Wednesday, January 30, 2013

The Signal and the Noise by Nate Silver

If there were a Wonk of the Year award, Silver would have won it in a slam-dunk last year -- this book was a bestseller, his political predictions came true almost perfectly, and he was all over the media (not least at his own, New York Time-affiliated FiveThirtyEight blog). I wouldn't be surprised if he'd also predicted the entire baseball season -- his other main claim to fame in the fields of wonkery was in devising a swell algorithm to predict the usefulness of players -- but I've been disconnected from sports for so long that I have no idea.

So: he's one of the current top Big Explainers, and The Signal and the Noise is his first book -- his official Big Explanation. Since it was designed to be a big bestseller, there is absolutely no math in it -- though Silver is a statistician, and his analyses rely heavily on Bayesian methodology -- and it, as it must, attempts to reduce all of life to one thing. (Oddly, for Silver, that thing is an equation, which is hard to do in a book with no math.)

Every single Big Explanation is wrong, with no exceptions, so this one is as well. Oh, it's pretty good, as Big Explanations go -- quite useful, in the right places, and a good tool for looking at a lot of situations in the actually existing world. But a book like this must insist that its Big Explanation covers everything in the world, and so Silver does, and so he's wrong, because nothing ever can do that. But his claim is elegant and not too obviously self-aggrandizing, so you can't stay grumpy at him for long.

If you know what Bayesian statistics are, you don't need to read The Signal and the Noise, only to know that Silver applies Bayes to baseball and politics, poker and weather forecasting, climate change and terrorism and the stock market -- all of which involve numbers and frequencies and lots of statistics, so they're fertile ground. If you only vaguely recognize Bayes -- if, like me, it's familiar while you read it, like the laws of thermodynamics and the carbon cycle, but slips out of mind immediately afterward -- then The Signal and the Noise will be pleasant and may make you feel quite smart. If you detest numbers and prediction, because the lord of the universe explained everything in this book you have right there, then you need to go sit in the corner while the grownups talk.

As long as no one takes Silver's Big Explanations absolutely seriously, it will be quite useful -- and thinking about probabilistic calculations in more situations would be a net positive for most of us. But I'm sure there will be a cult of Bayes -- like the Milton Freedmanites, I suppose, but more fond of brackets -- that insists that all of human life can and will be predicted. We always do have the stupid with us, though, so we can't blame Silver (more than mildly and half-heartedly) for that.

The Signal and the Noise has quite a lot of good thinking, some good tools, and an organizing principle that's vastly more correct than most similar books. For a non-fiction bestseller, this is about as good as it gets.

1 comment:

Argonaut said...

Speaking as someone who works with statistics, there very much is a "cult of Bayes," though it's mostly restricted so far to fairly scholarly cultism. I've been following some of the academic bloggers talking about this book, though I haven't read it myself. Interesting stuff, and I absolutely agree with your general point: Whether Bayesian or "frequentist," more people having a grasp of the fundamentals of probability would be a boon.

Post a Comment