The telltale cliff

Dan Luu has a wide-ranging post documenting suspicious discontinuities in data, such as marathon runners pushing themselves to achieve round number finishing times, especially half-hours; published p-values clustering just below the magic threshold of 0.05; and Russian voters apparently favouring turnout percentages that are multiples of five.

I wrote a much narrower post on the same subject a few years ago. But mine had gigantic underwater sea walls!

^{Hat tip to tosh on Hacker News.}