Team Efficiency Rankings: Week 7

For the first time all year, the top 10 teams from last week remained the same.  In fact, no team moved more than five places in the efficiency rankings, the least movement we've seen so far.  This does not means things won't change, however, especially considering the spate of critical injuries we've seen the past couple of weeks.  Something tells me the Bears won't remain the ninth-most efficient passing offense with Josh McCown at the helm for the next month (#analysis). 

The top was not entirely constant, as only the top-ranked Seahawks, second-ranked Broncos and ninth-ranked Chargers remained in their spots.  The Bengals sitting at fourth looks a bit curious, but the majority of the top-10 seems fairly reasonable.  In fact, apart from the offensively-uberefficient Eagles, these rankings seem to agree with the mainstream perception.

However, there is one particularly curious disagreement, as the team at the top of the league standings is nowhere near the top of these rankings.

Hey, This Looks Familiar
The Kansas City Chiefs may be the NFL's lone undefeated team, but their eighth place ranking is an embodiment of how the fundamental factors that determine who wins football games have changed.  Overall, the Chiefs possess the top-ranked defense and 26th-ranked offense.  The Broncos, angling to become the first 14-2 fifth-seed ever, are the exact opposite, with the top-ranked offense and 26th-ranked defense.  That Denver is second and has a GWP 11 percentage points higher illustrates how a prolific passing game is the ultimate trump card.

Plenty of writers have bandied about the 2011 49ers comparison with the Chiefs.  That really rises from the fact that Alex Smith is/was the quarterback for both defensive-oriented teams, but it's almost spot on:


In terms of EPA per game, we can see that the 2013 Chiefs and 2011 49ers occupy the same quadrant, though Kansas City's defense has actually been better.  That's likely due to their historically good pass defense—the Chiefs' minus-60.4 EPA against the pass is almost double the Seahawks' second-ranked pass defense.  The 2011 Niners were "just" plus-2.5, and no team has put up a full-season mark better since the 2009 Jets.

Meanwhile, there's is/was a perception that Smith simply avoids turnovers while an elite running game carries the offense.  But as we can see, the name value behind Frank Gore and Jamaal Charles didn't necessarily match the actual production:


Both the Chiefs and Niners actually had a below-average running attack.  In fairness, Smith has never been Brady or Manning, but it's not as if he is the glaring weak link in an otherwise flawless team.  The offenses got by with the advantages inherently provided by their defenses—both teams had the best average starting field position in the league, per Football Outsiders.

In today's NFL, no team is better equipped to win without a superstar quarterback than the Chiefs.  The 2011 49ers were essentially a coin flip away from making the Super Bowl, as their loss to the Giants involved tons of crazy late-game swings.  The Chiefs may or may not get the right bounces come this postseason (seven games is enough to know they're not a fluke), but they're certainly in position to take advantage of a weak AFC field.


Hey, This (Also) Looks Familiar

Below these rankings, take a look at the league average in all the stat categories.  Then compare those to the Dallas Cowboys.  This would be perfect if Dallas was ranked 16th, but the perpetually middling Cowboys are once again the barometer for passable mediocrity.

I've already earned the enmity of DC (and many other places as well) for coming out of the closet as a Tony Romo defender, but the numbers still don't lie.  Romo has had another slightly-below elite season, ranking eighth in EPA, 12th in EPA per play, and fourth in success rate. 

The defense and run game haven't been significantly worse this year, but the fall might be coming.  DeMarcus Ware and DeMarco Murray will both miss multiple weeks, which is more crippling than under normal circumstances given the Cowboys' top-heavy roster construction.  Like a college team with scholarship reductions, Dallas' salary cap mess has hampered them from building a viable middle class, meaning that they have less room for error with injuries than just about any other team.

The Cowboys have finished 9-7 or 8-8 five of the past eight years, and will likely end up in that neighborhood again.  They'll likely end up stumbling into a division title, especially if they can take a Week 17 home game against the Eagles to sweep the season series, but that's more an indictment on the other three teams.  Dallas' reward will be a first-round mismatch against either San Francisco or Seattle.


Quick Hits

- These rankings are organized by GWP, which leaves Carolina at 12th.  However, the Panthers have allowed the second-lowest opponent GWP, trailing only the Saints.  Their GWP net difference is sixth-best, behind Seattle, Denver, Cincinnati, Green Bay and New Orleans.  That's what a 26-point average margin of victory will do for you.
- The Cleveland Browns must have someone secretly working on a time machine to bring back Otto Graham.  The Browns' third-ranked defense has again created the league's most competitive five-to-seven win team.  Their pass defense is just a hair behind that of the Seahawks', but with a bumbling Brandon Weeden instead of Russell Wilson, that effort will go in vain.
-  Can there really be three teams worse than the Jaguars?  This graph would seem to defy that logic.  Then again, the three teams below Jacksonville employ Josh Freeman, Kellen Clemens and Mike Glennon at quarterback, so who really knows?

Check out this week's rankings below.

RANKTEAMLAST WKGWPOpp GWPO RANKD RANK
1 SEA10.800.5242
2 DEN20.730.48126
3 SF60.690.5586
4 CIN80.680.4878
5 IND40.670.541210
6 GB100.640.44325
7 PHI30.630.53515
8 KC70.620.48261
9 SD90.610.50232
10 NO50.610.331023
11 DAL130.610.54137
12 CAR120.550.381518
13 HOU160.530.55184
14 NYJ180.530.41169
15 DET110.530.471116
16 NYG150.510.55215
17 TEN200.490.611412
18 CHI140.470.49630
19 PIT220.470.402217
20 CLE170.440.47283
21 ARI190.430.522313
22 NE210.430.492511
23 BUF230.430.482014
24 ATL250.420.34931
25 WAS280.390.461728
26 MIA240.370.422724
27 BAL270.360.511921
28 OAK260.310.473120
29 JAC300.270.572922
30 MIN290.270.423027
31 STL320.260.502429
32 TB310.210.433219

TEAMOPASSORUNSR%OINT%OFUM%DPASSDRUNSR%DINT%PENRATE
ARI5.7384.92.76.3583.10.34
ATL7.2401.21.96.8571.40.45
BAL6.2283.01.56.5541.70.45
BUF5.2422.12.66.1594.40.45
CAR6.3462.91.55.8594.40.38
CHI7.0452.91.98.1574.40.32
CIN7.0422.42.45.7591.80.45
CLE4.9433.01.35.2611.80.49
DAL6.7401.92.36.4613.00.45
DEN8.4441.03.47.3633.10.46
DET6.9381.41.96.9623.70.47
GB7.7451.82.56.6601.30.46
HOU5.9393.91.85.4611.80.56
IND6.2481.31.16.2563.40.26
JAC5.3344.61.87.0551.30.40
KC5.5371.61.54.8584.00.32
MIA5.7373.22.26.4563.00.27
MIN5.4363.62.76.6552.80.36
NE5.2391.82.05.8553.20.35
NO7.4392.10.75.9533.90.30
NYG6.3335.92.56.1621.70.45
NYJ6.2404.92.15.4690.70.51
OAK5.8363.52.26.5561.40.45
PHI7.0482.11.76.7552.60.40
PIT6.4342.32.76.0641.10.39
SD7.9452.01.37.2451.20.38
SF6.8412.71.95.9613.70.49
SEA7.0472.13.85.1614.60.55
STL5.6361.52.37.4582.30.49
TB5.0382.72.06.6632.80.66
TEN5.9372.22.36.2552.60.52
WAS6.6452.52.47.3552.00.48
Avg6.3402.72.16.3582.60.43

  • 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

14 Responses to “Team Efficiency Rankings: Week 7”

  1. Anonymous says:

    http://www.advancednflstats.com/2013/10/team-efficiency-rankings-week-6.html

    I might be missing something, but it looks like there is a discrepancy for what last week's rankings were.

  2. Brian Burke says:

    We had a problem witht the format of the data source last week that required a minor re-code. I think it's working the way it should now. Apologize for any discrepancies. The 'last week' column above could be considered a revised and corrected ranking for last week.

  3. Anonymous says:

    I think that there might be a better way to measure offensive passing for this model than net yards per attempt, because ny/a values all yards, including those past the first down marker, as equal.

    This means that passing offenses that make the occasional big play are overrated by the model, which would explain how the Eagles were at #3 last week. This also might explain how it has the Jets and Colts passing offense as equal, when anybody (or EPA) can tell you Andrew Luck has been playing better than Geno Smith.

    Basically, NY/A has the Eagles, Niners, and Jets as better passing teams than the Colts when either observation or EPA can tell you otherwise.

    Would (SR%)(NY/A) make it more predictive?

  4. Anonymous says:

    How did Carolina's defense go from 8th best to 18th in one week?

  5. Anonymous says:

    The Giants, in the 1980s, played the field position game and wore down their opponents more times than not. But that kind of play (winning by net field position shifts) is something that people, and their stat models that focus on player-yardage/scoring stats, tend to ignore.

    So while I find these individual stats interesting, they don't really capture (IMO) a lot of important information because, by their nature, they can't. And that's where some QBs, despite 'great stats' fail, while others, despite 'poor stats' win.

  6. James says:

    Geno Smith has 8 more interceptions than Luck, which is going to have a big impact on his EPA but isn't necessarily predictive of future performance. Also my understanding is Luck is extremely efficient with his running, particularly on third downs, which will also increase his EPA but not show up in NY/A

  7. Anonymous says:

    I don't think that the 8 picks cause a gap that large. It is still better to consistently get medium gains than to get big gains less frequently. In general, QBs with the similar NY/A have a higher EPA if their success rate is higher.

  8. Anonymous says:

    I’m able to replicate all of the efficiency stats except for FUM%. Please correct me if I’m wrong. Fumbles / ( rush + completions + sacks )
    Using Arizona I have; 12 / ( 154+162+20) ; = 3.57% but Brian’s chart has 2.7%
    Has this computation changed since, http://www.advancednflstats.com/2007/04/fumble-rate.html

  9. Anonymous says:

    Okay, just figured out that he is including all pass attempts. Can someone direct me to the study were this might be explained. I’m not sure how fumbles are even possible on a incomplete pass.

  10. Anonymous says:

    Ok, let me try this again.
    Carolina's Def efficiency metrics compare favorably with Dallas and Cincinnati. Yet while Dal and Cin are 7th and 8th in D Rank, Carolina is 18th. What factors influence this ranking?

  11. Anonymous says:

    I'm having trouble reproducing the probabilities given at the NY Times (http://www.nytimes.com/2013/10/24/sports/football/week-8-nfl-game-probabilities.html). I'm attempting to reproduce them using the week 7 team efficiency scores from http://www.advancednflstats.com/2013/10/team-efficiency-rankings-week-7.html. So I'm using week 7 efficiency ratings to make predictions about week 8. Is that the correct way to do it?

    For example, for Seattle at St. Louis, I compute logits as:
    SEA: 0.46*7.0 + 0.25*0.47 + -19.4*0.021 + -19.4*0.038 + -0.62*5.1 + -0.25*0.61 + -1.53*0.55 = -1.9631
    STL: 0.46*5.6 + 0.25*0.36 + -19.4*0.015 + -19.4*0.023 + -0.62*7.4 + -0.25*0.58 + -1.53*0.49 = -3.5539

    St. Louis's Opp GWP is 0.50, equivalent to a logit of zero. Seattle's is Opp GWP is 0.52, good for a logit of 0.08. This gives St. Louis an opponent-adjusted logit of 3.5539 and Seattle one of -1.88.

    The difference is -1.67. When I add 0.36 for home field advantage and compute the odds ratio, I get 0.27, which gives the Rams a 21% chance of winning. But the NY Times has the Rams at only 11%. Am I using old model coefficients? Or doing something else wrong?

    Thanks.

  12. Anonymous says:

    BTW, the coefficients and methodology I'm using come from http://www.advancednflstats.com/2009/01/how-model-works-detailed-example.html

  13. Anonymous says:

    @Anon: Yes, coefficients are old ones. They are coefficients for YPC, not for SR%.

  14. Anonymous says:

    @korsakoff: thanks. After a little more digging I found http://www.advancednflstats.com/2011/09/weekly-game-probabilities.html. Are those the latest probabilities?

    I notice that they begin with:
    $C['const'] = -0.4619;
    $C['ahome'] =.748577;

    Unlike the 2009 coefficients, these don't have the property that the constant is equal to -0.5 times the value of being at home. Isn't this property necessary (as highlighted on the 2009 page)?

    I now get logits of
    SEA: 0.4467*7.0 + 0.03565*0.47 + -13.026*0.021 + -7.209*0.038 + -0.576*5.1 + 0.02749*0.61 + 14.24*0.046 + -1.587*0.55 = -0.5424736
    STL: 0.4467*5.6 + 0.03565*0.36 + -13.026*0.015 + -7.209*0.023 + -0.576*7.4 + 0.02749*0.58 + 14.24*0.023 + -1.587*0.49 = -2.5434088

    Once I adjust for opponents and take the difference I get -2.08097790767

    If I ignore STL's home field advantage completely I match the 11% probability shown in the NY Times. Trying other things based on the -0.4619 and .748577 gives me values that are too high. Am I still not using the right coefficients? If not, where can I find them? Or is the issue something else?

Leave a Reply

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