Good Books & Bad Correlations

Have been re-reading Nick Hornby's Fever Pitch, which I first read some six or seven years ago. We'll forgive him for being a gooner because it's a fantastic read and really captures what it's like to be a football fan. I'd thoroughly recommend it for anyone who has ever followed their team in a serious way. And by that I mean actually going to see your team play on a regular basis.

There is a small passage in it that had me laughing out loud in recognition, and it concerned the very first match he was taken to as a boy, and his reaction  to it:

I’d been to public entertainments before, of course; I’d been to the cinema and the pantomime and to see my mother sing in the chorus of the White Horse Inn at the Town Hall. But that was different. The audiences I had hitherto been a part of had paid to have a good time and, though occasionally one might spot a fidgety child or a yawning adult, I hadn’t ever noticed faces contorted by rage or despair or frustration. Entertainment as pain was an idea entirely new to me, and it seemed to be something I’d been waiting for.

The book follows him as a fan during Arsenal's double-winning season, so for a Spurs fan this is somewhat gauling, but it's very well written and has a great deal in it for fans of any football team.

Okay, so what else? Well, I'd like to talk about correlations, and how we must remind ourselves that correlations cannot be considered proof. Indeed sometimes correlations can give us completely the wrong picture. You all know what I'm talking about: "Ah, this has happened, therefore that must mean so and so". Or in other words, "...because of this, that".

Let me give you example of a correlation leading to an incorrect conclusion:

An old man goes to the doctor to complain that he is no longer able to satisfy his wife in bed. The doctor checks him over and everything seems fine physically, so he says to the old man:

"This is going to sound a bit strange, but I want you to make love to your wife in the normal way, but have someone standing over the pair of you waving a small towel. This will help you to satisfy your wife."
 
The old man thought this was very strange, but agrees to try nonetheless. He meets a young guy in the pub, who agrees for a small fee to be his towel-waver while he makes love to his wife. So he takes the guy back home, gives him the towel to waft and then jumps on his wife and starts banging her. He toils away, working up a sweat, but as much as he huffs and puffs, his wife remains unmoved, not an orgasm in sight.

Exasperated, the old man gets up off the bed and says to the young man:

"This is just not working. Look, I'll waft the towel and you have a go!"
 
Well, of course the young man obliges, he jumps on the old man's wife and starts jack-hammering her. Shortly after, the old man's wife starts screaming in ecstasy, her body arching as orgasm after orgasm rocks her writhing body. When the young man and the old man's wife are finished, they lay together on the bed, utterly exhausted.

After a small pause, the old man looks triumphantly down at the young man, and says, "There you go, sonny! Now THAT is how to wave a fucking towel!"

... Okay, I apologise. This may have just been an excuse to tell a rather lame joke, but it does also highlight how certain correlations can give us the wrong end of the stick. In football, we do this all the time. A team has been involved in high-scoring games in the last ten matches, so we automatically assume that the next game is going to be high-scoring also. Man Utd sit top of the league and Blackburn sit on the bottom. Man Utd have a great home record, so of course Man Utd must win (this last example is particularly relevant to me!).

I'm a great believer in stats, as even the most basic (goal supremacy) and well-known (ELO) algorithms can help to improve your hit rate above the average. But stats alone won't do the job as they can lead you down the exact same path as the old man with the sexually dissatisfied wife. A good run must inevitably come to an end, but largely these basic algorithms won't tell you, or even forewarn you when this is going to happen.

So what's the answer? Well, I suppose it's to be aware of correlations, to distinguish them from facts and to use as many different and disparate sources of information as you can.

...Oh, and also to beware of young men brandishing towels.