Given a large enough sample of random noise, you can find any finite pattern in it, akin to the infinite monkey theorem. The human mind is biased toward patterns.
I asked my astronomy teacher about this a week ago and told him the same thing you said- I asked him if it would be any harder to find concentric pentagons (if that's what they're called) in the WMAP since it is such a huge data set.
My teacher told me that he didn't read the paper, but one of his grad students read it and pointed out that the man doing the analysis of the data didn't account for the probability that those patterns would be there anyway, apparently a common mistake. I wish I knew the proper term that he used, but the semester is already over.
"No prior" is, I guess, the term he used. There's some prior probability of finding circles in "normal" noise. It's only news if there are more circles than that. And there weren't.
True, but if the first humans on Mars were to find the complete works of Shakespeare buried under the Martian soil, it probably wouldn't be prudent to dismiss it as a cosmological fluke.
Honestly, we have the statistical techniques needed to figure out whether a given pattern in the CMB is real or not. I'm no statistician, but folks that are can do this kind of thing for breakfast.
if you choose the encoding such that each letter in the alphabet encodes some Shakespearian paragraph, then you can find many lines of Shakespeare on this very discussion.
