website statistics
A few weeks ago in my team efficiency column, I wondered why teams were beginning to run more in the red zone.  Although passing continues to eat up an increasing proportion of offensive play calls, teams were on pace to call the most red-zone running plays since 2008.  Brian has done previous research suggesting that teams should be running the ball significantly more in the red zone on first down, upwards of 85 percent inside the 10-yard line.  As infuriatingly inefficient fade routes continue to land harmlessly on stadium turfs around the league, are teams beginning to adhere closer to what the numbers suggest?

First, we need to look at whether running the ball is actually still more favorable in the red zone.  Brian's data looked only at first downs in the above research.  To get started, let's compare the Expected Points Added (EPA) and Win Probability Added (WPA) generated from all pass and run calls in the red zone, regardless of down.  A few things to keep in mind when looking at the data:

- There is going to be variance in the data, especially as we hone in on more specific situations with a smaller sample size.  Some unexpected changes in EPA or WPA may be the result of that.

- These are based on the 1.0 Win Probability models.  Brian has since updated the model to account for significant changes to the game, so actual outcomes may not fit these estimates, especially since the data goes back to 2008.

- Our focus here is the difference in EPA or WPA between passing and running.  If the goal is to maximize a team's odds of winning, the coach is simply asking himself, "What is the more favorable option?"

- The data is not limited to "normal" situations (i.e., times when game situations will not dictate play calling).  There will be a bit of noise from that.

Note that all data are from the 2008-13 regular seasons, and that field goals, spikes, kneel-downs and aborted plays are excluded from the sample set.  To illustrate where certain calls may be more favorable, we will break the red zone into seven three-yard chunks (the 21-yard line is therefore included as "red zone territory" throughout this article).  Here's what the run vs. pass EPA and WPA charts look like from that time period:

Looking first at the EPA graph, we can see right away that running the ball is more favorable at all levels of field position.  This is especially pronounced near the goal line.  Virtually every passing staple rely on both horizontal and vertical stretches, but the latter option is gone when the space is that narrow.  The chasm begins to dissipate beyond the 10-yard line, though it never totally disappears.

The WPA graph portrays a smaller advantage for running, though the pattern is wonkier. Evaluating large samples of WPA is going to be problematic because it is particularly prone to game score and time, unlike EPA.  But while it's hard to draw as many big-picture conclusions, it is telling that WPA supports the broad hypothesis that EPA proposes—running the ball appears to be more favorable in the red zone.

However, is it possible that that advantage persists simply because these graphs look at all downs?  Teams who pass in the red zone might do so in less favorable situations, thus deflating the passing numbers.  It's interesting that the EPA is negative at almost every area of the field, as it at least raises the question as to whether unfavorable situations like third-and-long are skewing the data. 

To partially combat this problem, we can narrow our scope to look at the run vs. pass EPA by down.  This will highlight more game situations when running and passing are closer to being equally favorable choices.  The process is more applicable for first and second down—running is always going to appear more favorable on third and fourth down, since teams typically limit run calls to short-yardage situations on those downs.  Still, we know that the third and fourth down data will be skewed, so if the first and second down data paint red-zone passing in a better light, that would be extremely telling:

Even on first and second down, running the ball remains much more favorable near the goal line.  However, that advantage essentially dissipates beyond the 10-yard line.  In fact, while the trend is inconsistent, it seems to suggest that passing may actually be slightly more favorable. 

A particularly interesting trend happens on first down.  Between the 10- and 21-yard lines, first-down passing plays have generated an EPA/P of 1.76 * 10-3 , whereas running plays have accrued 2.00 * 10-2 EPA/P.  Similarly, on second down, passing plays have outstripped running plays by an EPA/P of  -1.79 * 10-3 versus -5.83 * 10-2. If passing is roughly 10 times as favorable as running beyond the 10, the play calling tendencies on first and second down should resemble a funnel, leaning heavy on the pass before moving closer to a 50-50 split inside the 10.  Eventually, run calls should take over as the predominant choice near the goal line.  Here's how it actually played out:

First down play-calling appears extremely suboptimal, as coaches continue to lean heavily on the run and bring up second-and-long situations.  Interestingly, the tendencies for second down almost perfectly fit the theoretical model.  The overall EPA/P is still negative on second down, but that does not mean the play calling adjustments are not working.  EPA is almost always going to be negative in the red zone.  The expected points for a drive are going to increase as an offense approaches the opposing end zone (duh), so positive plays are not going to result in a big jump in EPA/P, while negative ones will have a disproportionately large effect.

If anything, it appears the play calling corrections have helped stave off bigger EPA/P dips.  Take a look back at the first graph in this article.  Though passing is more inconsistent, the EPA/P levels for both running and passing are virtually identical on both first and second down.  But second down is obviously less favorable than first down (except in rare situations), so the fact that the EPA/P decline is not more linear may potentially be attributable to inadvertently optimized play calling.

Indeed, the early-season emphasis on the run has dwindled a bit, as teams are now running on 46.4 percent of red-zone plays.  That's much more in line with what we've seen over the past five years.  The article I linked in the first sentence measured the rushing rate in raw play total; we are now on pace for 2,155 red-zone rushing plays in 2014, as opposed to the 2,449 I measured at the time.

Ultimately, while running the ball will likely always remain more favorable near the goal line, simply due to the inherent limitations of field dimensions, passing is infringing further and further into a traditionally run-favorable area.  This makes sense intuitively—passing plays are becoming shorter, and the proliferation of these quick man-coverage beating route concepts has forced defenses to cover larger swaths of the field.  Therefore, the tiny red-zone passing windows we hear about so often may not be so tiny anymore.