The table above contains are our model driven NFL power ratings, which are updated weekly using a combination of game results, play by play efficiency, film grades from Pro Football Focus, and a variety of custom models.
These ratings also incorporate quarterback grades from 538 and will account for changes to the starting quarterback or when an elite quarterback goes on a hot streak, but don't account for other player level changes like an injury along the offensive line or skill positions.
To improve predictive quality, all our ratings start the season using a blend of the prior season's performance and football outsider's projected DVOA. Incorporating DVOA projections helps to account for additions made in free agency, the impact of a solid draft, the development of young players, and head coach changes.
This model has proven to be quite strong at predicting game outcomes. In back tests, the nfelo model has been more accurate than the closing line, and routinely achieves top decile performance in 538's pick-em competition.
In addition to our NFL power rankings, the table above includes some useful NFL team stats.
The Wins section compares how many wins a team actually has to how many wins they're expected to have based on a pythagorean wins model, PFF's film grades, and 538's pregame projections. Teams that have more or less wins than expected tend to regress to the mean overtime.
The Points section highlights points for and points against. While these stats seem simple, point differential is one of the most predictive measures of NFL team performance.
Finally, the EPA section details a team's per play efficiency based on Expected Points Added, which is an advanced stat derived from play by play data. This is a great way to determine the relative strengths of a team by looking at how they score points. For offensive EPA, positive values are better, while for defensive EPA, negative points are better.
NFL Analytics and Betting