Statistics and the Curse of Tippecanoe

For those of you who haven't heard the legend, take a quick look here:

http://en.wikipedia.org/wiki/Curse_of_Tippecanoe

Back/never left? OK.

I heard of the curse a long time back (grade school?), but wasn't much interested. A coincidence, nothing more.

Well, I just ran the numbers, and... wow. That'd be quite a coincidence

Bad enough that the first seven candidates for the curse did, in fact, die in office and the eighth came close (the ninth seems to have evaded the curse entirely). What's really disturbing is this: we've only had 8 presidents out of 43 who died in office. That's less than 1 in 5, but the 'curse' has claimed 7 of 9. The odds of seven consecutive hits on those not quite 1 in 5 odds are over 120,000:1.

All four of the presidents who were assassinated (ruling out the poisoning theories for Taylor) were curse candidates. The 8th candidate (Reagan) came quite close indeed to assassination while in office (lost about 1/2 his blood volume and the bullet stopped about an inch from his heart, IIRC).

We've only had ONE president who was NOT a curse candidate die in office - of the 34 non-candidates.

1/34 vs. 7/9. Less than 3% vs. nearly 78%. If as an engineer I was told that a product had failed 8 times, and that 7 of the failures were from a batch of nine, I'd reasonably assume there was something wrong with that batch and recall the other two from that batch. Sure, I'd check for other things (were all of that batch bought by one Wyle E. Coyote?), but I'd expect to find some defect in the manufacture of that batch that wasn't shared by the full production run.

I think there's more evidence that curses get less deadly over time (note that the 1st victim also had the shortest time in office of any US President - less than 31 full days), than there is that there's no curse.

And yes, I know that the 1st rule of statistics is that statistically unlikely events do happen.