Fivethirtyeight.com's performances

[Mr. Morden]

I'd be curious to see an analysis that looks at his probabilities for each race going one way or the other, and compares them to the actual outcomes.  For example, you take the races where he says the Democrat has a probability of victory of between 20% and 30%.  What %age of those races does the Democrat actually win?  Is it somewhere between 20% and 30%?  You could do that for various time steps, how does he do one month out, three months out, six months out, etc?

Because really, that's the test of whether his probabilities mean anything.  If events that he rates as having a 30% probability don't actually happen 30% of the time, then he's doing something wrong.

[Mr. Morden]

Cool.  If you like, you can also fold together the cases for both parties, and look at how many times the guy who was favored with X % chance of winning won, regardless of party.

So:

Favorites who were given a 50-75% chance of winning won: 4 out of 6 times, or 67% of the time
Favorites who were given a 75-90% chance of winning won: 7 out of 9 times, or 78% of the time
Favorites who were given a 90-100% chance of winning won: 58 out of 58 times, or 100% of the time

Of course, that's from his last update before the election.  It'd be more interesting to look at his projections from many months out, when not so many of the races are going to be in the 90-100% range.