I admit it… I’m a Nate Silver junkie. I love his stuff. He’s a geek. I’m a sucker for people who use data. I read Nate’s FiverThirtyEight blog all the time. I was looking forward to this book.
And this book didn’t disappoint. This book is about predictions, which Nate does for a living, and why some are good and some are bad. He covers so many topics in this book. How weather prediction works (and how it has gotten much, much better), how earthquake prediction works (and how it hasn’t gotten better and probably won’t), and his life as a professional poker player during the poker bubble.
The overall theme of the book is how one should make predictions. The “signal” is the thing you are looking for – the thing you want to measure, model, and use to make predictions. The “noise” is all the other stuff that also happens that could be an example of the thing you are looking for, but actually are things that are just random occurrences.
Because this book covers how to separate the two, Nate does use a lot of math. It isn’t incredibly complex math, but if you really, really, really hate math, you might get lost and bored. There are a lot of graphs in the book to make the math easier, and if you can just power through some of the text he uses to describe the math and just use the pictures, you’ll be fine. But for some folks, this just isn’t their cup of tea – you might feel like you are back in high school.
Having said that, I love math, and enjoyed the descriptions. It gave me a new appreciation for how weather forecasting has gotten better over time, and why it might be impossible to make earthquake predictions, beyond some basics (such as, there is an x% chance of a magnitude 8 earthquake in the San Francisco region within the next Y years).
The earthquake prediction stuff was actually very fascinating. This was a case where people ended up creating very complex models based upon past historical data, and used that to make some very specific predictions. And, some of these predictions happened, making the people who created the model look like geniuses. However, almost all their other predictions failed. this was a case of people creating a model based upon the “noise” (past earthquake s) and thus their model was overly complex and mistook the “noise” for the “signal”. As a guy who has to spend his days predicting what we should do next based upon complex historical data and current data points, I really, really appreciated this. Not only do you have to know what the data is to use, but you have to be able to identify the data not to use.
The pièce de résistance of the book is the end. After showing past examples of predictions that worked and didn’t, and using Bayes theorem (which he describes in great detail – remember how I said if you hated math you might have a problem with the book?) he tackles climate science predictions. This was fascinating. He acknowledges the critics of climate change, and how there is good science and math that can be used to show how some of the predictions are off, but then shows how accurate the science behind climate change is, and how many of the predictions have been true.
Once you have the past data on how weather predictions have improved, and earthquake predictions haven’t, it is very fascinating to then look at climate science predictions in this light. Very, very cool.
So, my general recommendation is – if you like Nate Silver, and/or you like math, you will love this book. Even if you aren’t a math geek, but you are a logical thinker, you will probably really like this book. if, however, you want to stay away from math as much as possible, you probably won’t care much for this book.