I was a nerd as a kid. I devoted a significant portion of my childhood studying math, Star Wars, comic books and Nintendo. To be honest, not much has changed. Subsequently, my introduction to the football world had to come at the intersection of those interests: video games. At age 11, I received a PlayStation 2 for Christmas with two games: Silent Scope and Madden 2001.
In case you don’t remember, the cover athlete that year was Eddie George, following the Titans’ infamous one-yard-loss to the St. Louis Rams. Growing up in Ohio, I now had an iconic Buckeye on the cover of a video game I adored. I became obsessed, and that’s how my love affair with football began.
So why did I take this stroll down memory lane? Well the point here, is that I came to love football through the running game.
Like many, I valued the physicality of the game, and the belief in a running back’s ability to physically assert their dominance in any given match-up. When we use descriptors like feature back and workhorse, it’s hard not to appreciate the unique talent they possess.
The admiration for running backs doesn’t stop with just their explicit contributions, but it stretches to the assumptions we make regarding their implicit contributions. Here, I’m referencing the notion that a strong running game will make life easier on a quarterback through the play-action pass. It’s not a new idea, and new Cowboys coach Mike McCarthy even referenced it in his recent press conference.
[protected-iframe id=”a4f700527045c5fde7d8cd2c5260c624-105974723-174174104″ info=”twsrc%5Etfw” class=”twitter-tweet”]
While the intuition here makes sense, sometimes, our intuition can lead us astray.
In the past, others have demonstrated that there is an overall lack of evidence supporting this conclusion. To this point, the general finding has been that play-action passing works regardless of how well one runs the football.
Using the SIS DataHub, I pulled data from 2015-2019 to compare the relationship between a team’s EPA per Rush with their EPA per Play-Action Dropback. The results are in agreement with previous findings, as there’s very little relationship between the two. If we compare success rates, we see a slightly stronger relationship, though still weak by most standards.
While there may be a tiny bit of correlation between these variables, there are actually others that correlate much stronger. Below, you’ll find a correlation matrix.
If you’re unfamiliar with the idea, it’s just meant to display a group of variables, and how strongly related they are. Numbers close to zero indicate that there’s not much of a linear relationship, while values close to 1 or -1 indicate a strong linear relationship.
Elements on the diagonal are the correlation of a variable with itself, so naturally these values are 1.
In terms of the correlation matrix, the variable that actually correlates most strongly with EPA per Play-Action Dropback is PFF Passing Grade. This seems logical, since a quarterback that grades well overall should find success on play-action as well as non-play-action passes. We can observe this relationship in the plot below.
Here we see a larger clustering of points around the line, though there’s still some large errors in terms of what this line would predict.
When it comes to linear models, an “error” refers to the distance between the line (what we predicted) and the point (what actually happened). Points above the line (or positive errors) indicate situations where a team had a higher play-action EPA average than we’d expect given their PFF Grade. Likewise, points below the line (negative errors) indicate teams who had an EPA average lower than what we’d expect based on PFF Grade.
I know I’m getting into the weeds here, but stick with me.
We’ve seen up to this point that there isn’t a general relationship between effective rushing and effective play-action. But does this account for how often a team commits to “establishing the run”, or whether a team is better at rushing than passing?
In other words, we might expect those teams with solid running backs but sub-par quarterbacks to benefit more greatly from play-action passing, as defenses would be more apt to focus their attention on stopping the run.
In order to investigate this, we can take a look at the errors from the previous chart and how they correlate with some of these features.
On the left, the X-axis represents the difference in EPA for non-play-action dropbacks and rush attempts (i.e. the difference in a team’s ground game vs aerial attack). The X-axis on the right is simply how often a team ran the ball. In both charts, the Y-axis represents the errors from our previous chart.
The idea here, is that if “establishing the run” helped set up play-action, we’d see teams with heavier run rates or more successful ground attacks have more positive “errors”. But generally speaking, we don’t observe large differences in the error distributions for either chart. The takeaway here is that once we’ve accounted for a team’s quarterback capabilities through PFF Grade, rushing rate and rushing performance relative to passing stats doesn’t tell us much about the effectiveness of their play-action usage.
While this doesn’t follow our traditional instinctual understanding, I do think there are some possible explanations for this result. I don’t recall who it was, but I read a quote from a former NFL coach that said, “If a player is thinking on the field, he’s not playing.” While the spokesperson didn’t stick in my mind, the idea certainly did.
When we discuss great defensive players, an attribute that often sticks out is their instinctive capabilities. With the NFL moving at faster and faster speeds, a defender’s ability to read and react instantaneously is paramount. In other words, there isn’t time for conscious thinking, but only subconscious processing based on instincts from previous experience.
Most NFL players engaged in football at amateur level for anywhere from 5-15 years. During those formative years, it’s more than likely that the importance of running the ball and stopping the run was drilled into their DNA. After all, at the high school or college level, a great athlete at running back can completely alter a game. The point here is that an NFL defender’s instincts are built on the foundation of those formative years. For that reason, NFL teams don’t need to establish the run, because the threatening nature of it has already been instilled throughout the amateur ranks. Instead, they simply need to capitalize on the preexisting framework. It’s the threat of the run rather than the actual running game that makes the difference.
Obviously, if a team only ran play-action and never attempted to run the ball, defenses would begin to adjust. But the reality is that we’re nowhere near this threshold, and thus it would serve teams well to increase their play-action usage overall. If this were to happen, it’s likely we’d actually see more efficient ground attacks as teams adjusted to the new status quo. Until then, however, it’s the quarterback and the passing game that will continue to reign supreme.
[lawrence-newsletter]