Forecasters are feeling pretty good after the US election. Witness a recent XKCD:
Nate Silver and the prediction markets “called it” right. So did a different model by David Rothschild, our keynote speaker this Sunday. (And that was nine months ago.) Silver and Rothschild even “got” the 49 states which have declared a winner. Our team is among many people cheering for the quants. (Particularly our guest speaker, of course.)
But let’s not forget that good forecasting isn’t just about the “call”. As Rothschild and Silver have repeatedly said, it’s about judging the probability of an inherently uncertain event. If these forecasts were indeed correct, then we should believe two things:
- A week before the election, Romney still had a 1-in-4 chance of winning.
- When the model makes ten forecasts like this in the future, we should expect at least 2 to be “wrong”.
If the models & markets are good, then it’s partly luck they look so good.
- Forecasting the US Elections (daggre.org)
- Nobody’s perfect: Nate Silver and the imperfect art of prediction – Boston.com (blog) (boston.com)
- America’s Chief Wizard Nate Silver Had the Best Election Night of Anybody and Here’s Why: a Guide (gawker.com)
- is nate silver’s win sociology’s loss? (scatter.wordpress.com)
- Microsoft Researcher Predicts Obama to Win (spectrum.ieee.org)
- The US presidential election is no contest (newscientist.com)
- A better poll question: who do you think will win? (staradvertiser.com)
- How to Make Money on the Presidential Prediction Markets (pbs.org)
- ’12 Campaign: Predicting the U.S. Election (research.microsoft.com)
- David Rothschild: Obama Reaches His Highest Likelihood of Reelection (huffingtonpost.com)