I recently read Nate Silver’s popular book, The Signal and the Noise. He’s the guy who correctly predicted the winner of the election in all fifty states. The book is about how people have attempted to predict various things – the stock market, elections, earthquakes, and so forth – and how they’ve both failed and succeeded in doing so.
The book is largely made up of examples of various prediction systems people have made in different areas. Silver writes in an informative yet entertaining style. Even topics I usually find boring, like politics or economics, he was able to describe in a way that had me engaged and wanting to learn more.
The second half of the book goes into how to make better predictions. A large part of it rests on Bayes’ Theorem, which is a basic law of probability that you would learn in the beginning of an introductory statistics course. And yet, he never actually states the theorem in terms of probabilities of events. I found this a bit frustrating, despite knowing what the theorem is, because he frequently refers to it as being an important tool in making predictions (contrasted to the frequentist approach, which he describes as inferior, for which he also includes no equations).
He even goes through some examples where he gives numbers for x (the prior probability), y (the conditional probability given the hypothesis is true), and z (the conditional probability given the hypothesis is false), and then he magically plugs them into the formula xy/(xy + z(1-x)). Well that works, but I really think it would’ve been more effective had he expressed it in terms of events A and B and conditional probabilities in terms of those events, rather than having to memorize what three letters really mean. But yeah, that’s just a gripe from me as a statistics major.
There’s a lot of interesting footnotes that I wish he had just included in the text at the bottom of each page, instead of putting them in a separate section at the end of the book. I kept having to flip back and forth between the main text and the notes section (and some chapters have >100 footnotes), only to see that most of them were just references to other materials. The non-reference footnotes would’ve been much better as actual footnotes in the text. Again, minor gripe.
It’s sometimes a tad repetitive, the message of prediction being a difficult problem that can’t be approached by just throwing numbers into a black box. But the anecdotes are interesting, Silver’s commentary is engaging and occasionally humorously snarky (“New Orleans does many things well, but there are two things that it proudly refuses to do. New Orleans does not move quickly, and New Orleans does not place much faith in authority.”). I enjoyed the chapters about poker and chess, since those were more “fun” examples involving games that I think makes the more abstract talk of prediction a bit more relatable.
So even though I’m not usually much of a nonfiction reader (Underground and Moneyball are really the only other nonfiction books I’ve read recently), because it was entertaining, I zipped through this 500+-page book in much less time than it’s taken me to read a lot of shorter fiction books.