At 10:37 p.m. on Tuesday, November 8, 2016, I wrote myself an email: “Nate [Silver] and Micah [Cohen] just told me Clinton is probably going to lose, that she’s an underdog … collapse in the Midwest.” I’d been watching the 2016 vote returns in our office. At 9:35 p.m. Donald Trump’s chance of winning the election was 26 percent according to our model; by 10:09 p.m. it had moved to 44 percent. At 10:31 p.m. a blog post of mine went up about the radical shifts among voters without a college education in Michigan. Still, it took someone saying it out loud to arrange all those particulate results and blog posts in a line that pointed in just one direction: Trump was going to win the election.