« Department Of WTF? | Main | Joe the Strummer »

November 04, 2008

Comments

This is one of the most remarkable correlations in US politics. And it is completely devoid of any causation. But I guess that there are millions of possible correlations, so one of them must be really strong. It's still kind of freaky, though.

Post hoc ergo propter hoc. If you look at enough correlations, some of them will be true by dumb luck. Let's just hope that this isn't the year when the string is broken.

If we assume that the Redskins have a 50-50 chance of winning, then the odds of getting 16 out of 17 right are the same as getting 16 heads on 17 throws. I'm too lazy to do the full expansion but it's a bit less than the probability of getting 15 heads in 15 throws, which is 1/32,767.

My beloved Redskins, taking one for the team. Would that other sports franchises be as publically minded. (yeah, yeah, I know.

It seems fitting to include a link to Shirley Povich's column about a Browns/Redskins game

but it's a bit less than the probability of getting 15 heads in 15 throws, which is 1/32,767.

It's higher, I think. The chance of 17/17 is 1/131,072. So the chance of exactly 16 is 17/131,072, or 4.25/32,768 (not 32,767. 15/15 odds are 32,767 to 1, p is 1/32,768), or about 1/7710.

It actually provides interesting evidence that Kerry really won in 2004, as that is the only time where the prediction has not held true!

Always remember kids: 42% of statistics are made up on the spot.

"The Republicans haven't won a presidential election since 1972 unless their ticket had a Bush on it."

And I don't to jinx him, but it looks as if Bush will break the "year-ending-in-0" curse more decisively than Reagan did, by not suffering any sort of significant injury.

You also have to avoid the "prosecutor's fallacy". That is, if you're searching the entire universe for a match for some set of characteristics, the odds that any one individual (or team) will match is astronomically low, but the probability that that SOMEONE (or some team) will match is much higher. (For real world examples of this, check the Wikipedia articles for "Sally Clark" or "People vs. Collins"- both are downright scary).

I'd point out that any given team had two opportunities to fit the pattern- either the party who won the popular vote last time wins if the team loses, or vice versa. I have no idea how many professional and major college sports teams play in Oct/Nov, but for the sake of the argument I'll estimate 60. Since each one can fit the pattern two different ways, that's 120 opportunities to fit the pattern.

The probability that one of the 60 teams fits the pattern would then be
1-(1- 17/131072)^120 = .0144 = 1/65.

Of course, if you expanded your universe of search to include things other than sports teams, or to sports events earlier in the year, the probability of a match becomes much higher.

Thanks for the distraction- now I can go back to gnawing my nails to the quick fearing a McCain upset!


It actually provides interesting evidence that Kerry really won in 2004, as that is the only time where the prediction has not held true!

Funny!

No causation!!! Oh, there's causation here. It's just in the opposite direction than most people would normally consider. When the incumbents end up being ousted, they usually have had reason to suspect their impending ouster. This makes them edgy and sullen - not good qualities in a football team's fan base for the purposes of maximizing home-field advantage. Confident and happy incumbents make much better fans. Since we're talking about the Washington Redskins playing home games, it should be pretty obvious what's going on here, shouldn't it?

If we assume that the Redskins have a 50-50 chance of winning

Well, there's your problem, right there.

it looks as if Bush will break the "year-ending-in-0" curse more decisively than Reagan did, by not suffering any sort of significant injury.

Ah, the Julian fallacy--the real curse is for leap years ending in 0. In the Gregorian calender we presently follow, 2000, like 1600, was not a leap year.

And every president elected in a leap year ending in zero since old Tippecanoe has died in office, although with Reagan, it took us a few years to notice.


If we assume that the Redskins have a 50-50 chance of winning

Well, there's your problem, right there.

it looks as if Bush will break the "year-ending-in-0" curse more decisively than Reagan did, by not suffering any sort of significant injury.

Ah, the Julian fallacy. The curse was for "leap-years-ending-in-0". Thanks to good old Pope Gregory, 2000, like 1600, was not a leap year.

Every president elected in a leap year ending in 0 has died in office, although with Reagan, no one noticed until much later.

But 2000 was a leap. It was 1900 that suffered with but 365 days.

2000 was a Leap Year, as are years that are multiples of 4 that are not multiples of 100 unless they are multiples of 400. 1900 was not a Leap Year, but 2000 was.

2000 is a leap year; 1900 was not. This is of small consolation to William McKinley.

(Back in the 80s, I worked for a database company. It was a frequently reported "bug" that our date functions thought that 2000 was a leap year.)

I read somewhere that the correlation is a perfect 17 for 17 if you change it from "incumbent party" to "party that won the popular vote". Of course, that doesn't roll off the tongue very easily but it's even more proof that Al Gore was robbed.

Dealing with the whole leap year business used to be a common exercise for people just starting to learn programming. Wonder if it still is.

That's why I'm hip to the Leap Year jive. Fortran in 1986 - YUK!!

COBOL to rule them all, and to the 60s bind them.

What's the correlation between Democrats winning the White House and every one of my college football teams losing?

And regardless: can we just stop them from losing, period? It's getting old.

Dealing with the whole leap year business used to be a common exercise for people just starting to learn programming. Wonder if it still is.

It was for me a few months ago...

[...not that I didn't know how to do it, I just had to get certified by my company.]

"I read somewhere that the correlation is a perfect 17 for 17 if you change it from "incumbent party" to "party that won the popular vote". Of course, that doesn't roll off the tongue very easily but it's even more proof that Al Gore was robbed."

No this just flips the 2000 and 2004 predictions. The Redskins lost in 2000, predicting the loss of the incumbent party. Gore had a larger popular vote.

They lost again in 2004 predicting the end of Bush. But he was too powerful *cue evil music*. He won the electoral college, the popular vote, AND evaded the Redskins-loss/curse!

Sheesh, do I have to explain the basics of astrology too?

And I don't to jinx him, but it looks as if Bush will break the "year-ending-in-0" curse more decisively than Reagan did, by not suffering any sort of significant injury.

Just keep the man away from pretzels for a couple of months.

Thanks -

Every president elected in a leap year ending in 0 has died in office,

James Monroe begs to differ.

How about this one from the New York Times in 1903:

"The letter in the alphabet which has most signalized itself in the names of Presidents is the letter n. It figures extraordinarily as the final letter of the surnames of fourteen successful candidates. Moreover, eight out of ten times when two men ran for President and Vice President successfully on the same ticket and when their name ended in n they were elected...

One of the two failures was when Van Buren and Johnson ran for a second term in 1840, and yet it was a man with an n at the end of his name to beat Van Buren. The man who beat him was William H. Harrison. The other failure was when McClellan and Pendleton were Democratic candidates in 1864. However, even then the two men with n at the end of their respective names defeated them, namely, Lincoln and Johnson."

After stumbling on this article awhile back, I looked into how the theory has held up since 1903, and if I remember correctly, it looks grim for Obama. I can't find where I wrote it down, however, and don't care to go through it again at the moment.

[...not that I didn't know how to do it, I just had to get certified by my company.]

Right, anarch. Letting that shiny new doctorate go to your head a bit, it seems. :-)

So I'm surprised by that statistic. It seems to me that with 32 NFL teams, and god knows how many baseball teams, each playing the others, which we can interpret as "the Incumbent", "the Democrats", "the younger", ... "will win" vs "will lose", "will win the popular vote" ... and so on. And then add an extra bit of combinations: "at home"...

Well, we wind up with a LOT of combinations. Some of which have to work out.

"If the Oakland As win 3 games of their first spring series against the Cleveland Browns, then the younger candidate will get more electoral votes than the older one."

Just keep the man away from pretzels for a couple of months.

And no hunting trips with Cheney.

I'm watching Alan Alda and Jimmy Smits campaign for President.

Also, I just finished reading Jasper Fforde's Something Rotten. Fforde is ffcking brilliant.

"If the Oakland As win 3 games of their first spring series against the Cleveland Browns, then the younger candidate will get more electoral votes than the older one."

That one is definitely true.

"I'm watching Alan Alda and Jimmy Smits campaign for President."

I'd so much rather have had Senator Arnold Vinick running, than Senator John McCain.

The comments to this entry are closed.

Blog powered by Typepad