That's cool, and I'm sure the author's conclusions are correct for that code base.

However, the skeptic in me can't help but point out that this is not nearly rigorous enough to say his method can pretty reliably determine TDD vs non TDD code in the general case.