The Signal and the Noise: Why So Many Predictions Fail–But Some Don’t. Nate Silver. Amazon
Silver’s fame as a statistical analyst and election prognosticator at FiveThirtyEight have thrust him into the limelight (to the applause of us kindred nerds). This book is an explanation of the philosophy behind his analysis and predictions.
This was a fascinating book to me–the subjects discussed range from weather forecasting (which has remarkably improved in the last 20 years or so) to Garry Kasparov vs. Deep Blue to stock markets to foreign policy, and more.
Bayes’ theory, which Silver uses and advocates in the book, emphasizes probability or “percent chances” of predictions, as opposed to making absolute declarations. The theory has as its foundation the principle of humility (whether or not it’s explicitly stated as such); make predictions, learn from what actually happens, and refine your models accordingly.
Admit what you don’t know. Learn. Don’t disregard data or results you don’t like. It’s actually refreshing (as well as interesting).
You don’t have to be a nerd to appreciate or enjoy the book (although it helps). Silver is actually a funny and insightful writer, and he casually throws in pop culture and obscure references to serious discussions. I certainly appreciate his sense of humor.
Probably the best summary of the book I can give (without giving away the book in a sense) is his discussion of Phil Tetlock’s distinction between hedgehogs and foxes. These are two kinds of thinkers: hedgehogs (popular TV pundits, columnists, bloggers, etc.) are specialists who have dug their heels in on their position; they insist that the world operates as they articulate and stubbornly refuse to hedge or alter predictions in the face of new data. Foxes, on the other hand, tend to be multidisciplinarians who adapt their predictions and approaches based on new data or additional expertise. Foxes are extremely cautious about their predictions, rarely if ever speaking in absolute terms.
The Signal and the Noise is a book that defends a foxy approach to analysis and prediction as opposed to a hedgehog-y approach. (It’s really more intriguing than it sounds.)
Kindle rating: 4/5 stars