The story goes like this: if the Washington Redskins win their last home game before the presidential election, the incumbent wins, and if they lose, the incumbent loses. It has correctly "predicted" the outcome since 1936.
It's amusing, and meaningless, of course. The two events have nothing to do with one another. But it does give us an opportunity for a little statistical fun.
Suppose the probability of the two events happening together (Redskins win/incumbent wins or Redskins lose/incumbent loses) is 1/2. It's not exactly, but it's a reasonable approximation for starters. The probability of it happening twice in a row is 1/4. The probability of it happening n times in a row is (1/2)^n. There have been 17 elections since 1936, so (1/2)^17=0.0000076. Impressive indeed!
Or maybe not. Open challenge (I might try to analyze this if I have time): argue that the success of this indicator of elections is not as impressive as the above paragraph suggests. I have the beginning of an argument started, but it's already way too long to put here.
It's kind of like the observation that an old NFL or NFC team winning the Super Bowl tends to coincide with a rising stock market that year. Historically, the NFL or NFC has won more Super Bowls (they have had better teams) and the stock market is up more years than it is down (for reasons that have nothing to do with football). Note that this predictor seems to work well only because the Pittsburgh Steelers (4 Super Bowl wins) are an old NFL team and correctly predicted bull markets! If the Steelers were counted with the AFC (their current conference), the predictor would not do much better than pure chance.
And that's the beauty of looking for these patterns. People just search around until they get something that works! That's why I am always so skeptical of stuff like this. Big deal! It's amusing--nothing more.
Bottom line: The world is full of improbable events. Improbable events are literally happening all the time. Each one is so unlikely, but there are so many opportunities for unlikely things to happen. Sometimes a coincidence is just a coincidence!
I have an academic paper on the subject of conditional probabilities, so this is more than just a passing interest. I might write more on this after the election--no matter who wins.
Leave a comment