Heidi Bond points out a few cool uses of statistical theory to show probable electoral college outcomes, including this site by Andrea Moro, an econ prof at Minnesota; I actually had more-or-less the same idea a month ago, but was too lazy to do anything with it.