Book Review: The Signal and the Noise by Nate Silver

Over the semester break I spent some time ploughing through books which were on my to read list. One of them was The Signal and the Noise by Nate Silver.

I kind of like Nate Silver’s writing, and I especially like his analysis but I had started the book, gotten half way through, got distracted and only picked it up again in January. So the review is more or less “I seem to remember this was fascinating” and “the content of this book should be fascinating but I’m not really sure I like it any more.

I like numbers. I like playing with them. I like manipulating them. I’m not very good at them; I don’t have many regrets in life but a maths and languages course up front might have been a better choice when I was 17 rather than pure maths.

I like that there is an increasing recognition that there is meaning in numbers and that the meaning needs to be interpreted. In many respects, that’s not that different to languages anyway. There is meaning in words; it has to be extracted; interpreted.

So to Nate Silver. Yes, he got the polls right in the last few US elections, and yes, he’s doing the start up thing with Five Thirty Eight now.

The focus of the book, to some extent, was the art of prediction, and his dependency on Bayes. It featured some case studies – baseball and gambling are included (although I really do suggest that you have a look at MoneyBall if you’ve any interest in the application of statistical inference and prediction to the baseball numbers as it’s a better read on that front). There was a section meteorology which was fascinating. A key point which he raises is perception and what people want from a weather forecast. Is it a weather forecast, or some entertainment?

One of the stories in it which fascinated me related to Deep Blue and the chess match with Garry Kasparov. What particularly interested me there was the idea that the computer behaved in a specific way, based on a bug. But the way it behaved rattled Kasparov and  caused some investigation as to what the long term outcome of that move could be.

I’m interested in machine learning so this is something which would catch my attention in a lateral way. We train computers to make decisions; sometimes it is not clear whether a given decision is based on a bug or some aspect of the training.

However, a couple of things annoyed me about the book. The Kindle edition has a frustrating number of typos. I can understand this in a scanned book I just think it’s a bit unforgivable now. And there are a lot of elements of the book where Nate Silver assumes he is writing for a uniquely US based audience. I don’t think this was ever going to be a safe assumption for him.

A couple of sections of the book fascinated me in a way that led me back to subject specific books of which one is earthquake prediction – we just aren’t good at it at all at the moment. As it is, I have a more than passing interest in earthquakes, volcanoes and rogue waves so which I finished this, you can make an approximate guess what other books were on my reading list.

I’m inclined to say that The Signal and the Noise is a fascinating book and well worth reading. But it’s difficult to grade in terms of is this a five star read, is it four or is it just average. I’m inclined to classify it as a book you should read, but be aware that it’s not a perfect reading book; there are elements of it which might annoy you. And you could skip it if you were so inclined. It is the sort of book that should help your Trivial Pursuit score and will open your mind. Oh and you’ll probably be left with the impression that Nate Silver is brighter than you are which isn’t always the most edifiying either.