Team Efficiency Rankings - Week 5

A lot of movement from week 3 to 4, as we'd expect. But actually, the median absolute (in either direction) movement was only 2 spots. The big movers were TEN, CIN, ARI, AND SEA.

How on earth did TEN move up 9 spots after a 38-14 divisional beatdown? Aside from turnovers, they played statistically well, especially on offense. Midway through the third quarter they were only down 7 with possession and a .31 WP, but a pick-6 quickly let the game get away. Turnovers are discounted by the model because they are so random. The biggest factor for TEN's climb up the rankings was opponent strength. HOU appears so dominant right now that just showing up on the field moves their opponent 5 spots.

CIN is climbing the ranks quickly because they put together a complete game against the Jags and their stomping at the hands of BAL is being averaged out by their performance since.

ARI is everyone's record-gazing outcome-focused darling of the moment. Their 4-0 record is deceiving, and I can already sense that they might be the post child for bad power rankings this season, like ATL was a couple years ago. First, they're 2 plays away from being 2-2. Second, their offense is terrible. They are tied for 28th in passing efficiency and dead last in run success rate.

Despite the apparently close score, SEA was soundly beaten by a very poor STL team. Russell Wilson posted just 0.2 adjusted net yards per attempt on Sunday. Laying an egg against the 23rd ranked team in week 4 will drop a team hard.

Typically, teams are bunched in the middle, so a relatively small change in their GWP can translate into a big movement in ranking. Also keep in mind that when a team's previous opponents move up or down, so does its own opponent adjustment. For example, the teams that have played CIN so far will get a bump in their rankings due to CIN's improvement. In other words, CIN is probably a better team than their previous stats had indicated. This early in the season, both of these factors can have a large effect.

With that, here are the efficiency rankings going into week 5. And as always, they are completely perfect and I personally agree completely with each and every ranking.

The inputs for the rankings are in the second table below, so if you have any questions why a team is ranked where it is, start there. Click on the table headers to sort.

Yes, that's right. The Saints are dead last.


RANKTEAMLAST WKGWPOpp GWPO RANKD RANK
1 HOU10.750.5243
2 ATL40.700.55117
3 DEN30.690.5765
4 CAR60.640.48319
5 BAL50.630.55113
6 PHI20.620.54171
7 SF70.620.47134
8 NE140.600.52520
9 CHI180.570.45226
10 NYG90.560.53226
11 MIA170.560.521614
12 CIN210.510.45722
13 MIN120.500.462110
14 DAL110.500.471417
15 GB100.500.471015
16 SD150.490.49299
17 ARI80.480.54258
18 DET220.480.51923
19 TEN280.480.581525
20 BUF130.470.461230
21 CLE190.470.56322
22 OAK260.440.532027
23 STL300.430.452312
24 NYJ200.410.512616
25 PIT250.390.511924
26 SEA160.390.482811
27 KC270.370.492728
28 JAC240.360.533021
29 IND290.360.481831
30 WAS320.350.40832
31 TB230.350.523118
32 NO310.310.472429



TEAMOPASSORUNSR%OINT%OFUM%DPASSDRUNSR%DINT%PENRATE
ARI5.4292.21.55.9642.50.46
ATL6.8341.40.05.8505.30.22
BAL7.4441.90.56.9622.50.59
BUF7.1455.62.06.8532.70.34
CAR8.4444.52.16.4552.00.37
CHI6.4345.10.55.5586.60.36
CIN8.0373.11.46.0410.70.44
CLE5.3364.21.16.3613.60.48
DAL7.0385.01.56.0631.00.52
DEN6.9451.91.85.9691.40.47
DET6.6392.20.96.4570.00.49
GB5.8491.90.95.7533.10.59
HOU7.4420.80.44.9543.70.34
IND6.3353.31.47.1571.10.45
JAC4.6480.91.77.2571.40.40
KC5.7454.33.57.4571.80.34
MIA6.5434.21.86.5722.90.36
MIN6.1390.01.95.4610.60.47
NE7.3460.60.87.3614.10.31
NO6.4332.60.58.1581.50.44
NYG7.7362.50.57.9545.00.36
NYJ5.8303.12.26.3473.30.46
OAK6.1321.20.67.4620.00.28
PHI6.6473.92.65.5684.20.52
PIT6.6330.81.26.2501.10.66
SD6.3373.21.05.8613.20.43
SF5.7530.91.45.3722.10.51
SEA4.8444.00.55.4602.10.61
STL5.6393.21.16.1516.00.41
TB5.9383.40.68.2653.80.50
TEN6.6352.82.47.5540.70.38
WAS7.3460.80.48.1563.20.69
Avg6.4402.71.26.5582.60.45

  • 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

43 Responses to “Team Efficiency Rankings - Week 5”

  1. Anonymous says:

    Why on God's green earth is Carolina still so high? Every single other team seems perfect to me though

  2. Joshua Northey says:

    It is funny that this is what 5 or 6 years running that the mainstream NFL media has wildly overestimated how good Dallas is. Pretty much every year since the Quincy as starter year, you hear about how great Dallas is going to be...I always wonder why? The season always seems to come down on my side.

    There are never enough pieces there. It is like because it is Dallas everyone takes each players evaluation and rounds up so the sum is way more than the parts. If you took this same personnel and stuck them on the Browns everyone would think they sucked.

  3. Anonymous says:

    could you elaborate on what you mean by Seattle being dominated by the Rams? Was it that Bradford dramatically outplayed Wilson and the fact that Seattle ran better doesn't compensate? Is the offensive interception rate damning in this instance in spite of its relatively random nature?

    It seems like the kicker outplayed the expected points on the 58 and 60 yarders. maybe the expected points would be 2 and he got all 6. that would seem to move the needle in Seattle's favor.

  4. David Cooper says:

    I have been trying to follow along with the math at home but I can't get the coefficients for the regression to match the GWP posted. I know there is some dissonance due to opp GWP not being factored in but is there somewhere I can find those numbers?

  5. Anonymous says:

    I believe you've said that you update the regression coefficients every year, I'm wondering what they are this year? (For OPASS, ORUNSR%, etc...)

    I wanting to try and use them to determine weighting for edges on a graph of teams to do a random walker type of ranking. (see http://arxiv.org/abs/physics/0310148)

    -pack_mentality

  6. Dave says:

    Brian- this post should be tagged so the links up top point to these latest rankings.

  7. Anonymous says:

    Brian could you improve your predictions by improving the strength of schedule adjustments? For example, calculate an expected number for net passing and success running based on the strength of your opponent, then adjust for under or overperformance?

  8. Anonymous says:

    "Why on God's green earth is Carolina still so high? Every single other team seems perfect to me though"

    Did you not see them nearly beat ATL?

  9. Anonymous says:

    did YOU see them get dismantled by the giants on home turf?
    Yes..1) Car is too high..2) Atlanta is overrated -
    3) Wsh is underrated
    they are not elite in any key stat -have very low pen yds which should regress and very high dint and no ofumbles
    All will regress. one weakness in Brian's system is that
    he counts turnovers when studies have shown they are essentially random.

  10. James says:

    Anon,

    Carolina leads the league in THE key stat: net yards per pass attempt. Also Brian's system heavily regresses turnover rates so you clearly don't know what you're talking about.

  11. Anonymous says:

    The ability to take criticism - to be open to improving one's system is key to ongoing success.
    Check out vegas futures here:

    http://www.vegasinsider.com/nfl/odds/futures/

    Why are Carolina so low? 50 to 1 for conference? 12th best? seriously? do you think those that make their living at predictions would leave themselves so exposed?

    Brian has been amazing giving insight and perspective but his prediction system has flaws I was just pointing them out...:)

  12. Anonymous says:

    Brian;
    one question? Why don't you include any stat on a team's ability to score td vs FG? It is obviously a huge discrepancy?Are you saying that by your research there is no observable repeatable skill in scoring a td vs not scoring one? for example, I know Pro football ref adjusts net pass giving credit for passing tds?
    Just curious..thanks
    Bob


  13. Anonymous says:

    James;
    Carolina was down a large amount early at home vs Nyg
    they were able to rack up pass yds on softer D.
    Score effects need to be applied.

  14. Michael Beuoy says:

    "The ability to take criticism - to be open to improving one's system is key to ongoing success."

    Anonymous - Agreed, as long as it's constructive criticism.

    What stat do you suggest Brian use to improve his model? Which coefficient in the regression should he adjust?

  15. Jerome "The Giant" Manson says:

    My Giants are about right. Our secondary is being killed with injuries, but we can move up a little this week because we get the Browns.

  16. Anonymous says:

    Michael;
    I have already suggested 1) score effects, adjusting for state of the game. - We know net passing yards Is an important stat BUT we also know teams leading will make the logical decision to give up passing yards/yards and yards in general for time off the clock.(Brian uses this fact in generating EP and WP). Therefore expected net passing yards
    (and run success rates) based on score and time remaining.

  17. James says:

    Anon, against the Giants the Panthers gained 267 passing yards on 33 attempts. Include 21 yards lost on 2 sacks and that's 7.6 NYPA. For the whole season the Panthers have a 7.9 NYPA, which means the Giants game hurt their overall statistics.

    So while you may have a valid point about game states, your example is contrary to your point.

  18. Anonymous says:

    The splits for cam newton definitely make a point that CAR's YPA may be inflated by game stats.

    1Q: 6.58 YPA
    2Q: 8.43 YPA
    3Q: 9.49 YPA
    4Q: 12.54 YPA

    Also the majority of netwons attempts have come when trailing where both his completion % and yards per attempt are markedly better. Granted we are dealing with small sample sizes.

    Also Carolina's passing success rate is close to league average rather than near the top of the league.

  19. Anonymous says:

    James ;
    sometimes it does pay to watch the games...
    thanks to Anon for providing more evidence!

    I have watched 70% of all games played and every Car game and they are simply not the 4th best team in the league.
    vegas reflects this having them 12th best in conference?!
    and only 3 point fav. vs Seattle this week even though Seattle are awful in opassing?!


  20. Anonymous says:

    "The biggest factor for TEN's climb up the rankings was opponent strength. HOU appears so dominant right now that just showing up on the field moves their opponent 5 spots."

    Does this create a feedback loop, where HOU's OppGWP is improved by playing them, which improves HOU, which…?

    I don't have the statistical modeling experience to examine the model and tell.

    Thanks

  21. Unknown says:

    Let me make ask a question and then make a comment.

    Brian--I know your model creates probability estimates of game outcomes (which you post on Fifth Down Blog). Does it have the ability to project a final score? Since EPA is a major component of your model and efficiency rankings (and live game probability model), my intuition would think it does. But I won't pretend to completely understand your model and how it works. It's slowly, but surely becoming more intuitive, but I still find myself re-reading the same past posts--which has helped me improve my understanding. But it remains limited. But would be a nice addition and test to see how accurate it is at predicting final scores.

    On the Carolina ranking, lest the critics forget, CAR lost two games they easily could have won--Tampa and ATL. They lost the ATL game cause Ron Rivera is chicken sh*&. According to some data I saw, CAR was 16 of 18 last year on 3rd or 4th and 1. And Rivera punts... Granted, I'm sure he felt like a genius when the ball was downed at the 1 yard line. But we know how that worked out.

    Not suggesting CAR should be #4, but I think there is a tendency to overweight one's perception of CAR based on the performance on the national stage against NYG. While it is true that the Dolphins could say the same thing (and not coincidentally, they are probably rated higher than most pundits would rate them), they are two plays away from 3-1.

  22. Unknown says:

    To follow up on why I ask about the score. There is a site that charges a BOATLOAD of $$$s for a similar "live" prediction models (with probability of winning after each play, etc.). However, they also include a predicted score--which I find interesting. But what also was interesting is how closely their live probabilities moved in sync with ADV NFL STATS. Certainly, there was some variance, but it was remarkably similar. Essentially, they're using some kind of Monte-Carlo mean outcome approach (or so they claim). I think they probably are using a regression model (and the rest if marketing to make it sound "cool" to the masses and understandable--eyes tend to glaze over when you say the word regression :) But just given my limited understanding of how your model works, I thought a score outcome could be part of the equation. Maybe not. I don't know. Curious.

  23. Anonymous says:

    More evidence Car is overrated -
    O believe they have +14% Expected O prob added
    BUT -14 % D prob added that makes them 50%
    They have played slightly under average sos 48% so
    they are a little below middle of the pack. Thats where I have them. That's where vegas has them. I feel Brian's expected Points Prob. seems to pick up something that his rankings don't (early in the season) Perhaps to the score effects I raised?

  24. Kulko says:

    @ Carolina
    I dont think before week 8 or so Brians Model reaches full maturity, as single game effects and opponent strength are hard to gauge on a small sample size. SO I would guess Caroolina being overrated is probaly true, but maybe less then you think.

    Looking at the details you see a top Offense (YPA Top, RSR good) undone by high turnover rate. The first part is highly predictive, the second part much less so, so you would think that last years success will eventually return.

    The defense OTOH is rated average to good again with a bad turnover rate, so should be average at worst in the long run.

    As for the garbage time argument, firstly all efforts by FO have shown that discounting the effects of garbage time actually lowers the predictivity of their models, and I would stringly believe this to be true for Brians model too. Apart from that how many Garbage time did you see th Panthers play in? Maybe 2 Quarters against the giants, but all there other games have been too close. For a defense to really let up the pressure.

  25. Brian Burke says:

    CAR is ranked so highly because of their league-leading 8.4 net YPA. They have a very high int rate, but that's getting heavily regressed.

  26. Anonymous says:

    Kulko:/Brian;

    "I think the efficiency numbers don’t properly gauge its true ability. Green Bay’s defense is 31st in the league in efficiency, but 13th in Win Probability Added, a more advanced estimate of its effect on game outcomes. "
    (from NY GB division playoffs 2011)

    Here Brian acknowledges a case where Win Prob. 'can' tell us something efficiency stats don't. And Brian adjusts his efficiency prediction accordingly. This is exactly my point with Carolina.
    Of course In Football one is "always' struggling with a lack of sample size.two quarters of stats (at this point) represents 12.5% of total.

    ...But Carolina at ~14% above is quite a bit off.Thanks for all responses.. I was only pointing out a potential way of improving these predictions early in the season?!





  27. Luca says:

    Is Carolina also being overrated because the model so heavily favors yards per attempt, and they've made fewer attempts? Carolina is 31st in the league in passing attempts, ahead of only Seattle. So while they have succeeded when they do pass, they don't pass very much.

    I mean, is the Rams' punter (who threw the TD pass on that fake field goal) not the most efficient passer ever? Every time he's thrown a pass this year, he has scored a touchdown! But being the most efficient is not the same as being the best.

    Maybe these rankings need to be interpreted as just that - EFFICIENCY rankings, not "power rankings." It's obvious that Carolina is not the 4th best team in the NFL right now, and they probably never will be this season.

  28. Unknown says:

    When I first posted my comment, I failed to complement Brian on what an AMAZING site this is. New reader and that was my first post. I've spent the better part of the last 3 days going through historical posts. I've never regretted quitting mathematics after DE and Calc III more than I do now. Gonna need to refresh my skills so I can understand this at a deeper level--though you do an INCREDIBLE job of making things simple to understand. I don't wish to knock FO--which is a great site, too--but I struggle to understand their methodology--and I feel their explanations are often convoluted. You do a great job explaining things here. Many thanks for all the work and helpful insights.

    I'll repeat what I said before. CAR is 2 plays from being 3-1. Would there be so many complaints about their ranking if CAR record was 3-1? What if we're not that generous and just say they "should" have beat ATL? At 2-2 with a recent win over ATL, would there be so many complaints? I don't know how much of the math is "shared" with Quant Coach's methodology but QC also has CAR ranked very high. Again, not saying I think they are the 4th best team in the league. But I think there is a too much focus on the randomness of their loss to ATL and even to TB. Full disclosure: I refer to the CAR QB as Sham Newton. So not a CAR apologist at all. And I tend to think CAR is overrated by the media. But that's what makes this type of work so interesting. When the results confirm/deviate from conventional wisdom (and/or my own intuition), it generally means there is something new for me to learn about how to think of the game.

    To me, the greater variance of the rankings and "perception" is the Saints--and there's been virtually nothing mentioned about them in the comments. I totally understand why they are ranked 32--in terms of the math. But consider, most fans/analysts would probably think the Saints could beat ANY team in the league. Whereas I doubt there'd be too many who think JACK or KC (each ranked several spots higher) would beat any of the top 5 or 6 teams in the league. That's obviously a reflection of the QB--and our perception of the relative QB's abilities. Which begs the question. Is the "market" (based on point spreads) behaving irrationally by installing the Saints as 3 point favorites against a superior SD team? Based on the probabilities from the forecaster, the Saints have about a 33% chance of winning the game--let alone covering 3 points. Answering those types of "market efficiency" questions that is of greatest interest to me. At what point does the data reach "critical mass", if you will, and increase in predictability? Have you ever studied past season to see the correlations of efficiency metrics as the season progresses? In other words, at what point in the season can we say "they are who we THOUGHT they were" based on the metrics?

  29. Anonymous says:

    It was my post/questioning that started the Carolina ranking?
    Advanced stats users like myself know not to put much wait on team w/l records.
    To be fair to Brian, his rankings are meant to be 'predictive' i.e. moving forward.So it remains an open question.
    BUT his methodology is projecting Carolina to be 4th best 62% 0r ~ 10 wins. My research disagrees...time will tell :)

  30. Anonymous says:

    Four games into the season and thread-jack whining because the model doesn't perfectly match someones eyeball test. Small-sample size maybe? Or perhaps your untested brain-model might be incorrect?

  31. Anonymous says:

    Seriously stupid comment Anon..

    Inside Vegas power rankings have Carolina just below average.
    Car is only 3points at HOME over mundane Seattle this implies they are virtually even?? 50 to 1 and 12th best in NFC on vegas futures that I provided....
    on and on..not a brain test cold hard facts...All successful models need to continually look for flaws..I clearly pointed on out...time to get your head out of the sand!

  32. Sampo says:

    To the Anon above:

    What if Vegas would rate Carolina as a huge favorite? People would bet like crazy on Seattle. Bookies don't want this. They want action on both sides to minimize losses. They go with the flow or wisdom of crowds. They can not give out odds that differ too much from the market. That is the reason why Vegas can be wrong.

  33. Anonymous says:

    Sampo..of course Vegas can be wrong by a point or two BUT yr fooling yourself if they are going to be off by 6 or 7 points! on a team as would be the case if Carolina was 62% vs 49% team?!not to mention the futures market..they would be out of business in a flash...
    Of course Vegas adjusts for crowd bias but its a poitn or 1.5 at most..! Seattle is not a crowd fav BTW ..so your rebuttal holds no water..sorry...Carolina ranking is IMO way off..

  34. James says:

    Re: Cam's YPA per quarter and based on game states.

    Using pro-football-reference.com's play-by-play finder, these are the Panthers' stats split by score differential:

    Panthers up 8+ points: 15.6 NYPA on 8 plays. (8.4 NYPA excluding a 66-yard gain).
    +7 to -7 Points: 7.5 NYPA on 48 plays.
    -8 to -14 points: 8.3 NYPA on 33 plays.
    -15 or more points: 7.9 NYPA on 31 plays.

    Let's throw out the 66 yard play: that brings the overall NYPA down to 7.9, which is still second best in the league (and new #1 Cincy has a 73 trick play boosting their stats). Carolina varies from 7.4 to 8.4 in the different states, and is perfectly average when down 3+ scores. Still think garbage time is adjusting their stats?

    Fine then. Let's also throw out all plays where the Panthers were down two or more scores in the 4th quarter:

    -8 or more points, 1st-3rd qtr only: 8.24 NYPA on 41 plays.

    Making their NYPA without garbage time and without a 66 yard play ... 7.9. Even if you take the worst of all three remaining scoring states (scoring margin between -7 and +7, 7.5 NYPA) the Panthers are the 3rd best passing team. The Panthers have been extremely efficient at passing, period.

  35. Anonymous says:

    James Thankyou for a clear example of how inaccurate it is to take a slice of the pie to fit one's argument!

    I never said it was 'only' Carolina's Opass stat that was in question ( that was the position of you and other carolina backers :)
    I questioned their 'ranking' which is based on four main stats opass & dpass & orun success and drun success
    To validate your claim you need to take all efficiency stats in the game states you selected. As well, the reduction from 8.4 to 7.9 is a 25% drop from below the av of 6.3 (.5/.2) so it is significant!

  36. Alan says:

    I love this site and almost always agree with everything posted. I practically tear my hair out listening to all the CW nonsense spouted by commentators and for that matter by coaches every week.

    But I have NFL Rewind, and thanks to the condensed games I watch over 50 percent of all the snaps taken in the NFL each week. And the Panthers are just not the fourth best team in the league. More starkly, this model ranks them as the best team in the NFC, period! Just...no.

    It's true that the ESPN power rankings do underrate them down at 24th. It strikes me that their true strength is somewhere in between, maybe 11th or 12th. They have a soft schedule, so depending on how everything bounces they might even squeak into the playoffs, but I would be pretty surprised to see them even win a playoff game--and you'd certainly expect them to win more than one playoff game if they were really the top NFC team and didn't have to face anyone in the postseason from the stronger AFC unless they got to the SB.

  37. Sampo says:

    At least it will be extremely interesting to follow the Panthers this season!

  38. Boston Chris says:

    Hey Brian,

    I'm wondering if there's something to the thought of how often a team passes? You have said many times that teams don't pass enough. Well on the extreme edges there must be teams that pass more than usual and less than usual. If you are correct that more passing leads to a better team, then teams that pass more than usual will be undervalued by an efficiency model that doesn't account for that? On the flip side a team that passes less than usual could be overvalued? I'd be interested to get your thoughts on that, since it kind of combines a couple of different thoughts on the site.

  39. Boston Chris says:

    Or put a different way, I know you have tested and thrown out a lot of different variables for predictability. Did you ever test run/pass ratio to see if it can be predictive? (I'm guessing its not just b/c it'd probably be only the extreme cases that are affected, and on top of that would be highly dependent on game situation, but still would be interested to know)

  40. Mike M. says:

    Passing more often is not the answer "if" your team goal is to win the Super Bowl.

    Just 1 team has won the SB passing on more than 59% of their plays, and no-one here could quess which team it was.

    Manning, Brady and Warner all failed.

    And to put it another way, SB winners with run/pass balance dominated throughout the postseason where-as SB winners who passed often win in much, much closer games in the playoffs.

    Example, of the 5 Patroit SB teams which team won by the largest margins in the playoffs ?

    The 2004 team which had almost a 50/50 run/pass mix and it outgained it's opp by the largest amount running the ball in the regular season.

    NFL History paints the same story over and over again.

  41. Brian Burke says:

    Mike M. - That's because winning teams tend to run the ball to take time off the clock in games in which they're already ahead.

  42. Alan says:

    Agreed, Sampo--and I think the most recent Panthers game illustrates my point. The defense and special teams played well, but the offense seriously struggled.

  43. Mike M. says:

    Brian, Yes I understand that. That's not what I'm talking about.
    Super Bowl winners that out-rush their opponents in the regular season by larger amounts win by larger amounts in the playoffs.
    Passing more often does not seem to produce bigger wins for SB winners in the playoffs, balance does.

    In the passing era since 2000, just 2 SB winners have been able to win by 17 pts or more throughout the playoffs and those 2 were not exactly powerhouse passing teams, the 2000 Ravens and the 2002 Bucs.

    From 1978 to 2000, it was routine for SB winners to win by 17 pts or more throughout the playoffs and those teams were not passing on a high percentage of plays, they had much better run /pass balance and that led to more dominace in the playoffs by SB winners.

    The powerhouse passing teams led by Brady, Manning and Warner could not duplicate such playoff dominance, and the team which came the closest was the 2004 Pats who had the best run/pass balance and the largest difference in out-rushing their opponents in the regular season.

    Example using the Pats, the 2007 team was a monster passing team in the regular season but did not dominate in the playoffs the way the 2004 Pats did, and the 2004 Pats team was far better at out-rushing their opponent then the 2007 team in the regular season.

    If the 2004 team had passed more often they would not have dominated as much in the playoffs.

Leave a Reply