Monday, July 15, 2013

Guest Reviewer Jeff on "The Signal and the Noise"

Prophecy: A Review of Nate Silver’s 
“The Signal and the Noise”

Part 1 - Introduction

This may be one of the best books I’ve read in the past several years. For this reason, and the fact that the book is quite long (534 pages) this review is also long, and I have chosen to break it into four parts. In my defense, I must say that there is more content (and thinking) in this single book than in the last 10-20 books I have read combined. This book is also a real paradigm-shifter.

I should clarify “best” here. For me best books (I’m an equation-focused physicist and therefore a slow reader) are defined more or less by me learning the maximum new stuff - useful stuff - in the shortest amount of time. For a deaf person, life consists of learning out how to pick up visible clues - lip-reading, for instance. For a slow reader, life consists of figuring ways to wade through fluff to get at substance. In Silver’s take on the universe, we live in an extraordinarily data-rich environment: the amount of data becoming available is increasing at exponential speeds in our modern world. It is increasing so fast (the US government now reports monthly on 45,000 different economic parameters) that we can barely even store it today. However, this translates to the fact that we are also all drowning in noise. With experience, I’ve figured how to get the relevant content of most books by just reading the title, the table of contents, the figure captions, and skimming the conclusion. Most of the rest is mostly fluff - like Styrofoam package filling in a box holding a single DVD. In my mind (even before reading this book) I have tended to think in terms of signal-to-noise ratio - what’s the “fluff percent” in something? In the book world, content vs. words seem to follow a power law (best seen on a log-log plot): there are millions of books with little content, a very few books with a LOT of content, and the rest fit on a logarithmic line in between. Silver’s book fits on the very “high” end of that power-law curve. Put another way, this is an unusually cost-effective book for a physicist to read.

This book concerns just one thing: prophesy. That word does not appear even once, because this is a calculatedly secular book, written by an individual who is an acknowledged gay atheist. Mercifully, Silver doesn’t try to flog either straights or believers with his proclivities, however (he’s my favorite kind of atheist). Nevertheless, throughout all human history, mankind has yearned for foresight: to know what is going to happen in the future. Where do I place my bets? What do I have to watch out for? What will kill my children? The Old Testament had its prophets, but those ceased with the death of the original apostles. If that doesn’t make sense to you, then you are a “dry” (not yet baptized) Mormon, because Mormons are that rare group in the modern world that believes in living prophets.

In short, Silver succeeds amazingly in showing when predictions work, when they don’t work - and most importantly, why they do and why they don’t. He draws heavily from his own life experience, mainly baseball, poker, and politics, but then fearlessly ventures into fields that are quite far removed from these. Silver is famous (and heavily sought after) for precisely predicting the outcomes of the 2008, 2010, and 2012 American elections. By precise, I mean that he predicted/forecast successfully the outcome of both presidential elections in 49 of 50 states. He made Gallup and other high-investment pollsters look like a bunch of high-school rubes in the process. He showed why the Talking Political Heads on TV (e.g., the McLaughlin Group) have had an abysmal success rate in political forecasting. In fact, their forecasting effectiveness has been significantly worse than the 50%-50% that you could expect from a coin-toss. Why bother to waste the electricity (and suffer the numbing ads) by watching them? Just flipping a coin would put you ahead of these political “experts.” Silver also explains why they fail so badly: fundamentally, they are pandering to a particular audience. Certain people gravitate to Fox “News” or MSNBC NOT for useful information (including news), but to feel good. If a substantial and predictable audience feels good, then the advertisers have hit their gold mine.

However, Silver also draws from a startlingly long list of experts in nearly every sub-field that involves forecasting. I can’t be certain of this, but it appears from his footnotes that he has paid for airfare and hotel rooms to travel for years back and forth across the United States to interview upwards of 500 individuals - just for this book. The interviewees range from people who you and I have never heard of to people like Donald Rumsfeld, the Secretary of Defense during 9/11.

The title of the book is “The Signal and the Noise.” An overlying thesis is that if we cannot sort between these two extremes, all our forecasting (all our understanding) is totally wasted and ineffective. If you develop your forecasting algorithms and apply them to noise, you will get garbage. Your predictions or forecasts will consistently fail. Massive amounts of thought and funding have gone into predicting the stock market – repeatedly over time – and all failed.

Implicit in this is that an understanding of how a system works is absolutely crucial... even if that understanding is demonstrably imperfect. (If that understanding is wrong, then all bets are off, literally and figuratively). As we will see later (in Bayesian thinking), this understanding (or the heuristics, as Silver calls models) can be steadily improved: made more and more consistent with the real world.

Some systems - baseball and chess, for instance - have very clear rules, and thus should be easy to understand. There are also readily-available and truly vast databases for each of these games. Theoretically then, if we understand the way things work AND have sufficient data, we should be able to forecast how things will play out - how to win the game, how to predict the hurricane. We should be able to prophesy what will happen in the future. This should especially be the case if we have a lot of money to throw at it, right?

It doesn’t turn out quite that way, and this will bring us back later to the book’s title.

Note from Madame L: Look for further installments of this review in the coming days.

No comments: