So I've been following FiveThirtyEight
for my electoral vote predictions lately (and having Electoral-Vote
text me every morning...). I really like their statistical method, running 10,000 simulations every morning with the new polling data. And the simulations are not pure Monte Carlo, but rather known ways that states and groups of states tend to move. The key graphs I wish to go over right now are:
The first image has the "Win Percentage" graph, which says that in 90.7% of those 10,000 simulations (or about 9070 of them), Obama wins. This figure comes into play later, but I want to focus on the more interesting chart, specifically the first two rows. The chance of Obama winning the Electoral vote but losing the popular is 5.07% (507 of 10,000 simulations) and the chance of the same happening to McCain is 1.10% (110/10,000). A cursory glance would make it seem like Obama has a five-times greater chance of pulling a Hayes
. However, given that Obama has a 90.7% of winning, then, if he wins, he has a 5.07/90.7 = 5.59% chance of that victory being a Hayes-ian one. But
, if McCain wins, he has a 1.1/9.3 = 11.9% chance of that victory being Hayes-ian — double
that of Obama.
For more on this wacky mathematics, check out Bayes' Theorem
, and be glad it's used to filter your spam!