Defensive Run-Pass Balance

I've been writing a lot about offensive run-pass balance lately, recommending that most teams should pass more often. So far I've ignored half of the equation--Defensive strategies matter too.

Unlike offensive play choices which are discrete defined options (such as either run or pass), defenses have a continuum of options ranging from a prevent defense to a goal-line stand. Defenses don't have to commit to one or the other. Instead, they can bias their strategy anywhere from 100% pass to 100% run.

With play-by-play data, we know what option an offense chose. Unfortunately, without coach's film of every game and an intimate knowledge of a teams' defensive schemes and intentions, it's nearly impossible to know where on the continuum a defense stood on any given play.

Still, it may be useful to see how often a team was run against or passed against, and see how successful each play type was. The table below lists each defense's proportion of passes faced on first downs. Also listed is the average success of each type of play as measured by Expected Points Added (EPA). The difference in average success between passes-against and runs-against roughly measures the imbalance in play selection.

For example, Detroit has faced 43% passes on first downs. Passes against them have averaged 0.65 EPA and runs have averaged 0.00 EPA. Perhaps teams should be passing more often against the Lions. The Vikings show the opposite indication--passes against them are far less successful than runs. Maybe offenses should run more often against the Vikings on first down.

For defenses, negative EPA numbers indicate better performance. In the EPA Difference column, high positive numbers suggest that offenses should pass more against the listed defense, and high negative numbers suggest that offenses should run more. You can click on the table headers to sort.


Team
% Pass
vs Pass EPA
vs Run EPA
Total EPA
EPA Diff
DET
43%
0.65
0.00
0.28
0.65
SF
55%
0.27
-0.06
0.12
0.33
ATL
42%
0.27
-0.02
0.10
0.29
PIT
48%
0.16
-0.13
0.01
0.28
CIN
56%
0.19
-0.05
0.09
0.24
MIA
45%
0.14
-0.10
0.00
0.24
CLV
47%
0.20
-0.02
0.08
0.22
CHI
49%
0.20
0.03
0.12
0.17
GB
37%
0.07
-0.10
-0.03
0.17
TB
39%
0.15
-0.01
0.06
0.16
BLT
46%
0.14
-0.02
0.05
0.16
ARZ
59%
0.02
-0.13
-0.04
0.15
SL
49%
0.18
0.03
0.10
0.15
TEN
51%
0.21
0.08
0.15
0.12
DAL
46%
-0.05
-0.16
-0.11
0.10
NYJ
43%
-0.03
-0.13
-0.09
0.10
OAK
45%
0.02
-0.06
-0.02
0.08
HST
54%
0.04
-0.03
0.01
0.07
NYG
55%
0.01
-0.05
-0.02
0.07
DEN
46%
0.07
0.03
0.05
0.05
CAR
43%
-0.10
-0.08
-0.09
-0.02
SEA
41%
0.10
0.15
0.13
-0.05
NE
45%
-0.06
-0.01
-0.03
-0.05
KC
51%
0.04
0.12
0.07
-0.08
JAX
44%
-0.08
0.01
-0.03
-0.09
PHI
51%
-0.06
0.04
-0.01
-0.10
IND
43%
-0.16
0.00
-0.07
-0.15
WAS
43%
-0.02
0.18
0.09
-0.20
NO
52%
-0.05
0.20
0.07
-0.25
BUF
39%
-0.19
0.09
-0.02
-0.28
SD
51%
-0.08
0.23
0.07
-0.30
MIN
42%
-0.57
-0.03
-0.26
-0.54
Avg
47%
0.04
-0.01
0.02
0.05

Against the Lions, offenses have scored an average of about 0.65 more points by passing on first down than by running on first down. Against the Vikings however, the reverse is true. Offenses have scored 0.54 fewer points by passing rather than running. But their respective opponents have chosen to pass against both teams at nearly an equal rate, about 43%. Does that make any sense? Why aren't offenses exploiting the relative strengths and weakness of their opponents?

Note: I limited the data to plays when the score is within 10 points, and to the first and third quarters when time is not a factor. This way, the effects of 'trash time' and hurry-up offenses are excluded. In this analysis I also limited the data to plays between the 20-yard lines.

  • 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

10 Responses to “Defensive Run-Pass Balance”

  1. Mark M says:

    Sounds to me as though Detroit loads up on the run on first down. Play action anyone? And perhaps draw plays against Minnesota would work on 1st down. Interesting stuff.

  2. James says:

    This says the Cowboys has the 2nd best defense, but they are the 21st DRANK team. What do you think is the best explanation for that casual disrepancy?

  3. Brian Burke says:

    Good point. I think it's probably due to a couple things. Sample error for one. This post is limited to when the score is within 10 pts and the game is in the 1st or 3rd qtrs. The DRANK includes all plays. The second reason is opponent-adjustment. I'd guess the Cowboys are getting a bonus for tougher than average opposing offenses.

  4. Guy says:

    More great analysis, Brian, both here and in the post on team offenses. Even with limited sample sizes, I'm sure many of these large pass/run spreads reflect the reality for those teams.

    However, I'm struck by the fact that the correlation between run EP and pass EP is slightly negative, both on defense and offense, not strongly positive as game theory would predict. In theory, for each team the run EP, pass EP, and total EP should be virtually identical. Yet we observe nothing like that. When reality and theory are this far apart, we can conclude that reality is wrong -- i.e. teams are making massive strategic errors both on defense and offense -- or we can conclude that our theory has a problem. I think the latter must be the case here (which doesn't mean teams aren't also making some mistakes).

    Clearly, there are massive differences in talent by team in terms of their ability to pass, and significant but much smaller differences in their ability to run. These talent differences must be much larger than the impact of the defense shifting its pass/run orientation. So even when defenses try to stop Indianapolis’ or Minn's passing game, they simply can't do that enough to bring the run EP and pass EP into balance. Or at least, they can’t do it without paying a very large price against the run. (For example, maybe frequent blitzes would bring down pass EP, but at too high a price.)

    That leaves the question, why don't strong passing teams pass even more than they do, especially against teams with weak pass defenses (and vice versa)? To some extent it's probably a function of the physical limits of the best receivers. (Especially on 1st downs immediately following successful pass plays.) And perhaps these teams should pass more. But I don't think it's a realistic model to expect a great passing team will have equal pass EP and run EP.

  5. Mark M says:

    Guy.. why not? The theory makes sense to me. If a team's pass EP is higher than its run EP in all cases then why ever run the ball?

  6. Brian Burke says:

    I think the reason that good passing teams don't pass against weak pass-defensive teams more is because they want low-variance outcomes. They will play conservatively and let the weaker opponent make mistakes.

    I think also that running is generally inelastic. No matter how much or little the defense focuses on stopping the run on first down, it's going to allow 3-5 yards in the vast majority of instances.

    For many offenses, I think passing has become a dominant strategy. No matter how much the defense is geared to tee-off with a pass rush and a dime-package scheme, passing may still be better.

  7. Brian Burke says:

    For all team-years from 2000 through 2009 (wk 11), the correlation between offensive run and pass EPA (between the 20's) is 0.20. For defensive run and pass EPA, it's 0.04. Not exactly sure how to interpret that. It does fit with what I've observed previously. Offenses generally drive the show, while defenses are more or less "there." Offenses are more determinative of outcomes than defenses.

  8. Anonymous says:

    Wow, the thing that really strikes me about this data is just how bad the Lions defense is. It's not just that they're bad, it's how much worse they are than every other team in the NFL. Remarkable. Have you done team-by-team analysis of this data for past years? Have other teams been as bad as the Lions are this year?

  9. Doug says:

    Brian,

    When we start breaking things down at a team-by-team level, is there still enough data for the differences shown to be statistically significant? Especially since so much data has already been excluded by the 1st and 3rd quarter within 10 points between the 20s restriction. I was just wondering, before we go too far in trying to explain stuff that could just be random noise.

  10. Brian Burke says:

    For most teams, there are about 100 cases in their sample, so for some teams in the middle, with smaller differences we can't be sure. For the overall league, definitely, there is certain significance.

    I'm not trying to explain each individual team's difference, but the league's overall discrepancy needs an explanation. When it comes to each individual team, we aren't looking to pronounce a scientific truth. Instead, each team is forced to make a decision about its own playcalling. It has to make a decision based on the data available, and this is the best we've got. For now, I'd just say either "run more" or "pass more" without much certainty over the actual numbers.

Leave a Reply