Quick! Your favorite 18th century English philosopher is . . . David Hume? Edmund Burke? Adam Smith or Bishop Berkeley? Don’t worry if one one comes to mind – Nate Silver, author of The Signal and the Noise: Why So Many Predictions Fail But Some Don’t, has a provocative suggestion: Thomas Bayes. Silver is a terrific statistician and a fine writer. His book won numerous awards and is becoming a non-fiction classic with sales continuing and an ever-widening impact. It was just named the book everyone at Northwestern University will read in the coming year.
Silver’s aim is to explain how predictions are made and why they are so difficult to get right. He mixes his own history with traditional hot-topic issues for prediction: gambling, weather, earthquakes, financial crises, and baseball. Silver has solid journalism chops, too, and his interviews provide further lens of analysis. Considering the explosion of available data and prognosticators clamoring for attention and support, informed human judgment about how to assess predictions and analysis is all the more essential. Supporting Silver’s arguments is an accessible account of statistics and quantitative literacy – the kind of description that would lead to lots of nodding heads if shared around a table.
Of course predicting the future is difficult. What makes it hard, even with advanced methods and powerful computers, are a cluster of tendencies, mistakes and methods. Silver makes it clear that there is no simple way to avoid them. Predicting the future cannot be reduced to an algorithm – even in closed systems. People are hard-wired to be overconfident in our ability to detect patterns. Silver cautions against any one ideal and instead advocates for understanding context, history, and causality as essential to making predictions and making sense of numbers. Good predictions should become better predictions over time as one learns more, adjusts, and the set of data grows larger. Silver endorses Bayes’s thinking and incremental improvements are at the heart of Bayes’ statistics.
Silver writes from experience. A University of Chicago math major, he followed his education with a consulting job that he hated and left after four years. To earn a living, he played online poker and obsessively studied baseball statistics, creating a better playing forecasting system called PECOTA. He sold it but remained active in the baseball statistical world, writing, editing, and forecasting.
In the run up to the election of 2008, Silver grew frustrated with the quality of political statistical analysis and decided that he could do better job. He created FiveThirtyEight, a website that looks at election forecasting, sports and statistics. The name comes from the number of voters in the electoral college. It is a superb source of information and I recommend subscribing. Silver’s writing and forecasting were better than most and he capitalized on it. The New York Times purchased FiveThirtyEight and Silver maintained his interests through writing, analysis, predictions and multiple media opportunities. Well on his way to becoming a leading public statistician (does such a position exist?), Silver’s The Signal and the Noise made him a national figure. ESPN purchased FiveThirtyEight in 2014 and Silver is editor in chief. One wonders what is next.
The Signal and the Noise marks another important step in our collective understanding of quantitative literacy. A college education should give graduates with mathematical skills and reasoning. For some, that is primarily the ability to solve particular kinds of problems. For others, it involves a curiosity and facility with both numbers and thinking about what the numbers might reveal. My sympathies rest with the latter. Quantitative literacy skills are essential to making sense of the world and making informed decisions. Silver’s work goes far in this ambitious project. I hope that many college students curl up with his book, subscribe to his blog, and decide on their own to take a statistics course. I would predict that they would find it helpful and relevant.