by Snowball »
15 Oct 2010 20:32
and BTW, there are stats "after the fact" and there are PREDICTIONS.
I predicted that Shane wouldn't score as many goals playing 4-5-1
Stats and Causality.
I am totally aware that correlation doesn't mean causality.
The classic is example is the number of Bishops in Northern
Ireland correlating with prositutes in Newcastle.
But basic experimental processes are different if you PREDICT the effect and which way it will go.
If I say, "Harte & Zurab being in the squad will lower the average number of goals conceded"
the degree of difference required is instantly halved (as I'm predicting in one direction.)
Of course it is true that it might be nothing to do with them, and I've taken a random two
players and it's pure coincidence. But "non-statistically speaking" if we were in the pub and I said,
"Zurab is a class act, better IMO than Mills or Pearce, and he'll increase competition. Harte is serious
class and has top-top experience and will bring a lot of stability to the defence despite his age and slowness
(but I expect him to get ripped every tenth or so game.) I predict we'll start letting in less goals and we'll
score a few more free-kicks and convert almost all our penalties."
That's a prediction. All we need to say my prediction came right is a significant difference in goals conceded,
and if we were arguing in the pub eight weeks later, we would all naturally look at goals-conceded.
That is, even if no-one uses the word, we search for statistical evidence to measure the accuracy of our predictions.
And YES, it MIGHT be that Griffin has improved, Pearce/Mills have played their ten best-ever games, Federici
has suddenly become world-class. It's POSSIBLE. But I made the prediction in this fictional pub, and it came right,
and there are sound reasons (their ability and the fact that they played, for example) to suggest that they ARE
the reason for the change in fortunes.
