I enjoy the fact that he uses two data points to support his conclusion that using two data points gives you better conclusions than guessing. Nothing circular about that reasoning at all. :-)