The Signal and the Noise: Why So Many Predictions Fail-but Some Don't by Nate Silver
Published in September 2012
How can we possibly review Nate Silver's The Signal and the Noise after the election? Silver is perhaps the most overexposed pollster of all time. His NYTimes FiveThirtyEight blog was a must read for many of us during the run-up to 11/6/12.
Try not to let Nate Silver election burnout dissuade you from reading his excellent book.
The Signal and the Noise has surprisingly little political content. This is not a book about political polling, or why changing demographics have put the the Republicans at a long term Electoral College disadvantage. Rather, this is a book about forecasting.
The genius of Nate Silver is not that he called the 2012 election. What is so great about Nate Silver and The Signal and the Noise is that he manages to turn any approach to forecasting that is not probabilistic into an anachronism.
Probabilistic thinking is hard to do and hard to communicate. We crave strong statements, decisive conclusions, and confident predictions. We confuse confidence with expertise, certainty with accuracy.
We had a hard time getting our heads around what it meant when the FiveThirtyEight algorithm, which was built with inputs from a range of polls, told us on 10/31 that the probability Obama winning the election was 79%. If we were an Obama supporter did this 79% figure mean that we should worry less about the election? At what probability should be stop worrying?
Why can't Silver just make a firm prediction?
The Signal and the Noise is a book length answer to that question.
You may not have known that you need to understand Bayesian probability theory to evaluate the forecasts of others, or make better forecasts yourself, but you do.
My favorite chapters in Silver's book were about his previous life as a high stakes online poker player. He was never quite as good as poker as at political forecasting, but he wasn't bad. Despite his winnings in online poker, Silver is quick to point out that he benefited from a large number of bad players (non probabilistic thinkers) rushing into the online poker world before the government cracked down. Anyone thinking of become an online gambler, or a day trader, would be wise to read The Signal and the Noise before making this career choice.
I also enjoyed Silver's chapters on the improving science of weather forecasting and the (almost impossible) task of earthquake prediction. His writing on baseball was of less interest to me (being a casual fan at best), but his ability to illustrate complex statistical thinking with examples from sports (and sports betting) is impressive.
The Signal and the Noise is on my shortlist for one of the best nonfiction books of 2012.
What are you reading?