A Strong Signal

In the months leading up to this last presidential election, I checked the FiveThirtyEight blog at least once daily. Nate Silver's considered analysis seemed like an island of reason amid a sea of insanity. And his methods were proven out on Election Day, when his forecast was correct for all 50 states (improving on his nearly perfect record in 2008).

The election may be over (thankfully), but there's still a solid dose of Silver's insight to be had in his book The Signal and the Noise: Why So Many Predictions Fail — but Some Don't. It's an extremely intelligent analysis of forecasting, and covers a wide range of topics, politics representing only a piece of the whole. Silver also examines predictions in a wide variety of other topics: economics, baseball, weather, earthquakes, infectious diseases, chess, poker, global warming, and terrorism.

The extensive work Silver has done to build his knowledge in these areas is abundantly clear from his book. The endnotes section of the book is a full third of its total length, citing sources for every fact and figure he uses. He presents quotes from interviews crisscrossing the globe, providing insights from true experts in each of the subjects. And he packages it all in effortlessly readable prose.

Nate has a brilliant way with words, particularly the use of analogy. There's really just one core concept he's trying to highlight in the book, but he goes at it myriad ways, demonstrating both where it is helpful in making predictions, and where the absence of it has been detrimental. But no matter how many times he comes at the same issue, he has a fresh approach, a comparison both entertaining and easy to understand.

That core concept he's explaining is Bayes' theorem, a method of trying to realistically assess the probability of an event. Moreover, it provides the means to readjust one's analysis as new information is introduced (this being the crucial flaw in the many failed predictions he cites throughout the book). He posits a very entertaining notion of a "Bayesworld," a reality where every two people of differing opinions that come into contact with each other have only two choices: 1) to come to a mutual agreed revision of their opinions (one or both of them having been convinced to change their opinion by the other); or 2) to agree to a wager between them ("if you don't believe I'm right, then put your money where your mouth is.").

I had some familiarity with some of the concepts in the book, thanks to my own background in gaming. Nevertheless, Nate Silver still had some good things to teach, and a fun way of teaching them. He also provided an intriguing look at many fields in which I had little knowledge.

I found The Signal and the Noise to be a great book, and an absolute must-read for any statistical wonks out there. I give it an A-.