More political geekery
Sep. 30th, 2008 05:36 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
gridloreFiveThirtyEight.com is a site devoted to statistical analysis of the various elections, primarily the presidential race, but also those for the Senate. It was started by Nate Silver, though he's added some other people since then. Silver made his name doing baseball analysis for Baseball Prospectus. His work there was good, and so is his
political work.
Silver uses a different method than Pollster.com. After weighting all of the various surveys by a number of criteria, he also uses state demographics and a couple of other items to create a simulation of the election. He then runs the simulation 10,000 times, and the results are what he goes by. It's an interesting technique, and it's the same basic premise as a lot of the stuff at Prospectus, including the Playoff Odds Report and Silver's system for projecting future player performance, PECOTA. In fairness, PECOTA hasn't proven to be much more effective than alternatives, such as Dan Szymborski's ZiPS, or even Marcel the Monkey. (Marcel is nothing but a weighted average of the player's last three seasons; beating Marcel is the real goal of any projection system, and they have mixed success.) It's a methodology that I've never seen in political analysis before. It's worth a look.
http://www.fivethirtyeight.com/
political work.
Silver uses a different method than Pollster.com. After weighting all of the various surveys by a number of criteria, he also uses state demographics and a couple of other items to create a simulation of the election. He then runs the simulation 10,000 times, and the results are what he goes by. It's an interesting technique, and it's the same basic premise as a lot of the stuff at Prospectus, including the Playoff Odds Report and Silver's system for projecting future player performance, PECOTA. In fairness, PECOTA hasn't proven to be much more effective than alternatives, such as Dan Szymborski's ZiPS, or even Marcel the Monkey. (Marcel is nothing but a weighted average of the player's last three seasons; beating Marcel is the real goal of any projection system, and they have mixed success.) It's a methodology that I've never seen in political analysis before. It's worth a look.
http://www.fivethirtyeight.com/
