Correction--There was a bug in the ointment the first time around posting the rankings this week. These numbers have been updated with the correct
There are a few sore thumbs every year in the rankings. A few weeks ago I wrote about how the then 2-8 Panthers were much better than their record indicated. Since then, they went 5-1 to finish 7-9. Do I think CAR is really the 4th best team in the league? No. But they were the 4th most efficient on offense and defense in 2012.
The following week I wrote about how the Ravens were a mystery. They were ranked 19th in the rankings but held a 9-2 record. I was grasping to explain their good fortune. Maybe no explanation was needed, as BAL went 1-4 since then. (1-3 if we throw out week 17's loss in fairness.)
It might seem like I'm cherry picking the model's 'hits' and ignoring its 'misses.' I'm obviously not pointing out how the model ranked PHI 2nd in its first iteration after week 3. Oops, I guess I just did. But it's not week 3 anymore, and we have a lot more information now.
There is one more sore thumb to be addressed. That's the 11-5 Colts, who are ranked 24th in efficiency. Their Generic Winning Probability (GWP) is 0.44, which if correct would make an 11-win record highly unlikely. It's very possibly they're better than 24th, but notice that their opponent average GWP is 0.46, so despite a low efficiency ranking, we should not be too surprised to see IND end up its winning record.
[Edit: 0.46 Opp GWP might not seem like much, but that's not far from the equivalent of having an additional home field advantage effect for every game, ...if that makes any sense.]
In the second table below, we can see where IND ranks in the core efficiencies. Despite Andrew Luck's promise, his net passing efficiency and interception rates are both slightly worse than average. IND's defensive passing numbers are also worse than average. They've been somewhat successful in the running game, but not nearly enough to make them a top team. These numbers have come against the easiest schedule in the league this season.
IND has been able to outperform their efficiency because of clutch play--good performance in high leverage situations. On offense, their EPA is ranked 11th while their WPA is 3rd. On defense, their EPA ranks 28th while their WPA ranks 21st. On both sides of the ball, IND has happened to save its best performance for when it matters most.
I had intended this next paragraph to be the case for IND to be a true 11-win team. I expected to illustrate how Andrew Luck was a rookie and that the offense's mediocre passing numbers was merely due to early season shakiness. But I was surprised to see he performed better in the first half of the season than in the second in terms of AYPA, WPA per Game, and EPA per play. Normally we'd expect rookie quarterbacks to improve over the course of season, but Luck has played consistently well in his first year.
So are the Colts really just the 24th best team? Probably not. But like everything else, the truth lies somewhere in the middle, somewhere between their 11-5 record and their 0.44 GWP.
MIN might be the bigger curiosity with 0.45 GWP and a 0.51 Opponent GWP, but that's a topic for another day. Here are the final efficiency rankings for the season. Raw stats are in the second table below. Click on the table headers to sort.
| 1 |  DEN | 1 | 0.72 | 0.47 | 2 | 2 |
| 2 |  SF | 2 | 0.69 | 0.51 | 3 | 3 |
| 3 |  SEA | 3 | 0.66 | 0.52 | 4 | 6 |
| 4 |  CAR | 4 | 0.62 | 0.52 | 7 | 7 |
| 5 |  GB | 5 | 0.59 | 0.50 | 8 | 9 |
| 6 |  NE | 9 | 0.58 | 0.49 | 1 | 29 |
| 7 |  ATL | 6 | 0.57 | 0.50 | 13 | 17 |
| 8 |  WAS | 8 | 0.57 | 0.50 | 5 | 24 |
| 9 |  HOU | 7 | 0.57 | 0.47 | 12 | 11 |
| 10 |  NYG | 11 | 0.54 | 0.52 | 9 | 25 |
| 11 |  CHI | 12 | 0.53 | 0.51 | 27 | 1 |
| 12 |  STL | 10 | 0.52 | 0.53 | 15 | 10 |
| 13 |  NO | 16 | 0.51 | 0.53 | 6 | 28 |
| 14 |  CIN | 15 | 0.51 | 0.47 | 19 | 8 |
| 15 |  DET | 13 | 0.51 | 0.50 | 10 | 19 |
| 16 |  DAL | 17 | 0.51 | 0.52 | 11 | 22 |
| 17 |  PIT | 14 | 0.51 | 0.47 | 22 | 4 |
| 18 |  TB | 20 | 0.49 | 0.52 | 14 | 23 |
| 19 |  BAL | 18 | 0.48 | 0.50 | 17 | 16 |
| 20 |  MIA | 19 | 0.47 | 0.49 | 21 | 18 |
| 21 |  BUF | 22 | 0.46 | 0.48 | 18 | 21 |
| 22 |  NYJ | 21 | 0.45 | 0.49 | 29 | 12 |
| 23 |  MIN | 23 | 0.45 | 0.51 | 25 | 13 |
| 24 |  IND | 27 | 0.44 | 0.46 | 16 | 30 |
| 25 |  SD | 24 | 0.43 | 0.49 | 28 | 15 |
| 26 |  CLE | 25 | 0.43 | 0.49 | 24 | 14 |
| 27 |  PHI | 26 | 0.42 | 0.51 | 23 | 20 |
| 28 |  OAK | 28 | 0.41 | 0.49 | 20 | 27 |
| 29 |  TEN | 29 | 0.39 | 0.48 | 26 | 26 |
| 30 |  ARI | 30 | 0.36 | 0.55 | 32 | 5 |
| 31 |  KC | 31 | 0.32 | 0.51 | 31 | 31 |
| 32 |  JAC | 32 | 0.31 | 0.48 | 30 | 32 |
| ARI | 4.5 | 33 | 3.5 | 1.7 | 6.0 | 56 | 4.4 | 0.42 |
| ATL | 7.0 | 36 | 2.3 | 0.5 | 6.7 | 56 | 3.6 | 0.21 |
| BAL | 6.3 | 39 | 2.0 | 0.6 | 6.1 | 56 | 2.3 | 0.53 |
| BUF | 6.0 | 45 | 3.3 | 2.2 | 6.1 | 53 | 2.2 | 0.43 |
| CAR | 7.0 | 44 | 2.4 | 1.3 | 6.0 | 57 | 2.0 | 0.42 |
| CHI | 5.7 | 36 | 3.3 | 1.0 | 5.4 | 61 | 4.1 | 0.40 |
| CIN | 6.1 | 40 | 3.0 | 1.2 | 5.6 | 57 | 2.5 | 0.42 |
| CLE | 5.7 | 40 | 3.2 | 1.1 | 6.1 | 58 | 2.8 | 0.48 |
| DAL | 6.8 | 40 | 2.9 | 1.2 | 6.8 | 57 | 1.4 | 0.42 |
| DEN | 7.4 | 46 | 1.9 | 1.5 | 5.2 | 59 | 2.9 | 0.38 |
| DET | 6.4 | 41 | 2.3 | 1.8 | 6.2 | 56 | 2.0 | 0.44 |
| GB | 6.6 | 40 | 1.4 | 0.9 | 5.7 | 57 | 3.2 | 0.44 |
| HOU | 6.6 | 40 | 2.3 | 0.4 | 5.8 | 59 | 2.6 | 0.41 |
| IND | 6.2 | 44 | 2.9 | 1.1 | 6.7 | 58 | 2.2 | 0.38 |
| JAC | 5.4 | 38 | 2.9 | 1.2 | 6.9 | 54 | 2.2 | 0.46 |
| KC | 5.3 | 42 | 4.2 | 2.1 | 7.2 | 58 | 1.5 | 0.37 |
| MIA | 5.9 | 40 | 2.6 | 1.7 | 6.2 | 60 | 1.7 | 0.35 |
| MIN | 5.3 | 41 | 2.5 | 1.3 | 6.0 | 58 | 1.6 | 0.40 |
| NE | 7.0 | 48 | 1.4 | 0.7 | 6.9 | 55 | 3.4 | 0.38 |
| NO | 7.2 | 41 | 2.8 | 0.6 | 7.4 | 55 | 2.5 | 0.42 |
| NYG | 6.8 | 43 | 2.8 | 0.8 | 7.2 | 53 | 3.9 | 0.29 |
| NYJ | 5.4 | 42 | 3.9 | 2.2 | 5.8 | 57 | 2.2 | 0.34 |
| OAK | 6.2 | 35 | 2.5 | 1.3 | 6.9 | 61 | 2.1 | 0.46 |
| PHI | 5.7 | 46 | 2.4 | 2.7 | 6.7 | 56 | 1.6 | 0.39 |
| PIT | 6.2 | 35 | 2.4 | 2.0 | 5.3 | 58 | 1.9 | 0.48 |
| SD | 5.7 | 36 | 2.8 | 1.4 | 6.1 | 57 | 2.5 | 0.40 |
| SF | 6.9 | 48 | 1.8 | 1.0 | 5.3 | 62 | 2.5 | 0.48 |
| SEA | 6.9 | 45 | 2.5 | 1.0 | 5.4 | 53 | 3.2 | 0.46 |
| STL | 6.0 | 42 | 2.5 | 1.0 | 6.0 | 58 | 3.1 | 0.48 |
| TB | 6.7 | 40 | 3.0 | 0.8 | 7.3 | 63 | 2.9 | 0.40 |
| TEN | 5.7 | 37 | 3.0 | 1.6 | 6.6 | 57 | 3.4 | 0.41 |
| WAS | 7.2 | 48 | 1.8 | 0.7 | 6.8 | 54 | 3.3 | 0.49 |
| Avg | 6.2 | 41 | 2.6 | 1.3 | 6.3 | 57 | 2.6 | 0.41 |
IIRC Baltimore has had an exceptionally good special teams year. Something that naturally comes up for me when Baltimore and Minnesota are mentioned is that both teams have large home field advantage. Have you ever tried doing home/away splits on the ratings?
Is there an easy way to see the Team Efficiency Rankings for just the former/latter half of the season? That way we could see which teams have improved over the course of the year (or perhaps it would just appear that way due to variance).
Denver may be #1, but Peyton isn't known for performing in the postseason. He'll lose in the AFC championship to Brady. Seahawks/9ers in NFC championship
Is it 2003 or 2013?
Peyton in the playoffs: 0.21 WPA/Game, 0.17 EPA/Play
Tom Brady in the playoffs: 0.18 WPA/Game, 0.11 EPA/Play
How do you calculate opp gwp...I just added up Minnesota's 16 opponents gwp and divided by 16 and came up with .515.
Minnesota had 11 of it's games against teams with gwp's above .51 - which means they played 11 of their games against teams in the top half of these rankings. 7 of those 11 were on the road. Hardly an easy schedule. They were also 7-4 against those top half teams.
IIRC, injuries aren't factored into the efficiency ratings. With the caveat of anything can happen in a 1 game playoff series, I like SF & SEA for the NFC championship—but SF is looking pretty beat-down right now. How does Justin Smith and Vernon Davis's injuries factor into a SF/SEA game?
Brian, looks like the last week column is screwed up (comparing w17 to final). It might be using last week's last week values.
I'll check the last week column. Thanks.
Opp GWP isn't simply a calculation of the average of all the opponent's GWP. That method would include games against the team of interest itself.
Last wk column fixed.
You changed our model from using YPC to SR for the offensive running game. I know that you did this with good reason but maybe this year it does not give Minnesota enough credit. Peterson was averaging 6 yards a carry and the team as a whole was 2.3 standard deviations above average.
San Fran all day!!
Can someone explain how Arizona is the only team with a above average strength of schedule?
That's gotta be a bug. Will check tonight.
I posted this last week, but I guess it was after everybody had moved on. Sorry to repeat, but I'm really curious about the answer!:
"After digging through the "how to's" of calculating team efficiency I had a question:
Does the model essentially assume a schedule strength constant of 1? It looks like the Logit of the opponent GWP is simply added in, without any multiplier on how important it is.
I guess putting it another way: If NE puts up 7.5 yards per attempt against teams with a 40% GWP - what should be expected when they play a 50% GWP team? Is it closer to 7.4 or like..5.4 yards per attempt? This is essentially what the model is doing, right? Inflating/deflating team stats based on who they played..?
Is this something that has been studied? Does it even make sense to put a constant in front of the opponent GWP Logit? How did you determine how important the quality of opponent is?"
I obviously do not understand the opponents GWP.
Detroit plays 6 games against teams that went 31-17, 13-3 falcons, 12-4 houston, 11-5 colts, 11 win SF, 11-5 seattle, and wrap it up with the poorer teams arizona (5-11), rams (7-8), titans (6-10), phi (4-12) , jac (2-14).
how that that remotely come up to the opponent's GWP of 0.47?
At footballoutsiders Aaron Schatz called the Colts the worst 11-5 team of all time. (Really, he means since 1991, but all time sounds better, and I suppose there weren't 11-5 teams before 1978 anyway.) Much more straightforward rating systems agree: pro-football-reference's simple ranking system, which just looks at point differential and opponents played, has them at -4.7, which is consistent with a 6-8 record.
As you say w.r.t. earlier predictions this year, there's really no need to try to explain away why the model is "wrong" here. When three systems that use three different means of measuring things come up with the same answer, it seems likely that it reflects something fundamental. Bet the under on the Colts next year, especially if they manage to win this weekend.
At footballoutsiders Aaron Schatz called the Colts the worst 11-5 team of all time. (Really, he means since 1991, but all time sounds better, and I suppose there weren't 11-5 teams before 1978 anyway.) Much more straightforward rating systems agree: pro-football-reference's simple ranking system, which just looks at point differential and opponents played, has them at -4.7, which is consistent with a 6-8 record.
As you say w.r.t. earlier predictions this year, there's really no need to try to explain away why the model is "wrong" here. When three systems that use three different means of measuring things come up with the same answer, it seems likely that it reflects something fundamental. Bet the under on the Colts next year, especially if they manage to win this weekend.
Tarr,
one might point out that all three "models" are wrong, and give the benefit of the doubt to what actually happened on the field. The colts really did win 11 games.
One person's wrong model is simply another man's Clutch Factor Index! The colts have a very high CFI, which explains everything :)
Andy-Yes. There is no coefficient for opponent strength. But it's not an assumption. It's a property inherent in the logit model.
And thanks to 'nottom' for catching the buggy opponent GWPs this week. The end of the regular season throws things off a bit and my script had simply divided by 17 games rather than 16.
Minnesota is able to do well in spite of rank because of their ability to sustain prolonged drives on both sides of the ball without a score once they have the lead to leverage a small lead into a win. They run a very short passing game even more so than most west coast passing games with many 5 yard routes or less. They also run the ball VERY well against 8 and 9 man fronts which defenses go into when they are behind and know the other team is likely to try to run the clock and can generate big plays against 8 and 9 man fronts. On defense they have been mostly a "bend but don't break" defense, even though against a high powered team like the Packers last week (and the first time they played them) they do manage to just barely break after a very long drive.
The probability of winning is significantly greater when they have the lead than most teams, but significantly worse when they are behind than most teams because of their inability to generate the big play passing or work the ball downfield.
But their record and style does open up your eyes to what's possible even at times without the ability to generate a big play if you can leverage a lead and work the clock.
"The probability of winning is significantly greater when they have the lead than most teams, but significantly worse when they are behind than most teams because of their inability to generate the big play passing or work the ball downfield."
I have not crunched the data to prove that MN has longer drives on either side of the ball, or proof that their change of winning is higher, but based on upon examining their style of play and routes they run and watching all of their games this year and following them closely I am confident they play a lower variance strategy when ahead that includes a limited number of big plays on each side on most games, and playing a higher success, lower risk of turnover strategy with the lead. It is probably worth examining more in depth for those interested but is just technically a hypothesis for now.
Additionally, Blair Walsh has been extremely good at converting short drives into 3 points due to his ability to kick very long fieldgoals (50+ or even 55+), and they are built to keep the game close as well.
They can have fewer conversions and lower efficiency as a result and still put up 3 points on offense. The defense has been a bit lucky with turnovers.
After watching Andrew Luck I have changed my mind.
Clutch exists... Eli Manning has "statistically" been insanely clutch, but he never convinced me because the defense would give him more opportunities. But Luck? Whole different deal. Different mindset, different approach, more aggression.
Clutch is real. It's called an adrenaline rush... More oxygen to the brain, more bloodflow, better performance. Heightened performance after an adrenaline rush can be scientifically proven.
Unfortunately, it also goes against you when the defense has the same adrenaline rush, and as such with a league full of parody, "clutch play" is perhaps normally distributed... and as always there are a lot of variables and luck involved.
But Andrew Luck? he's an outlier.
Giants set the bar last year for Lowest known Regular Season End GWP of a Super Bowl winner at 0.67. The 2007 Giants probably were lower but I'm not entirely sure. If going by that rule of at least 0.67 GWP the only 2 teams we have here is Denvah and San Fran. Well, I'm goin' with th' Denvah Broncoes to at least show up in the Super Bowl.
11 wins or more during any part of the season in a single year usually translates to a Super Bowl appearance. And since 1969 there hasn't been consecutive years where multiple teams with 11 or more wins in a row haven't made a Super Bowl trip.
Brian -
My apologies if this is obvious and I'm missing it, but how are you getting your OFUM% stat? Here's what I've got for Washington (data from NFL website):
Total fumbles: 26
Total offensive plays: 994
26/994 = 2.6%
But you have 0.7%.
I've tried these other combinations, but still can't get your 0.7%:
rushing fumbles/rushing attempts: 13/519=2.5%
rushing fumbles/total off plays: 13/994=1.3%
What am I missing? Do I need to add in kickoff plays, extra point plays?
Thanks in advance!