Tuesday, September 23, 2008

Yet another electoral scoreboard

Take the median value of all polls in a state over the last month for each state. Use that as the best guess of how the state will vote. I admire the simplicity of that approach, especially since I am also a big fan of the power of the median as opposed to the overrated mean or average. Calculating the median also saves energy or computing power. Their method is not only simple but powerfully predictive. This method, (Gott and Colley's Median Poll Statistics), predicted 49 out of 50 states correctly in 2004. It currently has Obama getting 273 votes due to adding IA, CO, and NM to Kerry's 2004 states. Which polls to use? They use (the same ones as) RCP although you could theoretically add more polls and get an even better result since the median method automatically eliminates outliers. Mr Gott has done a good job with the map, too. (It would make a good quilt, RISK board, or pendant of fuse beads AKA iron beads in Japan.)

No comments: