My Methodology page has a simple example of what I am doing below. You can also refer to my posts (here and here) at Advanced NFL Stats Community for more detail.
Here are the rankings, with lines updated as of the morning of September 23 (data courtesy of Killer Sports). First, a glossary of terms:
- LstWk - The team's ranking as of a week ago. Things will be volatile in the early weeks of the season.
- GPF - Stands for Generic Points Favored. It's what you would expect a team to be favored by against a league average opponent at a neutral site (if the team is at home, just add 2.5 points)
- oGPF - Stands for Offensive Generic Points Favored. It's the component of a team's GPF that is attributable to offense. I can derive this by combining the information in the point spread and over/unders. My post at Advanced NFL Stats Community has more detail.
- dGPF - Stands for Defensive Generic Points Favored. It's the component of a team's GPF that is attributable to defense.
Offense or Defense?The standard deviation of oGPF is 2.7 points and the standard deviation of dGPF is 1.7 points. This means that there is more variation in team strength due to offense rather than defense. You could interpret this as saying that defenses are somewhat at the mercy of what an offense can or can't do.
Another number of note is the correlation between oGPF and dGPF. They are positively correlated (correlation coefficient of 0.38). In football, a good offense can help out its defense and vice versa. Teams that score a lot move the ball down the field, so even if the drive stalls, they put the opposing team in unfavorable field position. So, the oGPF and dGPF rankings are probably not "pure" rankings of team defense and offense, but a linear combination of the two.