Let us now praise Nate Silver

This was's final projections for Tuesday's Presidential election based on state and national polling. 


These were the final results:


As you can see, they did pretty well. From a binary standpoint, the only real "miss" was Indiana, but he had them as essentially a tossup.

A lot of people have gone from obscurity to household name during this most recent election season, but as far as Internet punditry is concerned, Nate Silver (and his co-conspirator, Sean Quinn) has arguably seen his star rise higher than most, mostly because his projections - based on 10,000 simulations of the election factoring in weighted polling numbers (as well as other information) - held true. While many Dems were struggling with their own fears that somehow the treacherous GOP machine would somehow steal another election, Silver kept the faith in his system. And it wasn't just the Presidential race - while Minnesota and Georgia are still undecided, 538 got each Senate race right as well, with the exception of Alaska. And obviously some weird things are going on there.

Silver and Quinn don't spend much time on 538 talking about policy, even though they both are Obama supporters. They look at who is voting for whom for what reasons. They look at who has the better-organized ground game (Obama, unsurprisingly). If they see a poll that would appear to be an outlier, they don't openly dismiss it, but rather they dig deeper in an attempt to find out why, like when an IBD/TIPP poll showed McCain with in two points nationwide and ahead 74%-22% among 18-24 year olds (their conclusion: IBD/TIPP polled too small of a sample size fitting this description for it to be meaningful. It was later reported that the poll had only asked around 30 such people).

In short, there's good reason to listen to these guys, and I'm glad that they're gaining a lot of national traction with this effort. But for me, the job is not done. The largely positive reaction from the media towards Silver's efforts stands in sharp contrast to the way much of the world viewed his first data-driven exercise, Baseball Prospectus's PECOTA projection system, which attempts to use past performance along with age-and-injury regression models and the performance of similar players to try to quantify rough expectations for how a certain player (and therefore team) can be reasonably expected to perform. While there are a few things it can't account for - broken bones, for example - it's a similar system to the 538 projection system. And, not surprisingly, it's right more than it's wrong (for example, PECOTA foresaw the Rays' rise to contention this year), yet to hear many people tell it, those guys are a bunch of pencil-pushing nerds who have no idea what they're doing (see this site for examples).

So all I'm saying, in my roundabout way, is this: now that Nate Silver's methods have been proven true, can we lay off the baseball fans who want to bring up things like VORP in a discussion about why the Red Sox should sign Mark Teixeira and treat Mike Lowell like a sunk cost? Please?

| More

 Friends' Activity   Popular 
All Blogs
Follow the Phoenix
  • newsletter
  • twitter
  • facebook
  • youtube
  • rss
Latest Comments
Search Blogs
Phlog Archives