The Pythagorean expectation is a stat taken from Bill James and baseball, and modeled after the mathematical pythagorean theorem. In short, it looks like:
Winning Percentage = PF^N/( PF^N + PA^N)
Where ^ is the exponentiation operator, N is the Pythagorean exponent, PF is points scored by your team, and PA is points scored against your team.
While in baseball N = 2 works well, in football, by my calculations, that exponent varies, and can be 2.2 for a season or 2.8. Daryl Morey, back in the day, calculated a value of 2.37 for the teams of his era, and that's the value folks like Aaron Schatz (Football Outsiders) uses. In the modern era, I suspect the value is too small. A typical season will have an optimal Pythagorean around 2.5, post 2000.
That said, I went back through all the Matt Ryan - Mike Smith seasons, calculated the winning percentage of those teams, and their Pythagoreans, and ran off the chart posted here:http://codeandfootball.wordpress.com/2012/11/20/thoughts-on-nfl-week-11/
Mike Smith teams win more than they should, based purely on scoring stats. That's not something weird, it's just a fact. Getting down to the how of it can be fun, can be frustrating, but for now, just is, the way some folks automatically attract attention and others couldn't get any if they tried.