We’re only recently starting to have intelligent stats-based conversations about offensive line play. Pancakes aside (thanks Madden), the public has had to work mostly with hurries, hits, knockdowns, and sacks, and limited efforts to attribute those sorts of results to the offensive linemen themselves. 

That background makes the line between offensive line play and quarterback play blurry. Just search “sacks are a QB stat” on Twitter and you’ll see plenty of mentions over the last few years. There’s a mix of truth and deception behind any hard-and-fast heuristic like that, but the underlying concept is that pressure on the quarterback isn’t nearly as much the fault of the line in front of him as most people think. 

Don’t Blow It

Sports Info Solutions video scouts chart the primary reason for each sack, breaking it down into a handful of categories. In general, the goal is to use as many objective signposts as possible in determining this, so while we have a category for “QB fault”, that’s primarily reserved for the obvious cases like egregious pocket awareness or simply dropping to the ground on his own, which hearkens back to Peyton Manning’s heyday of sitting down before he could be overrun.

Sacks by Primary Reason, 2019-20

Sack TypeRate
Blown Block45%
Blitz/Overall Pressure18%
Coverage Sack13%
Rusher Untouched13%
Failed Scramble10%
QB Fault/Self Sack1%

Nearly half of sacks are the direct result of a blown block, which we consider a lineman being physically beaten within the first few seconds of a play. That’s a category that’s fairly obviously not tied to the quarterback. On the other hand, we could have a long conversation about how much responsibility the QB has over the other categories, and in the end it depends on the situation much of the time. For now, we’ll operate as though blown blocks are owned by the line, and from there we can start talking about the quarterback.

You might say that “pressure is pressure,” and if we care about quarterback performance we care more about whether there was pressure than whether he was responsible for it. And to that, I say…well, look at this table. Pressure that results from a blown block is much worse for a quarterback than other kinds of pressure. 

EPA Under Pressure, 2019-20

Pressure TypeEPA per Pass AttemptEPA per Dropback
No blown block-0.03-0.17
At least one blown block-0.17-0.65

Just to give a sense of how splitting out blown blocks can help us understand a quarterback’s ownership of pressure, let’s take a look at the signal-callers who have been pressured on at least 35 percent of dropbacks this season.  

Quarterbacks Pressured on at Least 35% of Dropbacks, 2020 (through Week 6)

QuarterbackBlown BlockNo Blown BlockTotal
Justin Herbert13%32%45%
Daniel Jones15%27%43%
Kirk Cousins15%27%42%
Lamar Jackson6%32%38%
Sam Darnold9%29%38%
Carson Wentz9%28%37%
Russell Wilson6%30%36%
Josh Allen7%29%36%

Consider the difference between Jones and Cousins and the names below them. Blown blocks have contributed much more to the pressure they’ve faced, but every one of the five players at the bottom has a higher pressure rate without a blown block. These are largely players who are known for being able to get out of trouble, but they also get themselves into trouble. Justin Herbert, for his part, has paired suspect blocking with self-imposed tribulations.

The Clock is Ticking

One way to evaluate how quarterbacks are responsible for and respond to pressure is with timing data. SIS charts snap-to-throw and snap-to-pressure times to help answer these kinds of questions. For this portion of the discussion, we’ll focus on 2020 only.

The plot below shows how the odds of a successful play (Positive%, or the percentage of plays with a positive EPA) fall as a play goes on, while pressure goes from 20% to 80% likely in the span of three seconds. Correspondingly, the Boom% (percent of plays adding at least one expected point) and Bust% (percent of plays losing at least one expected point) both increase with time, although the odds of a big play for the offense drop after the first few seconds.

To start looking at the split of responsibility between offensive line and quarterback, we can see how blown block rate compares to the time at which pressure occurs.

Pressure that occurs in the first two seconds of a play is much more likely to be the result of a free rusher, while pressures in the two-to-three second range are more often the result of a failure to sustain blocks. The two-and-a-half second mark is a clear peak, though, with blown blocks becoming less and less tied to pressure as the play progresses from there. That also happens to be the threshold used by the suite of Next-Gen-Stats-based “Win Rate” metrics to determine if a block was sustained.

So we know that a quarterback should get the ball out in around three seconds or fewer, or else he’s in territory where he’s more likely to be pressured than not and more likely to have a negative play than not. Some of that comes down to a quarterback making quick reads, but some of that is scheme-based. Can we figure out which players are holding the ball too long given their context?

To get at that question, we can create an estimate for how a quarterback’s internal clock should work based on the types of drops he takes (shotgun, play-action, number of steps), and compare to how quickly he actually gets the ball out. In this case, the “clock” is just the average time-to-throw for each drop type.

Slowest and Quickest Triggers, 2020 Quarterbacks

Snap-to-throw time compared to expectation based on the type of dropback

QuarterbackSnap-to-ThrowExpectedDifference
Josh Allen3.142.670.47
Baker Mayfield3.22.840.35
Lamar Jackson3.012.730.28
Daniel Jones2.912.630.27
Patrick Mahomes2.972.790.18
----
Ryan Fitzpatrick2.392.73-0.27
Matthew Stafford2.462.85-0.29
Drew Brees2.672.98-0.31
Philip Rivers2.512.88-0.37
Dwayne Haskins2.342.74-0.40

We’ve seen some of the names at the top of the list before! And for those curious, Wilson and Wentz appear just outside the top five slowest triggers. In a sense, we’ve triangulated a player’s tendencies by considering his blown-block-less pressure rate and his time-to-throw considering the depth of his drops, landing on a handful of consistent play-extenders. 

But anyone looking at the set of late throwers will realize that we haven’t just captured a tendency to hold the ball, we’ve also captured a player’s willingness/ability to maneuver within and escape the pocket. 

It turns out that the players who do tend to hold the ball longer make that work for them. Ignoring the value gained when they scramble (which is of course relevant in this conversation), those players have success late in the play in a way that the rest of the league does not. And they’re average on quick throws.

EPA per play by Snap-to-Throw Time, 2020

Quarterbacks< 2 sec2 - 3.5 sec> 3.5 sec
Allen, Mayfield, Jackson, Jones, Mahomes0.10.210.12
Others0.10.180.06

By contrast, the quick-trigger passers on the list have had good reason to get rid of the ball quickly because they have been much worse on throws late in the play and better on earlier throws.

EPA per play by Snap-to-Throw Time, 2020

Quarterbacks< 2 sec2 - 3.5 sec> 3.5 sec
Haskins, Rivers, Fitzpatrick, Brees, Stafford0.150.2-0.28
Others0.10.190.09

The Last Word

Obviously, we don’t want to just issue a blanket statement of “get yourself a man who can extend plays with his athleticism.” As shown earlier, holding the ball leads to pressure, pressure leads to sacks, and sacks lead to the dark side. But this evidence suggests that the quarterbacks who make it to the NFL level with that kind of tendency can put themselves on the positive end of the high-variance gamble that comes from letting plays extend past a few seconds.

The other key takeaway is that quarterbacks own some responsibility for their pressure stats, but pressure that comes from the offensive line’s failings leads to much worse outcomes than pressure that the quarterback has some responsibility for. Using offensive-line-specific pressure data to account for the split in responsibility between line and quarterback gives us better tools when evaluating pressure stats.