≡ Menu
Rivers was outstanding in 2013, despite this throwing motion

Rivers was outstanding in 2013, despite this throwing motion.

The Denver Broncos set numerous offensive records last year. The Chip Kelly Eagles had a fascinating offense that was lethal for stretches. The Saints offense was its usual efficient self, and the Chicago Bears under Marc Trestman had one of the best offensive years in franchise history.

Yet all of those teams had at least 61 drives last year that ended in a punt. San Diego , meanwhile, punted just 56 times. The Chargers only had 21 turnovers, which means only 77 San Diego drives could be clearly labeled as failures, or “bad drives.”1

That’s pretty impressive; the 2013 Chargers were just the 36th team during the 16-game era to have fewer than 80 “bad drives” in a season. On the other hand, the Chargers were one of just five of those teams to score fewer than 400 points. San Diego’s offense was very efficient last year, but the 77 “bad drives” statistic is a bit misleading. That’s because the team had just 158 total drives last year according to Football Outsiders, while the average team had 186 drives.

Why did the Chargers have the fewest drives in the NFL? A bad defense certainly helped limit the team’s number of offensive drives: San Diego forced only 82 “bad drives” all year, too. But the main reason was that the offense was not just efficient, but uniquely efficient. According to Football Outsiders, San Diego averaged 3:22 per drive, a full 15 seconds more than the #2 team in that metric, Carolina. And the Panthers were the only other team to average at least three minutes per drive. One reason for the long time of possession is that the Chargers moved at a glacial pace between plays, rating as the 2nd slowest team according to Football Outsiders. The other teams in the bottom four in pace were all run-heavy — Carolina, Seattle, and San Francisco — which marks yet another way in which the Chargers were outliers. In several metrics — first downs per drive, yards per drive, and points per drive — San Diego and Denver were the top two teams in the NFL.  But in pace, Denver ranked 4th, making the Broncos offense look and feel much different than San Diego’s attack.

Another reason the team’s average drive took so long to complete: San Diego averaged 6.85 plays per drive, with New Orleans second in that statistic with 6.35 plays. That’s because the Chargers had a very horizontal passing attack. According to NFLGSIS, Philip Rivers ranked 6th from the bottom in average length of pass at 7.75; only Jason Campbell, Sam Bradford, Matt Ryan, Alex Smith, and Chad Henne threw shorter passes. With the exception of Ryan, none of those quarterbacks came close, however, to matching Rivers’ league-leading completion percentage. What we have here is your classic hyper-efficient, short-area passing game, and the Chargers executed it beautifully.

In fact, here’s another unique part of the San Diego offense: it rarely targeted wide receivers. San Diego was one of just three teams to throw more passes to non-wide receivers than to wide receivers. Here’s how to read the table below: the Chargers threw 25% of all pass attempts to running back, 47.1% to wide receivers, and 27.7% to tight ends. Based on those percentages, San Diego ranked 4th in percentage of pass attempts to running backs, 30th in percentage of pass attempts to wide receivers, and 2nd in percentage to tight ends.

TmRBWRTERB % RkWR % RkTE % Rk
NOR32%40.8%27.6%2323
KAN44%41.1%14.6%13131
SDG25%47.1%27.7%4302
ATL25%54%21.2%62915
STL16%54.5%29.9%25281
HOU19%55.6%25.3%14274
CLE22%56.7%21.4%82612
OAK28%56.7%15.5%32530
DET25%57.2%17.7%52425
BUF24%57.6%18.9%72322
SFO16%58.7%24.9%22225
BAL20%59%21.2%112114
CAR16%59.8%24%23207
PHI17%60.3%23.1%20198
DAL15%60.4%24.4%27186
TAM21%61.2%17.5%91726
NYJ18%61.6%20.6%181617
IND18%62.3%19.5%16.51519
DEN17%62.4%21%211416
MIN15%62.7%22.7%281310
CIN14%62.7%23.1%29129
JAX20%63.2%16.5%101129
SEA16%63.6%20.5%241018
CHI19%63.7%17.2%13927
ARI18%63.8%18%16.5823
PIT19%64.2%17%15728
NYG18%64.6%17.8%19624
GNB15%65.3%19.3%26520
NWE20%66%14.3%12432
WAS11%66.7%22.6%32311
TEN14%66.7%19.3%30221
MIA11%67.9%21.2%31113

Throwing shorter passes to the tight ends and running backs is one way to keep the chains moving, but you have to be extremely accurate if your offense consists of short passes. As it turns out, San Diego had the highest completion percentage in the NFL when throwing to running backs, the highest when throwing to wide receivers, and the eighth highest when throwing to tight ends:

TmRBWRTERBWRTE
SDG85%65.3%66.7%118
DAL82%59.7%66.7%2108
NOR82%64.3%60.4%3424
TEN81%57.3%66.3%41711
DEN79%64.7%75.2%531
SEA79%61.8%65.2%6716
ATL79%63.9%67.9%755
PHI78%59.2%62.8%81121
MIA78%57.3%66.7%9188
CAR78%57.1%66.1%101912
CHI76%61.7%67.3%1186
CIN76%59.1%66.4%121210
GNB76%65.1%64%12220
BAL75%53.5%60%142825
ARI75%58.7%73.7%15132
BUF74%48.3%64.9%163217
TAM74%50.2%70.9%17304
IND74%56.7%58.9%182227
STL73%53.8%64.7%192719
DET71%54.1%61.8%202623
WAS70%56.6%62.2%212322
JAX69%56.8%58.3%222130
PIT69%62.7%71.1%2363
MIN69%57.3%65.3%21615
KAN69%55.9%65.5%252414
NWE68%60%58.8%26928
OAK67%54.9%57.7%272531
CLE66%49.7%65.7%283113
HOU65%58.6%58.6%291429
SFO65%57.8%56.6%301532
NYJ62%51.4%64.9%312917
NYG59%57.1%59.6%322026

Mike McCoy and offensive coordinator Ken Whisenhunt brought a new offense to San Diego, but the Chargers new offense wasn’t just about the coaches. Rivers was pretty clearly helped by a couple of newcomers on the field, too. Danny Woodhead caught 87% of his targets last year, and ranked 1st in both DVOA and DYAR. The other was Keenan Allen, who had an incredible 68% catch rate (of the 12 receivers with at least 60 catches outside of the slot (according to Pro Football Focus), Allen posted the best catch rate. Are those marks sustainable? Is anything San Diego did sustainable last year?

My guess is probably not. San Diego was incredibly efficient in 2013, even if it was efficient in a way that did not capture the public’s eye. The Chargers had the fewest three-and-outs in the NFL, even on a per-drive basis. The team ranked third in first downs gained despite ranking last in possessions. The Chargers were masters of the short-passing game last year, and were also very effective in the running game (ranking 1st in percentage of stuffed runs and 5th in short-yardage running). That seemed even more true down the stretch. Ryan Mathews easily led the NFL in carries in December, and in those five games, Mike Scifres had just fourteen punts.

But something about San Diego’s success seems unsustainable to me, and I know that’s not a very convincing argument. On the other hand, I’m not necessarily predicting that the offense will struggle this year, just that it will look different. The return of Malcom Floyd, a deep threat who missed most of last year with an injury, should help open up the offense. The same goes for Ladarius Green, the number two tight end who looked like an explosive playmaker in just his second season. But the idea that San Diego will again lead the league in length of drive, plays per drive, operate at a glacial pace, and have the fewest bad drives in the NFL, all seems pretty unlikely.

  1. The Chargers were 5/6 on fourth down attempts, so it’s not as though these numbers are skewed by failed fourth down attempts. []
{ 22 comments }
  • Edward Moretti July 28, 2014, 12:28 am

    And here I thought you were going to praise the Bolts. “Unsustainable”? “Unlikely”? Well, I disagree. Telesco brought in yet another RB who will be an upgrade to Ronnie Brown, Donald Brown from IND. According to PFF, “Brown averaged a whopping 2.4 yards more per carry than Trent Richardson (5.3 vs. 2.9) and had an elusive rating 25.4 points higher (73.8 vs. 48.4). By our metric, Brown calculated out as the most elusive back last year with at least 100 touches. He also was one of the biggest breakaway threats, achieving eight runs of 15 yards or more and 40% of his total yards on those runs.also according to PFF, Ladarius Greene is actually a pretty darn good run blocking TE and he will be seeing more of the field due to more 2 TE sets. DJ Flucker will have gotten rid of the rookie jitters and should be better than last year. Jeromey Clary, the worst player on our O-line, is on the PUP list. The return of Malcolm Floyd will only make our offense better. But the biggest hope for our offense to be even better than last year is the tiny speed demon from Baylor, the 7th rounder Tevin Reese. If he can learn to run routes well enough, he will be the kind of player that can truly stretch the field making it easier to dump off those short passes for more yards. So far he has looked great in practice. In practice. Can’t wait what someone as tiny as him will look like on an NFL field.

    Reply
    • Arthuro July 28, 2014, 2:25 am

      It feels to me that you read the final paragraph a little too quickly.

      Reply
  • Shattenjager July 28, 2014, 1:15 am

    With his injury history, the idea that Malcom Floyd is going to return and stay returned also has to be considered pretty unlikely at this point.

    Reply
    • Chase Stuart July 28, 2014, 11:05 am

      That’s a good point.

      Reply
  • Bryan Frye July 28, 2014, 7:44 am

    The Patriots minus Gronkowski – 28th in TE completion rate. Yowza. Do people still say yowza?

    Reply
  • LaVoNtE July 28, 2014, 11:24 am

    I wonder if the slow pace was on Rivers or just the game plan. I watched a lot of Chargers last year and it seemed every play the playclock would run down to 1 and Rivers would be clapping his hands over and over to get the ball. He must have burned a dozen timeouts last year doing this.

    Reply
  • Nick Bradley July 28, 2014, 12:10 pm

    Chase Stuart,

    I took a look at target distribution vs comp% and found a high correlation — R^2 of 0.927. So if you control for target distribution, you can factor out any QBs that boosted their comp% with dump-off passes to running backs. At the top of the list are Brees and Alex Smith. you get some really interesting results — BUF actually had a good Cmp% for the types of receivers they targeted, while GNB and PHI were on the opposite end of the spectrum.

    And San Diego? Well, they performed very well for the types of receivers targeted.

    formula below, here’s list:

    Tm Cmp% Exp Cmp% Cmp%+
    NOR 68.5% 66.14% 2.36%
    KAN 61.0% 59.27% 1.73%
    BUF 57.3% 55.76% 1.54%
    SDG 69.5% 68.12% 1.38%
    ATL 67.5% 66.26% 1.24%
    STL 59.5% 58.75% 0.75%
    DEN 68.3% 67.65% 0.65%
    CLE 55.7% 55.07% 0.63%
    DET 58.5% 58.15% 0.35%
    PIT 64.3% 63.96% 0.34%
    NYJ 55.4% 55.16% 0.24%
    BAL 58.6% 58.37% 0.23%
    IND 60.1% 59.90% 0.20%
    JAX 59.0% 58.88% 0.12%
    OAK 57.4% 57.29% 0.11%
    CHI 64.4% 64.29% 0.11%
    NYG 57.3% 57.21% 0.09%
    SFO 58.5% 58.44% 0.06%
    ARI 63.2% 63.16% 0.04%
    NWE 60.5% 60.61% -0.11%
    CAR 61.7% 61.86% -0.16%
    DAL 64.0% 64.25% -0.25%
    MIN 59.5% 60.06% -0.56%
    HOU 58.6% 59.17% -0.57%
    CIN 62.0% 62.66% -0.66%
    TAM 56.6% 57.63% -1.03%
    SEA 63.6% 64.69% -1.09%
    TEN 61.5% 62.60% -1.10%
    WAS 58.1% 59.46% -1.36%
    GNB 64.2% 65.86% -1.66%
    PHI 61.0% 62.66% -1.66%
    MIA 60.1% 62.05% -1.95%

    =0.04436405 + 0.200705685*(RBTGT%) + 0.583934509*(WRTGT%) + 0.127396973*(TETGT%)

    Reply
    • Nick Bradley July 28, 2014, 12:49 pm

      sorry that’s comp% to position vs. total comp%

      Reply
      • Nick Bradley July 28, 2014, 12:55 pm

        Here it is for pass distribution, R^2 is only .11 — lots of variance

        Tm Cmp% Exp Cmp% Cmp%+
        DEN 68.3% 60.49% 7.81%
        ATL 67.5% 61.70% 5.80%
        SDG 69.5% 64.19% 5.31%
        PIT 64.3% 59.67% 4.63%
        CHI 64.4% 60.05% 4.35%
        NOR 68.5% 64.33% 4.17%
        GNB 64.2% 60.57% 3.63%
        SEA 63.6% 60.54% 3.06%
        ARI 63.2% 60.30% 2.90%
        DAL 64.0% 61.98% 2.02%
        NWE 60.5% 58.82% 1.68%
        TEN 61.5% 60.03% 1.47%
        CIN 62.0% 61.45% 0.55%
        MIA 60.1% 60.18% -0.08%
        PHI 61.0% 61.16% -0.16%
        CAR 61.7% 61.97% -0.27%
        IND 60.1% 60.77% -0.67%
        JAX 59.0% 60.16% -1.16%
        MIN 59.5% 60.80% -1.30%
        KAN 61.0% 62.34% -1.34%
        NYG 57.3% 59.59% -2.29%
        WAS 58.1% 60.41% -2.31%
        DET 58.5% 60.91% -2.41%
        BAL 58.6% 61.12% -2.52%
        OAK 57.4% 60.24% -2.84%
        BUF 57.3% 60.53% -3.23%
        STL 59.5% 63.20% -3.70%
        SFO 58.5% 62.47% -3.97%
        TAM 56.6% 60.59% -3.99%
        HOU 58.6% 62.62% -4.02%
        NYJ 55.4% 60.69% -5.29%
        CLE 55.7% 61.52% -5.82%

        Reply
      • Chase Stuart July 28, 2014, 1:10 pm

        Wouldn’t it make more sense to compare target percentage at each position to total comp%? I ran a regression using target percentage for each position as my input, and total completion percentage as my output. The best fit formula was 1.58882895 -0.949888113 * RBTAR% -1.066092045 * WRTAR% -0.74847885 * TETAR%

        From there, you can calculate expected cmp%, and cmp% over expectation:

        Rk___Tm___Act___Exp___Diff
        1____DEN___0.683___0.605___0.078
        2____ATL___0.675___0.617___0.058
        3____SDG___0.695___0.642___0.053
        4____PIT___0.643___0.597___0.046
        5____CHI___0.644___0.601___0.043
        6____NOR___0.685___0.643___0.042
        7____GNB___0.642___0.606___0.036
        8____SEA___0.636___0.605___0.031
        9____ARI___0.632___0.603___0.029
        10___DAL___0.640___0.620___0.020
        11___NWE___0.605___0.588___0.017
        12___TEN___0.615___0.600___0.015
        13___CIN___0.620___0.615___0.005
        14___MIA___0.601___0.602___-0.001
        15___PHI___0.610___0.612___-0.002
        16___CAR___0.617___0.620___-0.003
        17___IND___0.601___0.608___-0.007
        18___JAX___0.590___0.602___-0.012
        19___MIN___0.595___0.608___-0.013
        20___KAN___0.610___0.623___-0.013
        21___NYG___0.573___0.596___-0.023
        22___WAS___0.581___0.604___-0.023
        23___DET___0.585___0.609___-0.024
        24___BAL___0.586___0.611___-0.025
        25___OAK___0.574___0.602___-0.028
        26___BUF___0.573___0.605___-0.032
        27___STL___0.595___0.632___-0.037
        28___SFO___0.585___0.625___-0.040
        29___TAM___0.566___0.606___-0.040
        30___HOU___0.586___0.626___-0.040
        31___NYJ___0.554___0.607___-0.053
        32___CLE___0.557___0.615___-0.058
        
        Reply
        • Chase Stuart July 28, 2014, 1:12 pm

          I guess great minds think alike!

          Reply
        • Nick Bradley July 28, 2014, 1:25 pm

          Yes I got the exact same results. I mis-labeled my spreadsheets.

          This is actually a decent measure of QB accuracy. It reminds me a bit of FIP in baseball (or xFIP).

          The only good quarterbacks near the bottom are Kaepernick and Stafford — no bad quarterbacks near the top. Alex Smith is still captain check-down. What’s kind of interesting about Kaepernick is that expected Cmp% is what his actual cmp% was in 2012. And if you apply the expected rate to 2013, you end up with the same QB rating as 2012 as well, 98.5 (98.3 in 2012).

          You can also look at RB/WR/TE component splits

          Reply
  • Chase Stuart July 28, 2014, 1:22 pm

    I just noticed something pretty interesting. The coefficients on all the targ% numbers are negative, but you would expect that RBs to have the smallest coefficient and WRs to have the largest. Instead, TEs have the smallest coefficient. That doesn’t make sense, since the comp% to RBs is 73%, while the cmp% to TEs is 64%.

    To put it in more practical terms, one would expect Miami to have the lowest expected cmp%; after all, the Dolphins threw the most passes to WRs on a percentage basis, and the 2nd fewest to RBs. But that’s not the case; the Patriots do.

    Miami threw 68% to WRs, 21% to TEs, and 11% to RBs. NE threw 66% to WRs, 14% to TEs, and 20% to RBs. Frankly, NE should have a higher expected completion percentage, and I don’t even think that’s debatable. So what’s going on here? It may be a one-year bit of randomness. But I think that’s something worth looking into — thoughts?

    The interesting result would be if throwing passes to the TE was somehow correlated with QB ability, or throwing passes to the RB was correlated with the lack thereof.

    Reply
    • Nick Bradley July 28, 2014, 1:34 pm

      I think your last sentence really hit the nail on the head. You often see backup quarterbacks just dumping the ball off in the flats to RBs, while hitting TEs in the seam takes a lot more skill.

      Strong-armed quarterbacks have trouble with screen plays, to both TEs and WRs.

      Reply
    • James July 29, 2014, 9:45 am

      I seem to remember Brian Burke showing that contrary to convention wisdom rookie QBs don’t throw the ball to the TE very often.

      Reply
    • chris July 29, 2014, 9:56 am

      You are proxying position for the main factor in completion %, pass length by assuming RBs catch shorter passes than wide receivers. I think that’s generally reasonable, but it doesn’t work for MIA and NE’s TE targets.

      MIA: 50 completions last year to TEs, avg pass length 5.38 yards
      NE: 53 completions to TEs, avg pass length 8.47 yards

      MIA: 28 incomplete passes targeted to TEs, avg pass length 7.86 yards
      NE: 39 incomplete passes targeted to TEs, avg pass length 11.67 yards

      NE should have a lower completion % than MIA. Brady’s passes to TEs were longer on average than his passes to other receivers, while Tannehill’s passes to TEs were shorter on average than his passes to other receivers.

      Reply
  • Joel July 28, 2014, 5:07 pm

    Your comment about inflating completion percentage by throwing to the running backs got me thinking. I took each quarterback’s completion percentage by position and multiplied it by the league average target rate by position to get an adjusted completion percentage. Then I subtracted the real completion percentage from the adjusted completion percentage to get a measure of inflation.

    It turns out that Alex Smith and Drew Brees have inflated completion percentages at just over 2% by this measure and that Miami represents the other end of the spectrum at about 1.7% under do to the high proportion of passes to wide receivers.

    I would post the table but it looked pretty bad with copy and paste (from Excel)… Let me know if there is a better way.

    Reply
  • Dave July 28, 2014, 6:09 pm

    Expected comp% I would think works pretty well for RB’s and TE’s but probably not as well for WR’s depending on if your passing game is more horizontal or vertical. It probably works Ok on average but certain teams will get misrepresented if they throw a lot of deep balls or a lot of screens.

    I think a better method is expected comp% over average by pass distance.

    Reply
  • Dave July 28, 2014, 6:12 pm

    Well I guess pro-football-focus did what I suggested earlier this year…..

    https://www.profootballfocus.com/blog/2014/01/27/2013-adot-adjusted-completion-percentage/

    Reply

Leave a Comment