Developing a Rating System 2

Direct and Indirect Comparisons

Well, I hope you're managing to collect some data. We'll look at how to organise and analyse that information a bit later, but first I want to talk to you about pair-wise direct and indirect comparison techniques.

Wow, that sounds complicated - I don't think I've ever heard of those before.

Perhaps not using those terms, though I'm sure you use those very techniques every single time you place a bet!

Let's take an example from Boxing. Suppose you're looking forward to investing in a fight about to take place between Mr Aggresive and Mr Basher. You'd probably want to see if they've ever fought before. Perhaps you find out that just last week Mr A knocked out Mr B in the 1st Round. Well, I'm not a boxing expert, but that's probably very good information to know. That is an example of a pair-wise direct comparison between Mr A and Mr B.

OK - that's obvious enough. I'd use that information myself, and I'd certainly want a rating system to use that kind of data. So what about indirect comparison?

Well, let's say that later on the same card Mr Crusher and Mr Demolition are going to have a professional disagreement, but that they've never fought before. That means there just is no way of directly comparing each other's strengths through a previous performance. We'll have to find a way of comparing them indirectly.

Aha - I see this now. This is like a form-line. Perhaps previously they have both introduced themselves to the same boxer, at different times of course!

Yes. Maybe we know that Mr C gave Mr A a thorough pasting, but also that Mr A walked all over Mr D. Although they've never met previously, I'd suspect that Mr C has the upper hand over Mr D.

I understand this, but it's all getting very complicated. What, with all these Mr Aggressive's and thorough pastings…

But I'll admit that you're right. That is the way I like to look into a bet, by looking at the form. But I certainly don't look into all the possible direct and indirect comparisons before I make a bet.

Maybe - but perhaps you would like to, and that's the kind of thing that attracts some people into automating the whole process. By using a ratings system, what we're attempting to do is use all of that kind of information in a precise mathematical way.

In addition, if you can find a good metric, that is if you can find a way of putting a number to phrases like 'a thorough pasting', or 'walked all over', then you've potentially got a very good ratings system.

OK - so the whole process is really just a way of trying to formalise what I do anyway when I make a bet. But there's much more to it when I make my own selections, for instance I might suspect that Mr A didn't really try against Mr C, and that Mr D is a southpaw and that…

Hang on!! All I'm going to say right now is that there are a number of ways of dealing with these problems, some quite effective, others less so. But it's fair to say that at the end of the day, your rating is just going to be a measure of something that you have defined. It will never be a perfect measure or predictor of sporting performance. Though if you construct your own ratings, using sound mathematical techniques, then they can certainly act as a good guide. You will know what they can do, and what they can't. You will also have the benefit of being able to back-test your theories and have a certain confidence in any future bets.

Read other installments:

 Part One| Part Three