≡ Menu

Luck leads the league with three fourth-quarter comeback wins

Luck leads the league with three fourth-quarter comeback wins.

Week eight brough us the two biggest blowouts of the season; in week nine, we saw the third most dominant win of the year and the biggest comeback of the season.

The first game involved Chip Kelly’s blitzkrieg offense. Nick Foles threw for seven touchdowns against the Raiders in one of the most lopsided (and surprising) games of the season. The Eagles held a 28-13 lead at halftime and 49-13 at the end of the third quarter; over the course of the game, Philadelphia held an average lead of 21.3 points.

At the other end of the spectrum, we have yet another Andrew Luck comeback victory. The Texans led 14-0 after the first quarter and 21-3 at halftime; on average, Houston held an 11-point lead throughout the game, but a 15-0 edge in the fourth quarter gave Indianapolis the win. That’s the highest Game Script of any team to lose a game in 2012, replacing…. Houston’s victory over the Chargers on opening week, when the Texans had a Game Script of -7.7 points.

In addition to the Colts-Texans game, the crazy comeback in Seattle now gives each of the Seahawks and the Bucs two of the five biggest comebacks/giveaways of the year. In week four, Seattle won in overtime against Houston despite trailing by, on average, 7.7 points in regulation. That was probably an even more crazy game than the win against Tampa Bay, where Seattle came back from a 21-0 deficit but only outscored the Bucs by 10 points in the fourth quarter. As for Tampa Bay, this was the fourth game of the season where the team lost despite having a 95% win probability at some point in the game. This was also the second time the Bucs lost a game with a Game Script of over 6.0 points, joining the come-from-ahead loss to Arizona.

Without further ado, the table below shows the week 9 Game Scripts data:

WinnerH/RLoserBoxscorePFPAMarginGame ScriptPassRunP/R RatioOp_POp_ROpp_P/R Ratio

In general, the usual trend held as teams with negative Game Scripts passed more frequently than their opponents, with a couple of notable exceptions. After Aaron Rodgers went down, the Packers executed a run-heavy game plan centered around Eddie Lacy, and therefore passed less often than Chicago despite trailing by, on average, 1.6 points.

The Dallas/Minnesota game was even more extreme. In some ways, it’s not surprising seeing Tony Romo and the Cowboys pass more often than a Vikings team with Adrian Peterson, but it’s the magnitude of the disparity that is jarring. Through five weeks, I noted that the Cowboys had the second strongest pass identity in the league. Things haven’t changed too much, other than an uncharacteristically ground-heavy performance (30 passes, 26 runs) in a 1-point loss to Detroit (Game Script of +1.2) in week eight. Against the Washington and Philadelphia in weeks six and seven, Dallas passed often despite holding strong leads, but against Minnesota, the Cowboys’ pass identity was off the charts. Dallas passed on 85.7% of all plays — tied for the highest mark of the season — despite holding an average lead of 1.4 points.

The other games all went according to (game) script, with the trailing team passing more often than the leading team, although the San Diego/Washington game was pretty extreme, too. This was an even game — Washington had a Game Script of +0.1 — so the large disparity in the teams’ pass/run ratios provides a window into each team’s true identity. For Washington, that involves a heavy dose of the run: Alfred Morris has 25 carries, Robert Griffin III had six runs, and Darrel Young had three touchdowns on five carries. Washington rushed 40 times for 209 yards and 4 touchdowns, while passing just 32 times. That’s probably the ideal split for the Shanaclan, as Griffin was very efficient: he picked up 291 passing yards (and was not sacked) on 32 dropbacks, and Pierre Garcon had 7 catches for 172 yards.

San Diego represents the other extreme. Danny Woodhead and Ryan Mathews each had seven carries, and that was it for the running game (although Philip Rivers technically added 14 yards on two scrambles). The Chargers star quarterback is having a magnificent season as part of the team’s new horizontal offense. But this was far from his best day: Rivers threw for 50 more yards than Griffin, but it took him fifteen more dropbacks.

As a reminder, you can view the game scripts from every game this year at this page. Finally, let’s close with some average field position data from week nine.

TeamBoxscore# playsAvg Yardline
New England PatriotsBoxscore7155.5
New York JetsBoxscore5855.1
Seattle SeahawksBoxscore6154.8
Carolina PanthersBoxscore7152.7
Washington RedskinsBoxscore7252.3
Minnesota VikingsBoxscore6750.6
Indianapolis ColtsBoxscore5849.4
Green Bay PackersBoxscore5548.6
Pittsburgh SteelersBoxscore7348.3
Cleveland BrownsBoxscore6848
Kansas City ChiefsBoxscore5447.1
St. Louis RamsBoxscore6947.1
Chicago BearsBoxscore7546.4
Cincinnati BengalsBoxscore9246
Buffalo BillsBoxscore7645.8
Houston TexansBoxscore6845.5
Dallas CowboysBoxscore6345.5
San Diego ChargersBoxscore6345.2
Philadelphia EaglesBoxscore5744.7
Oakland RaidersBoxscore9144.5
Tennessee TitansBoxscore6144.3
New Orleans SaintsBoxscore6643.3
Baltimore RavensBoxscore6743
Miami DolphinsBoxscore6140.6
Tampa Bay BuccaneersBoxscore6538.7
Atlanta FalconsBoxscore4838.1

Against the Saints, the Jets scored 26 points despite throwing for only 115 yards and picking up just 14 first downs. Field position played a key role in that production. The Jets last touchdown was set up by an Antonio Cromartie interception that gave New York the ball at the Saints 39-yard line. The Jets first score of the game was a field goal set up by another Drew Brees interception, which put the Jets on the New Orleans 48. The Jets did start one drive at their own 4-yard line, but on the second play of that drive, Chris Ivory ran for 52 yards, putting the Jets in Saints territory.

  • Arif Hasan

    Here you go, SOS adjusted:

    Rank Tm SRS
    1 CAR 6.90
    2 GNB 5.89
    3 SFO 5.37
    4 NOR 3.87
    5 DEN 3.75
    6 CIN 3.54
    7 DAL 3.39
    8 NWE 2.61
    9 KAN 2.35
    10 SEA 2.24
    11 DET 1.17
    12 ATL 1.05
    13 IND 0.33
    14 ARI 0.33
    15 BUF 0.14
    16 MIA -0.13
    17 SDG -0.22
    18 CHI -0.31
    19 CLE -0.48
    20 PHI -0.59
    21 TAM -0.83
    22 TEN -1.14
    23 MIN -1.80
    24 NYG -1.97
    25 OAK -2.05
    26 BAL -2.71
    27 HOU -3.32
    28 STL -3.36
    29 NYJ -3.37
    30 WAS -3.92
    31 PIT -5.34
    32 JAX -12.19

    Oakland’s change is enormous, while the Jaguars managed to get historically worse.

    As far as I can tell, the 2008 Lions went -7.6 while the 2011 Colts were at about -7.1. The Kansas City Chiefs last year were notable in having a record streak of games where they did not play a single second with the lead (until about Week 9 or so I believe, their win against New Orleans not counting as it came in overtime—where therefore there were no seconds played with a lead because the game ended as soon as there was a score) yet finished with a Game Script of -7.3 Jacksonville’s unadjusted Game Script is even worse than -12.2 at -13.2.

    A team of this ineptitude and failure comes along once in a blue moon. If this data is to be useful, then it will allow us to explain that the Jaguars (at least so far) were almost twice as bad as one of the worst teams in history.

    If only play-by-play data were available in 1976, so we could compare this year’s Jaguars to the Buccaneers (although era-adjustments would need to be made).

    • Arif Hasan

      Screwed that up:

      Rank Tm SRS
      1 CAR 6.92
      2 GNB 5.73
      3 SFO 5.36
      4 NOR 3.88
      5 DEN 3.74
      6 CIN 3.55
      7 DAL 3.45
      8 NWE 2.60
      9 KAN 2.39
      10 SEA 2.31
      11 DET 1.15
      12 ATL 1.04
      13 ARI 0.33
      14 IND 0.25
      15 BUF 0.11
      16 SDG -0.12
      17 PHI -0.18
      18 CHI -0.18
      19 MIA -0.20
      20 CLE -0.55
      21 TAM -0.85
      22 TEN -1.11
      23 MIN -1.79
      24 NYG -1.84
      25 OAK -2.50
      26 BAL -2.71
      27 HOU -3.26
      28 STL -3.35
      29 NYJ -3.38
      30 WAS -4.03
      31 PIT -5.39
      32 JAX -12.23

      Minor changes.

    • Chase Stuart

      Some comments.

      1) Thanks for putting these together every week — I enjoy reading them!

      2) Yes, the 2012 Chiefs were at -7.3.

      3) I have the 2011 Colts at -7.2…. which was not the worst score of the year. That belongs to the pre-Schiano Bucs.

      4) Below are the worst Game Scripts ever. Note that for pre-2000 teams, I used the estimation approach detailed in the original post.

      DTX---1952---dtx---NFL---1---11---0---0.083--- -12.4
      CRD---1944---crd---NFL---0---10---0---0.000--- -11.3
      BKN---1943---bkn---NFL---2----8---0---0.200--- -10.9
      NYY---1949---nyy---NFL---1---10---1---0.125--- -10.2
      NWE---1972---nwe---NFL---3---11---0---0.214--- -10.1
      CRD---1954---crd---NFL---2---10---0---0.167---  -9.8
      TAM---1976---tam---NFL---0---14---0---0.000---  -9.5
      ATL---1967---atl---NFL---1---12---1---0.107---  -9.5
      CLE---1990---cle---NFL---3---13---0---0.188---  -9.4
      ARI---2000---crd---NFL---3---13---0---0.188---  -9.3
      BOS---1970---nwe---NFL---2---12---0---0.143---  -9.3
      DAL---1960---dal---NFL---0---11---1---0.042---  -9.3
      BAL---1981---clt---NFL---2---14---0---0.125---  -9.3
      GNB---1949---gnb---NFL---2---10---0---0.167---  -9.2
      CHR---1948---cra---AAFC--1---13---0---0.071---  -9.0
      HOU---1973---oti---NFL---1---13---0---0.071---  -9.0
      NWE---1990---nwe---NFL---1---15---0---0.063---  -8.8
      CRD---1945---crd---NFL---1----9---0---0.100---  -8.7
      TAM---2011---tam---NFL---4---12---0---0.250---  -8.7
      BOS---1944---byk---NFL---2----8---0---0.200---  -8.7
      DET---2008---det---NFL---0---16---0---0.000---  -8.6
      BAL---1950---bcl---NFL---1---11---0---0.083---  -8.5
      DET---1946---det---NFL---1---10---0---0.091---  -8.5
      DET---1942---det---NFL---0---11---0---0.000---  -8.4
      NOR---1975---nor---NFL---2---12---0---0.143---  -8.4

      I suppose if someone really wanted to compare apples to apples, they would do the estimation approach for the 2013 Jaguars. But either way, it’s been a really ugly season in Jacksonville.

      • Arif Hasan

        Ah. I was using a rough, per-play method because it was easier on my PC (I need to switch to a SQL-based database or something because Excel sucks at play-by-play/large amounts of data) for the Colts and Lions. When you initially came out with “game scripts,” this was actually the first criticism I thought of—using plays as averages instead of time—but I shelved it because that is not how coaches think. Instead of “I have 15 plays left, better run it” they think of “4 minute offenses” and “2 minute offenses”.

        I forgot about the quarter/estimate approximation method you used when you introduced the concept! Very useful.

        If we were to limit the teams to a popular conception of the contemporary NFL, I would suggest 1960 for a few reasons:

        1) It feels right. Not many people talk or know about Slingin’ Sammy Baugh or Automatic Otto. But everyone knows Bart Starr and Johnny Unitas (a fair bit know about Tittle and Tarkenton, too).

        2) It’s a round number

        3) In 1960, Pete Rozelle became the commissioner

        4) The AFL was founded

        5) The NFL added the Cowboys, and a year later added the Vikings.

        6) A year later, the schedule switched from 12 to 14 games.

        But that is a discussion for another time and another post I imagine. I just felt that those five teams in the bottom six deserved an asterisk when defining “the worst game scripts ever”. None of those teams even exist in their incarnations (the Dallas Texans having folded before being refounded in the AFL, then turned into the Chiefs, while the Chicago Cardinals moved… a few times). It is curious that the Jaguars are finding ways/measures to be worse than the massively imbalanced ’40s teams.

        I knew the ’70s Patriots were bad, but I did not realize they were that bad! And when scoring was so low, too. It is interesting to see them with the “lowest” game script despite finishing with the sixth-lowest point differential of all time and winning 20% of their games (a combined feat that is mind-boggling by itself). I checked their wins, and it’s about what you’d expect—a come-from behind win against the Falcons to take the lead by one point after going down 13 into the fourth quarter; a sort-of come-from-behind win against the Redskins where they traded leads until some time in the fourth. The final score was not theirs, however, as a blocked punt led to a safety so that the Redskins could pull within one; a game where they were always ahead of New Orleans and won by 7.

        Thanks for the info! Of the teams that have won fewer than ten percent of their games, which ones have had the highest game scripts? Also, an offseason post on the worst game script seasons for every franchise would be interesting (so would “best” I suppose).

        And then the last thing I’m curious about is whether or not year-to-year game scripts would even be eligible for era adjustment, given that it is a differential and not a raw total. I imagine it would be a multiplier so that the distributions look similar as the average will always be 0 (so that the average highest and lowest teams would come out to similar values from year to year).

        Sorry for all the prattling; game scripts may be my favorite thing to come from this site of the dozens of things I have loved coming from the work overall (both here and PFR blog).

  • Arif Hasan

    In light of your Dallas comments, here are the Passer ID ranks up to Wk 9:

    Pass ID Rank Team PR ID
    1 DAL 135.8501941
    2 NOR 129.1289386
    3 DEN 128.7279674
    4 ATL 128.4679406
    5 CLE 114.2707673
    6 KAN 113.0626838
    7 GNB 112.6910923
    8 CIN 110.7024429
    9 DET 110.7000645
    10 NWE 109.5333214
    11 SDG 108.5837549
    12 MIA 108.2077689
    13 CAR 106.686108
    14 PIT 101.4891776
    15 CHI 101.2509308
    16 NYG 100.8233038
    17 ARI 100.8132825
    18 IND 100.4307652
    19 BAL 96.2746793
    20 MIN 93.95317722
    21 TAM 93.19978483
    22 STL 89.9821243
    23 SFO 89.81560333
    24 PHI 89.42324244
    25 TEN 88.32085035
    26 HOU 85.98515408
    27 OAK 85.66933924
    28 SEA 85.17872006
    29 BUF 78.45927586
    30 WAS 75.41535232
    31 NYJ 65.64714964
    32 JAX 61.25504247

    Not a lot of surprises here, except I expected the Vikings to rank higher because of Adrian Peterson’s oft-advertised low-carry total (a three game stretch of 10, 13 and 13 games) as well as the fact that despite their terrible efficiency, they have been in close games. I suppose I also expected Kansas City and Cleveland to rank lower because of the proficiency of their passers—especially because Kansas City has Jamaal Charles. Given that they rank fifth in game script, they could simply be passing despite the lead in order to maintain “balance”.

    Obviously, the above numbers are not adjusted for strength of schedule, because that would not likely provide useful data when it comes to passer IDs.

    • Chase Stuart

      Awesome work again! Thanks for putting this together. I agree that these should not be adjusted for SOS. No surprise to see Dallas at #1, but I like seeing the confirmation!

  • Arif Hasan

    Opponent Passer IDs. Not sure if I did this right, but this could be used to determine how coaches see differential defensive weaknesses. If a defense is very strong against the run and very weak against the pass, coaches could prefer high-percentage, low-yardage passes instead of runs in order to drain the clock.

    oPR ID Rk Team oPR ID
    1 NYJ 131.59
    2 ARI 121.25
    4 WAS 120.29
    3 MIN 120.67
    5 NYG 115.64
    6 PHI 115.41
    8 JAX 112.51
    7 DET 112.91
    9 CLE 109.87
    10 OAK 108.68
    11 BAL 104.59
    13 MIA 104.13
    12 DAL 104.42
    14 TAM 102.17
    15 BUF 99.64
    17 ATL 97.88
    18 KAN 97.69
    16 DEN 97.97
    19 SDG 95.74
    20 CIN 93.19
    22 STL 91.05
    21 CAR 92.70
    23 TEN 90.76
    24 GNB 89.27
    25 NOR 88.91
    26 PIT 86.73
    27 IND 86.55
    28 HOU 85.12
    29 NWE 84.27
    30 SEA 81.78
    31 CHI 81.13
    32 SFO 75.51

    I am curious about the code you use to make things look like tables in comments, incidentally.

    • Arif Hasan

      Not sure how the ranks got messed up. Again, for posterity:

      Rk Team oPR ID
      1 NYJ 131.59
      2 ARI 121.25
      3 MIN 120.67
      4 WAS 120.29
      5 NYG 115.64
      6 PHI 115.41
      7 DET 112.91
      8 JAX 112.51
      9 CLE 109.87
      10 OAK 108.68
      11 BAL 104.59
      12 DAL 104.42
      13 MIA 104.13
      14 TAM 102.17
      15 BUF 99.64
      16 DEN 97.97
      17 ATL 97.88
      18 KAN 97.69
      19 SDG 95.74
      20 CIN 93.19
      21 CAR 92.70
      22 STL 91.05
      23 TEN 90.76
      24 GNB 89.27
      25 NOR 88.91
      26 PIT 86.73
      27 IND 86.55
      28 HOU 85.12
      29 NWE 84.27
      30 SEA 81.78
      31 CHI 81.13
      32 SFO 75.51

      • Chase Stuart

        Thanks. That is pretty interesting. I may try to do something with this later! I guess the results make sense: the Jets game script is not very good, but teams don’t really run against them. Meanwhile, the 49ers see a lot of runs, and they have a high positive game script. You wouldn’t expect that, because of the reputation as having such a dominant rush D. Food for thought, I suppose.

        To make it pretty, wrap it in tags like this (but don’t include the number 1, obviously):

        <1p1r1e1> text <1/1p1r1e1>