Distribution Of All Statistics

A multi-university research program has compiled a profile of the effects of all known statistics that have ever been collected.

;-)

Special explanatory segment for the humor impaired (prob'ly mathematicians).

This is both true and a joke. It's a joke primarily because:

There is no such actual study. Such a study would be the equivalent of "find every acronym nobody uses". And it would have more data points. The page starts with a fabricative introduction to lead one into the ideas involved.
It's a joke because it's true. If you compiled every statistic ever compiled, distributed them on a scale between "true", "useless" and "harmful", because you sampled a very large population, and you rolled all its error bars up into one metric, you'd get a bell-shaped curve, almost normalized. Then, because you get what you measure (in both the input samples and the curve), a vanishingly few statistics would "EXACTLY match reality", and a vanishingly few would be "dangerously misleading". Statistics like, "Bread causes crime because 90% of all crimes are committed within 24 hours of eating bread." The families of folks spending their life in jail due to such PseudoScience would consider this sloppy thinking very dangerously misleading.

"There are liars, damned liars, and statisticians." -- MarkTwain