The Signal and the Noise

I am a schedule reader, which is usually nice because the hype has often died way down before I get to something and I can read it on its own terms. But then we get situations like this one, where the air has gone pretty spectacularly out of Nate Silver's balloon in the time since I bought the book. When his star was burning bright in the wake of the 2012 election though, everyone wanted to know how he looked at data to make his predictions. The insights he offers aren't especially unique or profound: generally, people are not as good as they could be at making predictions because of various biases, particularly confirmation bias, and also the tendency to not learn from mistakes and adjust models to take new data into account. He then sets about presenting various examples of prediction and its failures: political talking heads, weather, the 2008 recession, earthquakes, the trajectories of baseball players' careers, before spending the second half of the book writing a love letter to Baynesian statistics. It's much too long, about 450 pages before footnotes, and the prose style is fairly dry so it's very hard to stay interested. A big miss for me.