As the NFL has become more of a passing league, offensive lines are built around protecting the quarterback and defensive lines are built around chasing the quarterback. "Pass protection" and "pass rush" are phrases thrown around to represent something that we haven't measured well. But how to define a pass protection metric (and its defensive corollary, the pass rush) is something that we have to ask ourselves in analytics. To make it measurable, to understand the influence of linemen, to improve our QBR metric, "pass protection" has to be translated from the words into pass protection numbers.
As useful as it is for us to ask the question, it is more important for offensive and defensive coordinators to ask themselves what pass protection and pass rush mean. If it is as important as so many analysts and insiders say, the guys controlling the game need to have their own ways of evaluating it. After all, if you want to manage it, you better measure it.
But offensive coordinators haven't been asked to measure pass protection. They have always used a surrogate stat, such as sacks allowed. Defensive coordinators similarly haven't been asked to measure pass rush; instead, they have used sacks, hurries and knockdowns as surrogates. They know there is more to it, and that's where the (cliché coming) "eye test" comes in. Analytics can do better than these surrogate stats -- particularly incorporating other information -- and thus can get closer to a metric that is manageable and more reflective of actual on-field play.
So we want to define a measure of pass protection/pass rush that captures important elements of overall line-of-scrimmage control. The measure defined here might not be the only way, but it has a lot of useful information. And, as we've run it by people we know in the NFL, it has been reflective of their perceptions.
To illustrate the concept we are trying to represent, let's offer up some scenarios:
In this case, did Indianapolis' pass rush control the line of scrimmage or did Peyton do what Peyton always does -- take away control just as you think you have him? If Peyton completed the pass, does it matter? What if Peyton hadn't been flattened, but Mathis pulled up and just lightly hit him, in the way that NFL players can hit anything "lightly"?
This example was meant to be vague, begging for more info. So here is another.
Scenario No. 2: Detroit Lions DL Ndamukong Suh faces an early double-team but eventually bullies his way through, flushing Cincinnati Bengals QB Andy Dalton outside after three seconds in the pocket. Dalton gets an errant pass away before Suh gets close enough to hit him. Suh is frustrated and has to tell himself, "I'm good enough, I'm smart enough, I won't hit the quarterback."
Suh exhibits some self-control here, but did Detroit's pass rush have line-of-scrimmage control? Suh took three seconds to actually force Dalton to have to move from the pocket. The average time before a quarterback releases the ball or runs is about 2.75 seconds, and Dalton still took a little time after getting flushed to throw.
As with Mathis, at least Suh forced Dalton to feel pressure, which is a significant factor in lowering the effectiveness of the play. A throw under duress is typically completed only about 44 percent of the time, whereas a throw without pressure is completed about 66 percent of the time.
(Let's also define "feeling pressure." Quarterbacks are under duress when the pass rush forces them from the pocket, alters their throwing motion or if there is a defender clearly in their line of sight. A quarterback hit is when they get hit while throwing the football. These things plus sacks all constitute "feeling pressure.")
Scenario No. 3: The Arizona Cardinals send a cornerback and a linebacker on top of their three-man rush to go after Seattle Seahawks QB Russell Wilson, but the offensive line holds them off for 2.5 seconds in the pocket, allowing Wilson to find WR Golden Tate for a first down.
In this case, Arizona sent extra pass-rushers and Wilson had 2.5 seconds to throw -- less than the league average -- with no pressure getting to him. Maybe Seattle had to keep an extra pass-blocker in for the blitz (and that matters some), but the Seahawks' pass protection mostly beat the Cardinals' pass rush this time.
The point of the examples is to show that pass protection isn't a yes/no thing. There are partial degrees of protection and, hence, partial degrees of a successful pass rush. Sacking a quarterback after two seconds in the pocket is a pretty dominant pass rush. Getting pressure on a quarterback at 2.5 seconds to flush him from the pocket is not quite as good. Using five pass-rushers to get to a quarterback in the same time that four pass-rushers do it is a little less effective pass rush because it required a sacrifice elsewhere to accomplish it.
Defining pass rush and pass protection through examples like this is leading toward the use of the league-average pressure charts from this article looking at Eli Manning and Ben Roethlisberger, which is modified to look across the league and shown below.
When pressure occurs early, that means impact far above the average line in the chart. When it occurs late, that means it is only a little above the average line. If no pressure occurs and it's in the first second or two, that's not a big deal -- that's pretty average. If no pressure occurs after five seconds, the pass protection was quite successful.
At the end of a game, there is a pass protection or pass rush value ranging between 0 percent and 100 percent, with 100 being best and 50 being average. For instance, in the Sept. 29 game between the Kansas City Chiefs and the New York Giants documented in the above article, the Giants had a pass protection value of just 34 percent. On the defensive side, their pass rush wasn't much better, controlling the Chiefs' pass protection on just 40 percent of the passing plays. For Kansas City, it then had a pass rush value of 66 percent and a pass protection value of 60 percent.
As valuable as pass protection/pass rush are, we also know that it tells only a part of the passing-game story. Quarterbacks and receivers definitely matter on the offensive side, while the pass coverage matters on the defensive side. The seasonal correlation between pass protection/pass rush and passing-game expected points added (EPA) is between 30 percent and 40 percent. The lack of perfect correlation is suggesting the value of offensive skill players and of the defensive secondary, a study for another day.
With this solid correlation between pass protection/pass rush and the overall success of the passing game, it's easy to see why people talk about pass protection and pass rush to the degree that they are a common cliché. We hope that quantifying the concept allows it to become a meaningful metric for which to better measure and understand the game.
Statistics from Pro Football Focus were used for this article.