The signal and the noise: why so many predictions fail -- but some don't by Nate Silver

Started: May 21, 2015
Finished (ish): 4 July 2015
Pages: 457

There is a genre of book that I tend to pick up in airports.  No, not thrillers or romances.  Intellectual books of various kinds that pique my interest by addressing something that's top of mind for me.  Or at least,  they create a top-of-mind subject  by virtue of being right in front of me at a time when I know I'm going to be trapped in a chair for hours in the immediate future...albeit without quite enough oxygen for full mental processing. But still.

I picked this one up as a result of a recent visit to Saskatoon.  The blurb appealed to me, and work and workish books were on my mind, particularly questions about work-related predictions.  I told myself that the book would be interesting, and perhaps even useful.

And it was interesting, if nowhere near as profound as the cover blurb promises. "Could turn out to be one of the more momentous books of the decade."  Um...no.  Everything momentous in this book is more ably and rigourously covered in Thinking: Fast and Slow.   And at least some of the interest of the book comes from factors outside of the direct content of the book itself.  But that doesn't mean that the book was dull or pointless.  Just not profound.

What do I mean by all of that? And what is the book about?

Nate Silver famously made a 100% correct prediction of the results of the 2012 American presidential election, at a time when pundits were calling the election too close to call and pollsters were suffering public angst about the present and future of political polling.  He parlayed that success into this book, which discusses the art and science of prediction by examining its successes and failures in a number of different fields.

What did I learn from the book?   The coolest thing is that for the first time, I feel that I have some understanding of  Bayesian statistics.  Thank you Nate Silver.  I've heard Bayesian statistics discussed in passing many times, without really knowing what was meant.  Silver not only explains what they are, he illustrates how they are used and why they are significant.  It's not like Silver invented Bayes Theorem, but good communications skills matter, and his explanation is a useful contribution.

I also learnt the current state of a number of fields, like weather prediction and earthquake prediction, which is kind of interesting, if not life-changing.

But part of the interest of the book comes out of my 'meta-learning'. Silver talks about himself and his own experiences with prediction throughout the book.  Don't blame Silver: it's a modern non-fiction book style to write about yourself while you write about your subject. But one of the distracting side-effects of this style is that I felt that I was learning about Nate Silver as I read, and what I primarily learnt was that Nate Silver is smart, opportunistic, and ultimately kind of shallow.

No, I'm not sure why that really matters.  But Silver has apparently built his career by learning something about Bayesian reasoning and the kind of decision-making fallacies that we are all prone to (as described in Thinking: Fast and Slow), and then finding fields that are under-serviced in terms of the rigorous application of these principles to prediction.  He's helped develop a program that took a new approach to predicting the performance of baseball players, he made a bunch of money in the great internet poker boom in the mid-2000s, and after that crashed he focused on poll-based political predictions.

So, while the book is interesting, it isn't profound or life-changing, or really applicable to ones' own life circumstances.  A 7 out of 10, maybe?