Saturday, 14 July 2012

How To Rate the Ratings

I've frequently discussed various ratings methods on this blog, and ratings in general have been talked about, picked over, criticised and praised at different times and at different places on a multitude of betting-related websites over the years.

As you probably know, I am a fan of ratings and statistical analysis - in my view they are a very useful aide, a handy addition to a bettor's armoury in helping to increase profitability, even if they are not the definitive answer. Even on their own, a decent set of ratings will dramatically increase someone's hit rate above random guessing... But there is an issue with a lot of rating methods that often goes unnoticed, or is not discussed. In the worst cases, the problem is even deliberately ignored.

So both for advocates of rating methods and also for those of you that feel they are, ahem, over-rated, today I'm going to discuss one of their shortcomings, or perhaps I should say one of their potential shortcomings when they are used in a particular way.

I suppose there's no more classic example of a ratings system than Elo, which is perhaps the best-known of them all, and perhaps one of the most heavily employed. As many (most) [all] of you will know, this was a system first developed to rate chess players, but which has subsequently also been heavily used to rate two football teams playing against each other. This converted Elo method is often known as Rate Form.

It's not my intention here to review exactly how the Rate Form method works. Most people already know how to implement it, and if they don't then it's easily found on the web. But we'll use Rate Form here to demonstrate a common obstacle that many rating methods face when using it for football betting. It's certainly a surmountable problem, but I suspect that it's not surmounted by everyone out there.

Rate Form might throw out figures like this:

Man Utd (RF rating 1,745) v Aston Villa (RF rating 987)

Please note, I have just dreamed up these Rate Form figures out of thin air just to demonstrate the point. They are not real values. The idea with these Rate Form ratings is to subtract one from the other, and perhaps add-in a home advantage value to arrive at a final figure.

[Man Utd (1745) + home advntge (100)] - Aston Villa (987) = 858

For the moment, we'll ignore the rather crude way in which a straight 100 points is popped on for the home team's advantage as that's not my focus here. Disregarding that commonly-used "tweak", we have nonetheless arrived at a ratings total of +858 points.

"Aha!", people will say, "that's a home win. A final figure of +858 points is significant, so we can plonk some money onto Man Utd because they're going to win."

Erm, no I don't think so. It's right here that the problem I'm talking about appears. Just what ARE 858 points? What is the exact unit of measurement that we're discussing here? Is it 858 sausages? Is it 858 elephants? They may as well be elephants because as things stand, unless we can successfully convert these points into match odds and compare them against the odds on offer at the various bookies, then we have absolutely no idea if there is any value here or not. We have to find the correlation or the rating is utterly useless.

I do hope you can see what I'm banging on about here, because I've had so many discussions with people about ratings, and I'm always somewhat astonished by how many people arrive at a points value in the various methods they employ - and then go blindly ahead and strike bets based on that implied value. To my mind, they have done very well to build-up all their ratings in the first place, but then have fallen at the final hurdle because they haven't used those ratings to then generate match odds from them.

Rate Form, goal supremacy, Game Form, Power Ratings, Score Ability, shot on target, corners, number of farts per match: I don't care what ratings systems you employ, they can probably all be used to generate match odds from their individual ratings - but if they can't be transformed, then they should only be used as a general guide to what may or may not happen in the match, but certainly not used as vital information on which to place bets.

Okay, Mr smarty-arse, high-and-mighty, condescending toss-pot, how exactly do we convert all our lovely ratings into match odds?

Well, all will be revealed shortly. It's my earnest intention to continue with my intermittent series called "Compiling Match Odds" in my very next post, and when I do I will endeavour to show exactly how ratings, such as Rate Form, can indeed be used to generate match odds. There is no real mystery behind it and certainly no magic, but hopefully it will allow some of you to convert your ratings into something that you can compare with the real live odds on offer... and afterall, that's what it's all about, isn't it?


  1. You are a bloody tease, Eddie!

    Looks like I've a little longer to wait for my simpleton's explanation of MLE, and now am also looking forward to the next rateform analysis.

    But your efforts are appreciated, certainly by me, and I suspect, by the silent majority :-)

  2. Hah, sorry Dave. It's not my intention to tease you, rather a case of laziness on my part. Often I don't have any idea what I'm going to write and then suddenly I realise that I've bitten off more than I can chew.

    Made a bit of a rod for my own back with this one, but I'm a good way through writing the next post on match odds, so I hope to have that up soon enough.


  3. Another great post. Can't wait for the final post connecting all the dots :-)

  4. Hi Eddie, really enjoying your posts!! Just tried to click link for latest post compiling match odds part 3 but page can not be found. Please can you help not sure if it is me or it's really not working. Cheers for your help

  5. Hi Keith

    That's my fault. I accidentally posted it up before I'd finished it, so I've removed it for the moment. I'll try and get it up tonight or tomorrow.



Note: only a member of this blog may post a comment.