Benford's Law must be the most mind-bending thing I've read in a week. It turns out that in statistically random data, using any sort of scale, the number 1 is far more likely to occur than any other number. No matter what the data or scale is.
If you don't understand what I mean, go read the article. He's actually pretty good at explaining it. After that, scroll down to the applications to detecting financial fraud. That just completely blew my mind. It turns out that this law is good for picking out data that has been rigged in some way.
(via, btw, this awesome blog post, via Krugman)