Team Efficiency Rankings - Week 14

Very little movement this week.

1 DEN10.730.5041
2 SF20.680.5033
3 HOU30.620.4777
4 CAR40.610.5296
5 SEA50.600.5268
6 NYG60.580.53520
7 ATL70.560.481717
8 NE80.560.49132
9 WAS90.550.51223
10 CIN100.550.481315
11 GB120.550.511214
12 DET110.530.491019
13 STL130.530.531512
14 CHI140.520.51302
15 TB150.500.52825
16 PIT180.500.48254
17 MIA170.500.492016
18 DAL160.490.521422
19 NYJ200.480.532411
20 BAL190.470.481818
21 SD220.460.492810
22 NO240.460.531130
23 BUF210.460.481921
24 CLE260.450.48269
25 IND250.450.461627
26 MIN230.430.502913
27 PHI280.400.502324
28 TEN300.400.492226
29 OAK270.390.492128
30 ARI290.380.52325
31 KC310.350.512729
32 JAC320.310.503131


  • Spread The Love
  • Digg This Post
  • Tweet This Post
  • Stumble This Post
  • Submit This Post To Delicious
  • Submit This Post To Reddit
  • Submit This Post To Mixx

26 Responses to “Team Efficiency Rankings - Week 14”

  1. nottom says:

    I was really hoping that the loss to KC would finally start moving Carolina down this list, but I guess not.

  2. Eric Peterson says:
    This comment has been removed by the author.
  3. jerome manson says:

    Wow with all the bad play calling by Gilbride, I am surprised to see the Giants at #6.

  4. Unknown says:

    can somebody point me which article explain the GWP stat and how it is calculated, Carolina ranking really is disconcerning.

  5. Anonymous says:

    Carolina is a real problem this year! I was the first to point this out 6 weeks back. Their win total is not important. But their pythag expected win% is 37% accounting for regression/luck this huge difference is not fully explained. What's crazy is their numbers when you adj. for SOS are good enough to win most SB's!!. We know the game changes so it's very possible efficiency in scoring or preventing scoring has become repeatable skill? Or perhaps
    will the increased emphasis on the passing game & the high variation/randomness/& value in pass interference calls
    the luck factor has increased. Just some ideas where model needs to change!

  6. Anonymous says:

    CAR has lost a lot of very close games to good teams. Their schedule so far ranks 4th toughest in the league, and they could easily be 8-4 instead of 3-9 with it.

  7. Anonymous says:

    Re: anon
    Easily 8-4 u can't be serious?! they have been outscored by 57 points! to be 8-4 (legitimately -not like Indy) they need a ~+60 point margin. That's a difference of ~8 points each & ever game sorry you don't make any sense.

  8. Ian Simcox says:

    How are Carolina losing? I know others are concerned about these rankings, but when you look at the dEPA v oEPA visualisation, the teams they are nearest are Green Bay (8-4), Baltimore (9-3), Buffalo (5-7) and Seattle (7-5).

    They hold pretty decent company when it comes to offensive/defensive performance - so it must be special teams that is killing them. I'll dive into this some more later.

  9. Ian Simcox says:

    Just to follow up - it looks like special teams that's the cause of Carolina's losses. Their total WPA is -3, oWPA is 1.5, dWPA -2.26, meaning there is an unaccounted for WPA of -2.24, which is by far and away the worst in the league.

  10. Anonymous says:

    I read somewhere that Carolina is 0-13 in coin tosses this year. Odds are so low that I almost assume its inaccurate. Still, it was a legit source, so it could be true.

    My non-ground breaking point is: A 13 game sample size is crazy small and records have little to do with the underlying odds. Sometimes the rarest of events do occur. You can lose 13 straight coin tosses, so to speak.

    Point differential only says so much. Can be skewed by defensive scores (see Chicago game) or a desperately playing comeback (see Tampa game).

    If you take Brian's numbers at face value, as I always have, the entire narrative of a season changes. The Redskins are 3-0 since their loss to the "lowly" Panthers. What a turnaround. But from my view, the numbers said that the Panthers game would be tough, and lo and behold it was.

  11. Anonymous says:

    Vikings have sure collapsed since the first few weeks when they were 5-2 or whatever.

  12. Anonymous says:

    I'll add that one stinker of a game (see Giants) can throw point diff down the tubes. Sometimes you line up and get tattooed. Who knows why it happens, but it does just the same.

  13. Anonymous says:

    Final point: Re Carolina. The model can never predict mistakes at the most crucial times, either.

    Cam Newton is 2-0 against the Redskins with zero picks. He throws a pick every 35 throws but didn't do it in either Redskin game. Huge in-game impact, but the underlying model can get much more granular than the simple int rate.

  14. Anonymous says:

    Re Anon;

    As I stated & we can clearly show luck (as defined by Brian)
    IS part of Carolina's difference. But doesn't explain all of it. The gap is simply too large.something meaningful is not being accounted. My guess is coaching is important component.As Belichuck has over performed exp. win totals from efficiency stats for 10 straight seasons & a total of 2 wins per season since 2000!

  15. Anonymous says:

    Echoing previous commentators: Football Outsiders has Carolina ranked last in the league in special teams. Anybody watching their games can see that it's exceptionally bad. Does ANS take into account whether a team is so bad in a couple metrics that it belies the traditional underweighting of that metric in overall rankings?

    Also, the Rivera's coaching seems overly simple and conservative for most of the game. Perhaps other coaches see this and make better adjustments in the 2nd half / 4Q?

  16. Eric Peterson says:

    any attempt to make the model work so that every team fits our preconceived notions (i.e. their record) for how good they are is simply overfitting the data. Of course you could decide where each team should be in the rankings and then create a model which makes all teams fall in that order. But that model would probably be a real turd at predicting future outcomes. This isn't to say that maybe the model could be improved. Maybe the model is wrong (all models are wrong...some are still useful). But the model isn't wrong b/c Carolina is ranked 4th. Anyone who's done any work with data in their lives knows that there will often be outliers...there will always be data points that just defy the model.

  17. Anonymous says: are ignoring the facts.. Brian has in-depth articles on the role of luck in NFL. We CAN account for this..& the difference of Carolina cannot be explained just by this. This is a clear opportunity to redo improve the model.

  18. Anonymous says:

    The analogy to Carolina's coin flip record this year is great. It's (roughly)like a team with .5 GWP being 0-13. Imagine how many people would be crying if TB was ranked 16th in the model with not 1 win to their name. Outliers happen folks.

  19. Anonymous says:

    Re: Anon;

    The key question is - is it more likely an outlier or a flaw in system. Again it is how large the Carolina error is.
    using your flipping coin analogy. 7 straight is 'heads' is not that uncommon 50% chance in 100 flips BUT chances of 13 straight over 800 to 1 or over 25 seasons. could be yes? likely No!

  20. Eric Peterson says:


    i'm not ignoring anything. my point has nothing to do with luck. rather my point is that regardless of how much of the variation of a model you an explain (using log regression for example) there will always be points that should be "defined" as the one outcome (win in this case) and they come out the other (lose). Even as you define reasonable parameters for what will likely 'cause' a win versus a loss, there will always be events that turn out the other way. if brian's model shows that what carolina accomplished in 12 of their games should make them 9-3, and they're 3-9, then there are 6 events out of 12 that didn't end up the way his model would have predicted. given that carolina is something like 0-12 in gms decided by less than 7 in the last 2 years, these are likely just events that were borderline gms that something didn't go carolina's way. anyone who's ever done a logistic regression can tell you that there are tons of points that should have been fitted as a 0 or as a 1, but in the end were the other. the real question is, what is the likelyhood of that happening 6 times in 12 observations out of only 32 possible teams? Is it enough of an outlier to consider changing the model? or just short term back luck?

  21. Anonymous says:

    Thankyou for your perfect contradiction..

    "my point has nothing to do with luck"
    "Is it...just short term bad luck?"

  22. James says:

    If you think the model is flawed or broken because of Carolina, how do you suggest improving the model without overfitting it? Let's work on some constructive critism.

    Also, every year for the past few there's been one team that's stuck out (Carolina, San Francisco, Atlanta, San Diego), and to me it seems that if the model predicts 31 teams reasonably well then the model is working and the outliers are just random outliers.

  23. Anonymous says:

    Ive already made numerous suggestions but Brian and others haven't responded.
    To repeat:
    1. We know certain teams CAN out perform their expected win %
    consistently. NE has done this every year for the past 10 straight. This isn't luck - so why is it happening?

    a) their efficiency stats are deflated since they play
    a high number of blowout games (i.e. score effects)

    b) game management play calling &coaching decisions in critical situations

    Any others?

    Let's start with the fact that we know the model is inefficient in this area from the Patriots data.And stop offering red herrings such as outliers/luck etc in defense

  24. Anonymous says:

    I do think it is possible that certain teams can "fall through the cracks" and just aren't predicted well by the model - remember it is based on league wide data. Which means that we are assuming you essentially operate like every other team.

    This is very very good assumption. Remember, the ANS model is statistically indistinguishable from the vegas opening line in picking straight up winners (see "prediction accuracy"). This means that for every team that falls through the cracks (the model predicts poorly) there are teams that are represented extremely well by the model. (note that vegas, to the best of my knowledge, doesn't typically agree with the ANS model on those "fall through the crack" teams)

    Also, besides the patriots - the vast majority of teams might still fit within what one would expect based on a binomial distribution. As a rough estimate: Google binomial distribution calculator, plug in the teams GWP (assume all of their opponents are .50 GWP) plug in their wins as "success" or whatever its called and remember this:
    *one would be just as surprised to see a team way exceed their projected win total as be way below their projected win total
    *there are 32 teams - and here in the comment boards we always pick the most extreme ones.

    as for the patriots; search this site for patriots - there is plenty of discussion as to what may be happening.

  25. MSL says:

    Been perusing some of the comments regarding the Panthers. Hopefully, I can help explain--as someone who's watched all the Panther games this year:

    Panthers can be easily 7-5 this year. There were at least 4 games (not including others they blew late in games) where the Panthers were leading by 3-8 pts in the 4th quarter, with less than 1-2 min left, while the opposition had no Time outs and had to go more than 80 yards up the field to kick a FG or score a TD, and Miraculously beat the Panthers due to horrible coaching and game management.

    And as others have said:

    Yes, Panthers coaching is very conservative (on both offense and defense), and is particularly flummoxed during 2nd and 4th quarters as the opposition makes successful adjustments. Carolina's defense during 4th quarters (particularly the last 6 min) has been horrendous.

    Yes Carolina has the worst special teams and it's not included in this model. Their ranked dead last in the NFL.

    Yes, There are other models that include special teams, where the Panthers rank around 15-18.

    I hope that helps, explain?

  26. Anonymous says:

    Carolina loses because they have coaches with bad instincts that can't make adjustments during a game.

    Anyone who watches them play regularly knows this.

    Bring on Bill Cowher, please.

Leave a Reply

Note: Only a member of this blog may post a comment.