Wednesday, February 8, 2017

Prognostication Debunked

Boris Zelkin writes a long column at American Greatness, a pro-Trump webjournal, about how famed statistician Nate Silver is getting too many important political and athletic calls wrong. I won't summarize Zelkin's arguments, you can read them if you choose.

Here is my take on statistical modeling. It is based on increasingly sophisticated projections of past trends. Truly, in our only semi-chaotic world more often than not past trends continue. That's the value of what Silver and others like him provide.

While projecting existing trend lines is often accurate, it is no great feat. The really tough predictions are those which identify before it happens those points at which trend lines are going to change direction, going to perform in ways they have not previously or not recently behaved. In short, spotting inflection points.

Examples: the person who, in the midst of a bull market rally, correctly calls the timing of the next market crash (c.f., Martin Zweig). Or the person who spots that what has worked in the last several presidential elections will prove inferior to a new, never-before-tried strategy a la Trump.

Statistics doesn't give us great tools for identifying inflection points. Intuition is probably better, if not by much. Unfortunately, as Wall Street wisdom holds, intuition predicts far more inflection points than actually occur. There's always someone saying "this time is different;" only infrequently is it true.