Quite possibly the most interesting book I have read in sometime, Nate Silver's The Signal and the Noise is a wide ranging work which revolves around one theme: what can people accurately predict? It turns out very little. In fact, this work is replete with examples of how falsely optimistic experts are about their forecasting ability. Silver’s overall goal is to work within the parameters of what is possible to predict, and give the best possible approximation of how this can be done.
Certain events are fairly easier to predict than others. Baseball, one of Silver’s favorite examples, is a game played under proscribed circumstances, with the rules well-known, and the statistics of the game results recorded for many decades. In this data rich and rule laden environment, forecasting is far easier than, say, in more complex systems. Earthquakes are sometimes anticipated by minor quakes, but sometimes they aren’t. There simply is no way to adequately predict the behavior of fault lines many miles under the earth. They are too complex for our models.
Yet complexity isn’t the only measure of our ability to predict. Silver claims that meteorological predictions is one of the great success stories of this book. The atmosphere is a complex system, but it is governed by simple processes that can be observed and recorded. In the last 30 years, with advances in computer modeling, forecasting has improved dramatically. Weather prediction, often the butt of jokes, is actually a very successful forecasting method.
All in all, Silver’s book is a paean to Bayesian method of forecasting. Created by Scottish clergyman Thomas Bayes in the seventeenth century, “Bayes' theorem is a formula that describes how to update the probabilities of hypotheses when given evidence.” That is, it allows us to further change and refine our predictions based on new evidence. A common sense approach, but one that has not always been taken.