal and there is a fair amount of concurrent qualitative analysis that goes on aside from the raw numbers, but I always try and limit this as much as possible. If the numbers I'm using aren't doing the majority of the work, then they are the wrong numbers to be using.

Invariably, on finding out how simplistic the method, most enquiring minds want to know why I don't make it more complicated. After all, there are a great many more factors that I could be considering and, if taking all these into account, wouldn't I identify more betting opportunities and, ultimately, more profit?

The factors they suggest are things like draw bias, trainer form, jockey form, course suitability, ground, and the list goes on - depending on how deep a grasp the questioner has on the ingredients that go in to making a horserace result. In fact, the list includes all the things intelligent people would think required to develop as accurate a predictive model of horseracing as possible.

So why then does my approach fall so far short of this theoretical standard? Surely, if I'm not considering these things, my methods - and associated profits (no matter how consistent) - are somehow substandard?

My defence to these questions has never been that convincing. Sure, I can show them a consistently profitable account but, deep down somewhere, I know they are right: others are certain to have developed more complex models that bring greater profit.

I've heard stories of large punting syndicates who have developed models of huge breadth and complexity, and believe them to be making huge profits as a result. But does their existence make my (infinitely less) profitable approach inferior?

On one level, the answer to that question is obvious. On another level, I can say I've found (after many unsuccessful years) an approach that works, that suits the limited amount of time I'm able to spend punting, and that brings in consistent profits. And after all, I've heard about these syndicates and their methods, but I don't know for certain what methods they are using. Perhaps they are equally simplistic in their approach; only they bet a lot bigger?

This has been my defence - and after the recent discovery of something called Occam's Razor - I'm becoming more confident in it. Occam's Razor is a scientific principle asserting that, when trying to explain anything complex, the fewer assumptions you consider, the better.

Actually, that definition is a bit simplistic. It's really about asking yourself whether the addition of a new assumption to a predictive model increases the accuracy of your outcome, or decreases it. In other words, if you add an extra element to those you usually use to assess the likely outcomes of a race, you must ask yourself: is this making the picture clearer, or am I just adding a whole extra batch of uncertainty to something that worked quite well in its simpler state?

Last week I wrote about trainer form, and how difficult a concept it is to assess. And does it really have any predictive use? I'm still none-the-wiser. And the same can be said (in my eyes) about draw bias, course suitability and jockey form. On the one occasion I tried to assess the draw bias at a particular track, I spent hours coming up with a measure that, in itself, contained so many assumptions that I am doubtful it had any predictive use whatsoever.

That's not to say that all these factors cannot be used with success - provided you have the time and the mind to ensure the measure you are using adds clarity, rather than more ambiguity. It's likely the large syndicates have found ways to add buckets of clarity. But this in itself should warn us that when introducing any new element to our punting strategy, we should run it in parallel to our existing profitable methods and paper trade it first. One thing you can guarantee is that the big syndicates don't let anything in to their own models until it has proven its worth in the real world.