The Statistics of Small Numbers

Dr. Nick makes an insightful observation

I’ve noticed for this blog anyway, if a post is going to get commented upon, it will be commented upon the first day it appears, and generally not thereafter. I wonder if this is true for blogs in general? I think it might be.
...
Yes. One might even expect a … wait for it… power law to apply, somehow?

I think it's more like a Poisson Distribution or, in cases where the numbers aren't so small, a Cauchy or even full-bore Guassian.

Wonkette makes a similar observation in a different context

The Huffington Post yelped, almost Drudge-style, "NBC/WSJ Poll: 2% Of African-Americans Give President Bush A Positive Rating," with the helpful modifier, "UNBELIEVABLE..." And, indeed, it is. For a good reason: It's not exactly true. While two percent grabs headlines and make Tim Russert all wriggly ("Only 2 percent -- 2 percent!" as he spurted last night), the more significant number is buried in report. Dug up by Dan Froomkin, it's this: Out of 807 surveyed for the poll, only 89 were black. As Froomkin puts it, "there is a considerable margin or error." Our statistics are rusty, but it seems like it could be high as +/- 10 percent. Sure, there's a possibility that Bush has a -8% approval ratings from blacks. Or maybe Kanye West should investigate whether NBC/WSJ care about what black people really think.