Nate Silver’s book, The Signal and the Noise, is about making predictions. It examines what leads people to get predictions wrong, and what leads people to get them right—like Silver did when his model correctly predicted the result of the 2012 presidential election in all 50 states. That triumph led Jon Stewart to crown Silver “the lord and God of the algorithm.” Silver has since taken his approach to statistical analysis to FiveThirtyEight, where he is the editor-in-chief.
Ahead of the release this week of the paperback edition of his book, I spoke to Silver about sports, politics, and the theories we shouldn’t trust.
Noah Gordon: You’ve written a new introduction for the paperback edition of The Signal and the Noise. Have statistical predictions improved much since 2012, when many analysts overrated Romney’s chances? Do people put too much faith in them?
Nate Silver: I’m not sure I’d say people put too much faith in statistics or analytics. On the whole I’d say people put too little. But I think there’s a gap between the increasing desire to use analytics and where the craft is in terms of the practice. One irony about 2014 is that you actually had Democrats making a lot of the same mistakes that Republicans did in 2012. In defiance of where the polls stood they thought, oh, Democrats are going to keep the Senate. They came up with reasons why, like “our turnout operation will save us,” or “the demographics are favorable for us,” or they’d cite 2012 as a precedent. Until it actually happened, and the election was kind of a disaster for Democrats. You know, they were making the same mistakes. Less high-profile mistakes because people aren’t as concerned about the midterms as the presidency, but still—the average poll wound up overrating how Democrats would do by about three or four percent…