Now that we have the bare minimum of games played to derive some meaning to how good football teams are relative to each other (and we really still don't, but here we are anyway), it's time to roll something out that I've been doing for a few years: an objective ranking system for NFL teams (well, 2 rankings).
The first one I'd like to talk about is simple, and you likely already know it. It's just a pythagorean rating system: points scored squared divided by the sum of points squared and points allowed squared. That's simple enough; however, I've done a few more things to make the rankings more meaningful. I've included a stat called E[X], or expected wins. This stat is the number of wins this team is expected to win if they played their entire schedule over again. Since that’s not particularly meaningful, I then include aE[X], which is the adjusted win total (where they only play the games remaining on their schedule, and then add the wins each already has.
It's important to note that I DO factor in home field advantage. The home team wins about 57% of the time, so when I factor how a team should perform in an upcoming game, I factor that in.
The difference of these two expected win totals is what's interesting about this statistic. By definition, this is the difference between what a team is expected to do and what the team has actually done. I'm not sure you can attribute the difference to team skill or luck (surely it's a synthesis of both), but it's interesting to see what teams overperform their "peripheral" stats. You'd expect a healthy does of regression to the mean as the season goes on, but in my 3 years performing this analysis, I haven't really reached a conclusive resolution to this.
Here's the rankings for Rating (and then GRITS – Games Rescued In Tight Situations)
The first thing that pops out at me is that Carolina is in an extreme state of flux. A 38-0 pounding of the Giants will do that to you, of course. Part of the problem (early on) is that one game can have an inordinate effect on a team's rating. It's hard to imagine Carolina being the 3rd best team in the league (and GRITS has them as the 3rd-worst). San Francisco's ranking is equally troubling. For those prognosticating about the NFC North, Chicago's 11.4 wins has them comfortably ahead of GB (8.6), but just 1.2 games ahead of DET (10.2).
Honestly, though, I don't think these GRITS rankings are all that good. The problem that pyth ratings in football is interconnectivity. Teams just don't play enough to really be linked together like they should (in baseball, teams have 162 trials, an order of magnitude greater). This is why I developed another proprietary statistic, which is the diametric opposite of GRITS. I call it, somewhat funnily, OATS (Opponent-Adjusted Total Score). What OATS does is measure how a team fared against it's opponents relative to each opponent's opponents. It's somewhat hard to explain, so I'll use an example. Chicago scored 40 points on Pittsburgh in Week 3. As a team, Pittsburgh allows 24.7 points per game. So, relative to what Pittsburgh normally gives up, Chicago was 15.3 points better. Do that for each team (and average them), and you start to get an idea how good Chicago is relative to the rest of the league. Do the same thing for defense, add them together, and you get an idea how good they are in general.
In GRITS, a high margin of victory is actually harmful (because it inflates your rating, so your losses mean more and your wins less). In OATS, there's no such qualm about it (and many times, games that seem to be utter blowouts really aren't. SEA won by 28, but were expected to win by about 21).
One last stat (that doesn't mean very much) is Evenness. That's just the absolute value of the difference between a team's offensive and defensive ratings. If the evenness is low, then aren't better at one or the other. With a higher number, they definitely are. For examples, in my rating, Dallas is slightly above average in both offense and defense, so they have a low evenness. On the other hand, the Giants have a horrendous defense but a league-average offense, so they have a high evenness rating (which, when I type it out, seems unintuitive. Perhaps it should be oddness?)
Another bonus of OATS is that it gives you a predicted spread for an upcoming game. If you want to know if the Bears should be favored to beat the Lions, you can see that Chicago, as a team, is 2.78. The Lions are at -3.33. Since it's at Detroit, you'd give them 3 points of HFA, so add it together and you could say that the Bears should be around 3 point favorites (a word of warning; football DOES NOT WORK LIKE THIS. There are always, always, matchup considerations among other things. I haven't done extensive testing against the spread, but this isn't much (if at all) better than picking at random.) If you check Vegas, you'll see that the Lions are actually 3 point favorites.