History is easy to explain. We can always come up with some story for why this or that occurred. And, if the story sounds plausible enough, the explanation will be accepted as true. But can we ever know that the explanation is correct?  How would we ever test it?

If we try to think like a scientist, as opposed to a politician, a theologian, or a used car salesman, we consider our explanation of observed behavior to be a theory. We then test the theory by drawing predictions from it. If the predictions (about as yet unobserved behavior) hold true, we may have some corroboration for our theory. (Of course, no number of correct predictions can prove that the theory is true.) But, it only takes one false prediction to disprove the theory.

Such is the asymmetric nature of our knowledge.

In forecasting, I like to complain about the worst practice of using "fit to history" as the sole consideration in forecast model selection. Any fool can fit a model to history, or even create a model with a perfect fit. But what is the predictive value of such a model? Can it forecast worth a darn?

The nice thing about forecasting is that we can test our models. We use them to generate forecasts of the future (e.g., sales of item X), and then observe how accurate those forecasts turn out to be. If they turn out to be fairly accurate, then maybe our "theory" is right, maybe we do have some understanding of the mechanisms underlying customer behavior. But if the forecasts turn out terribly wrong -- as they so often do -- just maybe we don't know WTF we're talking about.

The New Snake Oil: Golf Swing Analysis

So what about the "swing analysis" provided by professional golf commentators? I'm no fan of golf, and after watching a bit of the Pebble Beach Pro-Am on television this weekend, I'm less a fan of golf commentators. They are not very scientific.

Every once in awhile the commentators delivered a slow-motion swing analysis to explain why a particular shot went good or bad. As in the video link, the explanation (=theory) of the shot all sounds quite brilliant: The toe fanned open, the flatness of the wrist, the down-and-through alignment of the clubface, and the "fully released finish" (which is properly referred to as the "happy ending" when using your putter).

Of course, it is quite easy to provide an explanation for why the shot did what it did AFTER you've seen what the shot did. If these commentators wanted to impress me, they'd provide the swing analysis WITHOUT having seen where the ball ended up, and then predict where it would land. Then we could actually test whether their analysis was any good or, as I suspect, just plain silliness.