Wednesday, October 29, 2008

Assorted thoughts, October 29 edition

- I noticed a shuttle to take people to the early voting booths this afternoon on its way through campus; presumably, the idea of voting over a span of time should increase the turn out, and so too should the availability of a shuttle to take you there as well.

Some may think that this increased voting would cut against the rational voter hypothesis-- quite the opposite, actually, as decreasing the cost of voting yields a higher turnout rate. Are these shuttles/early voting times varied across states? That's testable!

- I've heard a decent amount recently about the Washington Redskins proclivity to predict the winner in an upcoming presidential election. The theory goes as follows: In their last home game prior to the election, if Washington wins, then the incumbent party remains in power, and if they lose, the challenging party assumes the presidency. In 2004 it did not hold true, but in every presidential election prior to that it did, all the way back to 1936. I think I'm hearing about this because I receive Pittsburgh television broadcasting, and the Steelers are playing in Washington on Monday night.

Let's assume that this were a causal relationship-- through what mechanisms could this be true? I'm having a hard time even pulling something out of thin air. Would a better Washington team mean, across the board, the rest of the NFL cities are doing worse, and that means we want to keep the same party in office? Elections are on a Tuesday and the NFL plays on Sunday/Monday-- is there enough of a national Redskins following to have a national impact on elections immediately following the outcome of a game? Is there reason to believe that the success of Washington's football team is coincident with political cycles?

Yes, it's spurious correlation, but from 1936 to 2000-- that's 17 straight successes for a (presumably) 50/50 outcome. That just doesn't happen. (Though I believe there was one season where the St. Louis Rams lost the coin toss every single week, or something close to that.)

- What do you do when there's malinvestment in the economy? Add more credit!

-Here's an interesting result from Vegas concerning the ever intriguing Game 5 of the World Series. Just to clarify-- the possibility still exists that you could have bet the Phillies to win every single game of the World Series, have had it paid off four times, yet Tampa Bay could take home the trophy. Who doesn't love Vegas?!

By the way, the decision to postpone the game was a big edge for Tampa Bay-- Philly now has had a few days to think about winning 3 innings. Hamels could probably have gone another inning or two-- now he's out. And Balfour is a reliever who is used to working on sporadtic pitching schedules-- now he's basically on full rest coming into the 6th inning, with Price in the pen in case things drag on. And Philly doesn't want to go to Tampa. And Tampa has seemed to have their big players begin to produce now. In baseball, things even out over the long run, and even a World Series isn't long enough to always determine the best team, so 3+ innings tonight certainly could fall anywhere. But if you're Tampa, you've got to feel better after 6 innings than you did right before the first pitch of Game 5.


Justin M Ross said...

On the Redskins:

If we actively searched for an event over an entire domain of events (not just NFL) that occurred at least as often as elections, then by way of probability theory we should find some event that would have a perfect track record on predicting election outcomes. Ex: Suppose we asked people to guess each the outcome for each of 1,000 coin flips, then eventually we would find someone who would get all 1,000 correct. Essentially, I suggest to you that the Redskins prediction power stems from the infinite monkey theorem!

Matt E. Ryan said...

That is an interesting way of looking at it, and not incorrect. I do offer the following volley:

It could be possible that there would be another event that could have the similar predictive power-- but how "simple" could the event be?

What comes to mind is what I've heard said about making comparisons about baseball players. Many times, the question will involve many conditions-- "How many players have hit X home runs, have Y RBIs, stole Z bases, hit R doubles and had Q walk-off hits in 7 consecutive seasons?" For each condition you add, you lose some of the importance of the comparison.

To find such a non-conditional predictive event...that is what is interesting to me. It's the last Redskins home game, as opposed to the "rushing yards of Washington's QB when the game is at home, the receiving yards of the backup punter plus the number of fumbled interceptions when on the road."

Anonymous said...

Great theorem! I like what actually happened:

"Not only did the monkeys produce nothing but five pages consisting largely of the letter S, the lead male began by bashing the keyboard with a stone, and the monkeys continued by urinating and defecating on it."