I want to talk about a spreadsheet.
Okay, it wasnt literally a spreadsheet, it was a statistical model running on a blog hosted by the New York Times, but bear with me. Tuesday night I sat on my couch with a laptop open and a bag of pretzels I definitely didnt need, watching the election results roll in state by state, and by around 11pm it was obvious what had happened. Obama won. Fine. Thats the headline everyone already has. Whats stuck with me since isnt the outcome, its watching Nate Silver and his FiveThirtyEight blog call all fifty states correctly while half of cable news spent October treating him like some guy who does math for fun and shouldnt be taken seriously.
Ive been reading 538 pretty much every day since Labor Day. Not because I care that much about politics, honestly I find most of it exhausting, but because the actual mechanics of what Silver does are just interesting to me as someone who likes data. Hes not predicting anything in the fortune-teller sense. Hes aggregating hundreds of individual polls, weighting them by past accuracy and sample size and house effects, and running the whole mess through a model thousands of times to get a probability. Thats it. Thats the whole trick. And in the last week before the election he had Obama at something like a 90% chance of winning, which a lot of pundits on TV treated as either wrong or somehow biased, because the "race is a toss-up" narrative was more fun to argue about on air for six hours a day.
Turns out treating a probabilistic forecast as a coin flip prediction was the mistake, not the forecast itself. A 90% chance means the underdog wins 1 time in 10, not 0 times in 10. People conflated "not certain" with "not confident," and a bunch of professional talking heads who get paid to have opinions about elections did not come out of this week looking great. I wont name names because I dont feel like getting angry emails, but if you watched any cable news in October you probably know exactly who I mean.
What I actually want to write about, though, is the bigger thing this points at, which is that this was a genuinely good week for the idea that a model built from public data and open methodology can outperform a room full of experienced people with sources and instincts. Silver published his methodology. He wrote it up in plain english on a blog anyone could read for free. He wasnt hiding behind a paywall or a consulting fee, he just showed his work, and the work held up. Thats the part that gets me. Its basically the same argument a lot of us make about open source software, except applied to punditry instead of code: show me how it works and let the results speak, dont just tell me to trust you because youve been doing this since before I could vote.
None of this is really about politics for me. Id be writing this same post if the model had correctly called it for Romney. The interesting story is a public, transparent, kind of boring statistical process beating a bunch of confident guesses, live, in front of everyone, with a scoreboard at the end. That doesnt happen very often. Usually punditry is the kind of thing where nobody ever really gets proven wrong because the claims are vague enough to wriggle out of later. This time there were fifty concrete answers and Silver got all fifty.
Small tangent because I cant help myself: I also spent part of election night comparing turnout maps on my phone, which two weeks ago would have meant squinting at a five inch screen, and now, thanks to the iPad Mini that showed up on my doorstep last Friday, meant a slightly less miserable seven point nine inches. Its not a device I was sure I needed. Turns out for reading in bed and checking maps on the couch its exactly the size that makes sense, better than lugging out the big iPad and smaller than hunting for my readers to use my phone. Doesnt change my life. Does change what I reach for on the couch, which honestly is most of what a gadget review should tell you and rarely does.
Anyway. Watch the aggregators next time some pundit tells you a close race "could go either way" a week out. Sometimes it really is close. Sometimes theres just a guy with a model who already worked it out and is waiting for everyone else to catch up.