With the exception of San Diego still up at number 2 this week, the rankings are starting to make sense. Or, is it that perception is catching up with the stats? There were some big movers this week, including Tennessee, New England, Dallas, Atlanta, and Chicago.
The team rankings below are in terms of generic win probability. The GWP is the probability a team would beat the league average team at a neutral site. Each team's opponent's average GWP is also listed, which can be considered to-date strength of schedule, and all ratings include adjustments for opponent strength.
Offensive rank (ORANK) is offensive generic win probability, which is based on each team's offensive efficiency stats only. In other words, it's the team's GWP assuming it had a league-average defense. DRANK is is a team's generic win probability rank assuming it had a league-average offense.
GWP is based on a logistic regression model applied to current team stats. The model includes offensive and defensive passing and running efficiency, offensive turnover rates, defensive interception rates, and team penalty rates. If you're scratching your head wondering why a team is ranked where it is, just scroll down to the second table to see the stats of all 32 teams.
Click on the table headers to sort.
| RANK | TEAM | LAST WK | GWP | Opp GWP | O RANK | D RANK |
| 1 |  PIT | 4 | 0.81 | 0.53 | 2 | 3 |
| 2 |  SD | 1 | 0.81 | 0.37 | 1 | 5 |
| 3 |  NYG | 2 | 0.75 | 0.53 | 7 | 2 |
| 4 |  IND | 3 | 0.71 | 0.53 | 4 | 13 |
| 5 |  KC | 6 | 0.67 | 0.53 | 10 | 9 |
| 6 |  TEN | 12 | 0.65 | 0.52 | 13 | 1 |
| 7 |  GB | 5 | 0.64 | 0.48 | 5 | 15 |
| 8 |  PHI | 9 | 0.61 | 0.48 | 9 | 7 |
| 9 |  MIA | 13 | 0.60 | 0.54 | 16 | 18 |
| 10 |  HOU | 8 | 0.59 | 0.57 | 3 | 26 |
| 11 |  NE | 18 | 0.55 | 0.53 | 17 | 14 |
| 12 |  WAS | 11 | 0.54 | 0.55 | 14 | 20 |
| 13 |  BAL | 14 | 0.54 | 0.50 | 18 | 11 |
| 14 |  DAL | 7 | 0.52 | 0.58 | 6 | 21 |
| 15 |  ATL | 24 | 0.52 | 0.49 | 11 | 23 |
| 16 |  CHI | 10 | 0.50 | 0.51 | 28 | 6 |
| 17 |  CIN | 21 | 0.48 | 0.47 | 15 | 25 |
| 18 |  NYJ | 20 | 0.48 | 0.48 | 21 | 10 |
| 19 |  CLE | 25 | 0.47 | 0.55 | 22 | 17 |
| 20 |  NO | 17 | 0.46 | 0.40 | 12 | 28 |
| 21 |  MIN | 16 | 0.45 | 0.51 | 27 | 8 |
| 22 |  DEN | 19 | 0.41 | 0.47 | 8 | 30 |
| 23 |  TB | 28 | 0.40 | 0.48 | 19 | 29 |
| 24 |  SEA | 15 | 0.39 | 0.43 | 26 | 12 |
| 25 |  SF | 22 | 0.38 | 0.47 | 23 | 19 |
| 26 |  CAR | 26 | 0.37 | 0.50 | 31 | 4 |
| 27 |  DET | 23 | 0.36 | 0.54 | 20 | 24 |
| 28 |  BUF | 30 | 0.30 | 0.51 | 25 | 32 |
| 29 |  STL | 29 | 0.26 | 0.43 | 30 | 22 |
| 30 |  OAK | 31 | 0.26 | 0.47 | 24 | 27 |
| 31 |  JAC | 27 | 0.25 | 0.59 | 29 | 31 |
| 32 |  ARI | 32 | 0.23 | 0.45 | 32 | 16 |
Here are the team efficiency rankings through week 7.
| TEAM | OPASS | ORUN | OINT% | OFUM% | DPASS | DRUN | DINT% | PENRATE |
| ARI | 4.2 | 4.7 | 4.2 | 1.7 | 5.9 | 4.3 | 3.2 | 0.49 |
| ATL | 6.2 | 4.2 | 2.0 | 0.3 | 7.0 | 4.1 | 4.5 | 0.36 |
| BAL | 6.4 | 3.6 | 2.5 | 0.8 | 6.0 | 4.2 | 2.2 | 0.39 |
| BUF | 5.6 | 4.6 | 3.3 | 0.7 | 7.0 | 4.8 | 0.6 | 0.35 |
| CAR | 4.7 | 3.5 | 5.2 | 2.9 | 5.5 | 3.9 | 5.2 | 0.40 |
| CHI | 5.7 | 4.0 | 5.6 | 0.3 | 5.5 | 3.6 | 3.4 | 0.36 |
| CIN | 6.4 | 3.8 | 2.5 | 1.6 | 6.3 | 4.5 | 3.9 | 0.40 |
| CLE | 5.9 | 3.9 | 3.8 | 1.8 | 6.9 | 3.9 | 3.4 | 0.42 |
| DAL | 6.8 | 3.9 | 2.8 | 0.0 | 6.5 | 4.5 | 2.9 | 0.61 |
| DEN | 6.9 | 2.9 | 1.4 | 1.7 | 6.8 | 4.8 | 2.5 | 0.53 |
| DET | 5.7 | 3.5 | 3.1 | 1.0 | 6.4 | 4.9 | 3.6 | 0.55 |
| GB | 7.0 | 4.4 | 3.8 | 0.6 | 5.5 | 4.7 | 4.0 | 0.43 |
| HOU | 6.7 | 5.2 | 2.5 | 0.3 | 7.5 | 4.2 | 1.7 | 0.36 |
| IND | 7.3 | 3.7 | 0.8 | 1.2 | 6.0 | 4.8 | 2.6 | 0.32 |
| JAC | 5.5 | 4.2 | 5.4 | 1.1 | 8.1 | 4.6 | 1.9 | 0.39 |
| KC | 6.4 | 5.0 | 2.0 | 0.3 | 6.2 | 3.7 | 2.2 | 0.36 |
| MIA | 6.3 | 3.8 | 2.8 | 1.0 | 6.5 | 3.8 | 2.2 | 0.23 |
| MIN | 5.7 | 4.7 | 5.6 | 1.4 | 5.9 | 3.8 | 2.9 | 0.46 |
| NE | 6.1 | 4.1 | 2.0 | 0.0 | 6.8 | 3.8 | 3.4 | 0.39 |
| NO | 6.6 | 3.9 | 3.5 | 1.1 | 5.9 | 4.0 | 2.0 | 0.43 |
| NYG | 6.8 | 4.8 | 4.6 | 1.8 | 5.0 | 3.5 | 3.1 | 0.45 |
| NYJ | 5.6 | 4.9 | 1.1 | 1.3 | 6.1 | 3.5 | 2.3 | 0.55 |
| OAK | 5.5 | 4.7 | 3.8 | 0.8 | 6.2 | 5.0 | 1.5 | 0.63 |
| PHI | 6.3 | 4.9 | 1.6 | 0.8 | 5.6 | 4.0 | 4.6 | 0.58 |
| PIT | 7.4 | 3.9 | 3.7 | 1.4 | 5.9 | 2.8 | 3.6 | 0.32 |
| SD | 7.6 | 4.1 | 2.2 | 1.9 | 4.7 | 3.5 | 3.3 | 0.38 |
| SF | 5.7 | 4.1 | 3.9 | 0.9 | 6.4 | 3.6 | 3.1 | 0.56 |
| SEA | 5.4 | 3.7 | 2.9 | 0.0 | 5.9 | 3.3 | 2.4 | 0.46 |
| STL | 5.0 | 3.8 | 3.1 | 0.0 | 5.6 | 4.6 | 1.9 | 0.42 |
| TB | 5.6 | 4.0 | 1.4 | 0.7 | 6.4 | 5.3 | 5.6 | 0.38 |
| TEN | 6.4 | 4.3 | 2.9 | 1.2 | 5.3 | 4.0 | 4.3 | 0.54 |
| WAS | 6.3 | 4.1 | 2.8 | 1.2 | 6.4 | 4.7 | 2.6 | 0.35 |
| Avg | 6.1 | 4.2 | 3.1 | 1.0 | 6.2 | 4.1 | 3.0 | 0.43 |
Sorry, but your Ranking is making actually less sense to me then last week.
The game in SD went actually mostly like your modell predicted, i.e. the Chargers outplayed my Pats both on offense and defense. But instead of crying out "Yeah I got it right" your system then goes all whoops and raises the Pats by almost ten places?
The only problem I have with this model is it does not incorporate play calling/coaching. Success Rate considers play calling by evaluating the success of a play. For example, an otherwise efficient team, may have a coach who always chooses to punt/kick on 4th down or attempt a 2 point conversion when the game is tied with no 1 second remaining. Such poor coaching would not be adequately accounted for in this model. If Success rate were incorporated, I think this model's accuracy would greatly improve.
Pats jumped up because SD's rating was so high going in. NE's OppGWP increased, and because so many teams are packed tightly together in the middle, NE jumped to the top of that heap.
"With the exception of San Diego still up at number 2 this week"
I wouldn't be so quick to count out San Diego. They came with a foot to the left of overtime against New England even though they turned the ball over 4 times to zero, including two completely unpredictable fumbles.
In a model that (rightly or wrongly, that's not the argument here) doesn't take fumbles into account, I would say that means San Diego is a pretty good team - they're just finding ways to lose that you simply can't predict.
I thought it was interesting that SD was surprisingly high, and NE was surprisingly low, because their actual W-L records were largely determined by fluke plays - and then NE beat SD due to fluke plays. I wonder if there's a point at which it becomes enough of a pattern that it actually makes sense to factor it in.
dfan - if it's genuine fluke then no, it shouldn't be factored in. With the NFL only being a 16 game season it can be quite easy to see 'a pattern of bad luck' as it were and assume it's a permanent quality of that team.
If every game was a 50/50 you'd still expect to get one team at 4-12 and another at 12-4. It's such a short season that it's hard to tell whether something is a trend or it's just really unlucky.
Brian, any reason the Saints D is 28th???? I know that your formulas are different from Football Outsiders, and both are different from the NFL's "yards against", but what gives? The NFL rates them 3rd (too high), FO rates them 14th (iirc--seems about right), but yours 28th (seems awfully low)?? Four of the Browns 6 scores came as a result of 2 INT TD's, and two fluke special teams' plays. A 5th one was set up by a long DPI on 3rd down that put them on the 4 yd. line. Against the Cards, pretty much the same. They beat the other four teams on the sched, and the other loss was to the Falcons in OT where our kicker should have won it--thus the D gave up more yds/pts than they would have. In fact, iirc, the DEFENSE has only allowed 20 or more points 2x--against the Falcons (no shame there) and against the 49ers. What gives??
That does seem low. They must have had some games against some seriously bad offenses. Who have they played so far?
Saints have played so far the Vikings (27 O rank), 49ers (23), Falcons (11), Panthers (31), Cardinals (32), Buccaneers (19) and Browns (22). That's an average of 24.
I guess if you allow horrible offenses look average against you. Your defense is not a good defense.
Brian, I just want you to know I'm looking forward to your article next year about how the new, currently-unnamed Cowboys coach will look to have "turned the team around", when it's really just a matter of Romo's return and regression to the mean. Thanks.
Some good points raised in the comments. I think it's interesting that the Titans are ranked so high with a terrible defense. Then again Indy did it for a lot of years with a bad defense and an awesome offense.
The comments are the best part of this site.
James-sounds like a great idea for a post at the community site. You could just write a Madlibs-type article with blanks for the team name, the injured QB, and the new coach.
Brian - Do the "Orank" and "Drank" columns reflect Strength of Schedule adjustments? In prior explanations of your methodology, it seemed to indicate that the Strength of Schedule adjustment was more global, and reflected the overall strength of the teams. Your comments above seem to indicate that the adjustment is down to at least the Offensive and Defensive level. Just curious.
Have you considered using some form of stadium factor similiar to here: http://www.feinsports.com/?p=184
No on the stadium factors.
Michael- Yes, ORANK and DRANK are adjusted for opponent strength. E.g. CHI's DGWP is adjusted for the OGWP of their opponents.
I think it would be better to have some sort of value rather than a rank. It's hard to tell how bunched up teams are just looking at ranks.
The difference between 2 to 32 could theoretically be as big as the difference between 2 and 1.
We do have a value! It is the GWP.
I'm confused, CAR has a .11 GWP advantage on STL, yet is .49 to win this week. I thought homefield only accounted for 7 or 8 percentage points...
Also, it appears that you are basing DEN v SF on a game played in SF? The game is in London.
Anon, you are misreading the chart. GWP is the probability of the team winning against an average team, and Opp GWP is the combined GWP of the teams faced so far this year.
This is NOT a projection of who is going to win the games this week. That'll be in a different post.
Can you add starting field position to your stats, if available. This should capture the importance of special teams and turnovers.
Also, forced fumbles are not random, only the recovery of fumbles are random.
I know this is a model in development, but tt seems to me there will need to be some "Special Teams" efficiency taken into account at some point in the future.
Those yards made/given up have made a huge difference in SD's results.
James, I was refering to the probabilities in the Fifth Down this week. SF v DEN is based on home field for SF.
Brian, just wanted to say I stumbled onto this site last week and it is very interesting stuff you're doing here. Here's what I did...it was probably a stupid idea, but it did help me this past week.
I mixed your Efficiency and Success Rate rankings with Total Offense and Total Defense rankings from sportsbettingstats site. I believe they sort by total average points scored, respectively on each side of the ball. I then summed only the rankings. The lowest sum total was #1, the highest sum total was #32.
That seemed to help incorporate scoring and create somewhat of a power ranking for me.
I also agree that SD is rated too high...they've burned me too many times this year...
Again, some good stuff you're doing here and I hope you keep up the good work!
Math challenged individual here. Can someone tell me what the odds of a team with an 81% chance of winning would win 2 games in 7?
math challenged guy here - tried to figure this out on my own with the help of the internet and some basic high school probability.
If I'm right the chances of a team with a 81% success rate winning just 2 of 7 is .003412.
Am I right? If I am - I'm highly skeptical that SD really has an 81% winning probability.
What I like most about these efficiency ratings is it attempts to pull out all the luck factors to get a better predictions of future performance.
Football Outsiders and Brian seem to differ on special teams and I think turnovers as well with respect to luck. My gut would lean Brian's way.
But at times I think there is some plain old boring observation that provides powerful evidence. Last year I noted the top 3 NFC East teams were awful playing other top teams in the league (used DVOA rankings - Dall, NYG and Philly vs top 15 teams other than themselves). They were some god awful record like 3-12 negative 168 pts. The stats from FO and Brian suggested they were strong teams. But it just struck me that if that was true it seemed highly unlikely they would have such a horrible record against other good teams.
As if turned out Dallas was routed by Minnesota. Now that could just be bad luck - shit happens, but I think it more likely that the system was missing something or giving some other element too much credit.
I look at SD's stats my first instinct is they are so unlucky. But then I think the odds they are that unlucky to put up a 2-5 record against such a weak record and I think - there's something more here that the stats don't see.
jamaron--If you come back to read this comment, it is DEFINITELY SD's special teams. Their ST is SOOOOO bad that it is almost canceling out the value that their offense gives them. [They are a little better than the Panthers or Cardinals :P]