Some Post-Oscar Thoughts on Forecasting Politics Done Right

Nate Silver is taking some guff for his foray into Oscar predictions. What is revelatory in this 538 post is how his venture into understanding why he missed two of three contested Oscars tracks his approach to baseball projections.

The model may be wrong, but that’s fixable. Which is why PECOTA gets better every year. What isn’t, as Nate so politicly admits, are the vagaries of unprojectable circumstances. Nate found out that projecting six Oscars with a dubious data set focuses much of the attention on the vagaries and the unprojectable. Um, he got them wrong.

Which is why his protracted explanations in this post are both admirable, he’s trying to figure it out, and a little sad–didn’t we trust him because he knew that already?

Regular readers know that I admire Nate’s work, but that I also think his great insight into projections is one of marketing. Not statistics. Nate figured out how to get everyone to ascribe the failure of his subjects to follow his model to his subjects, rather than to him. That isn’t a bad thing, it is a perfectly fine (perhaps brilliant) way to convey the confidence interval, but it doesn’t do much to help us explain the large swath of the numbers (in my case Baseball, in Nate’s, all of them) that are unpredictable.