In short, I think beating the pythag is random, but teams that minimize the randomness allowing for more one-Run (or 2-Run maybe) situations over the course of a season are better off in respect to actually beating their pythag.
Again, just a theory.
Someone on this board a few years ago shared the theory that beating pythagorian projections was tied to bullpen ERA, especially back-end bullpen ERA. Something about the ability of your closer and your best set-up men to hold small leads and avoid turning 1-run wins into 1-run losses.
Of course, you could also end up with a Danny Graves who turns 3-run wins into 1-run wins. That also helps you beat pythagorian projections.