> Similarly finding any shape in a random set of points is much more likely than the odds of any one shape.

Obviously. But that’s not the point (no pun intended). My point is that most of the "shapes" would be just an unstructured shape—if you can even call it a shape. "Familiar" shapes will be much much unlikely to form that "uncommon" shapes. (Hopefully this is obvious because the number of familiar shapes are much much fewer than uncommon shapes.)

Let me use another example to help you understand the point. Suppose a monkey is given a typewriter and a sheet.

Is the probability of getting The Declaration of Independence is as likely as the probability of getting one particular gibberish sequence of characters? Yes.

Should we surprise if the monkey types any proper one-page English essay? Yes.

In case it's not obvious, that's because the number of possible ways to write a proper one-page English essay, albeit humongous, is nothing compared to the number of possible ways to arrange characters in one page. In other words, it's very very very unlikely to happen.