To have a competitive defense in the NFL, you need a formidable pass rush.
11 out of 14 playoff teams this past season ranked in the top half of the league in “quick pressures” (pressures that occurred in 2.5 seconds or less from the snap) per Next Gen Stats. In this most recent draft, we saw 8 offensive linemen drafted as teams want to bolster their protection units as a way to counteract increasingly sophisticated pass rush attack.
Evolving Pass Rush Metrics: From Sacks to Pressure Rate
Pressure Rate has been one of the more formative rate statistics in measuring pass rush performance given its higher relative stability year to year than sacks.
Pressure Rate is also a measurement that shows signs of predictiveness from college to pro for pass rushers. We can see below that collegiate pass rushers (150+ pass rush snaps) with higher pressure rates tend to perform better at pressuring the QB in the NFL in their first 4 years (controlled for team that drafted them), with a Pearson correlation of 0.43.
Positional Versatility and Pressure Rate
Pass rushers can also come in many different forms. Micah Parsons primarily played as an off-ball linebacker at Penn State and has taken the league by storm as a formidable edge rusher with the Cowboys. Jalen Carter had a productive rookie season and at Georgia was seen rushing the passer as a Defensive End (DE) in 3-4 sets, a traditional Defensive Tackle (DT) in 4-3 sets, a Nose Tackle, and sometimes was even used as an outside linebacker (OLB). Collegiate defensive coordinators are finding new and innovative ways to get a player a QB pressure.
Below are pass rush pressure maps by defensive gap alignment (as charted by PFF primarily since 2018) separated by primary game position (also provided by PFF). At the bottom you will see frequency of gaps primarily aligned to when rushing the passer (colored by most frequent). The top two percents are pressure rate by gap and pressure to sack rate (P2S rate), which would show the rate at which a QB pressure results in a sack.
As we can see in the pressure maps, regardless of your primary position, players should see snaps at various spots across the line. What we can glean from these pass rush charts is pass rushers can increase their pressure rates through more stunts or blitzes to confuse the offensive line given they are rushing the passer through a gap they normally don’t align in. One way to illustrate this is to examine the difference in pressure rates among defensive tackles (DT) relative to inside or off-ball linebackers (ILB) who primarily pass rush through the same gap (often in a delayed blitz).
There is a level of pre-snap match-up identity (i.e., OL calling out who to block) that directly eliminates the person in front of you in blocking, which would explain the DT difficulty vs ILB. We see a similar trend when you compare OLB vs DE. They generally pass rush through the same gap, but the difference of stand-up rushers versus hand in the ground rushers appears to provide better pressure rate figures for the stand-up rusher. What isn’t observed is the relative effect of DT & DE in terms of run defense, which would provide a counter factual to why you wouldn’t send an up-end pass rusher on every play.
The pressure maps highlight a few different trends with regards to relative pressure rate performance by gap, primary position, and coming from what can be thought of as a manufactured pressure situation (i.e., stunts or blitz).
Adjusting Pressure Rate with Play-Level Context
To confirm the relative effect of these indicators on pressure generation at the collegiate level, we constructed a Bayesian Generalized Linear Model to understand the relative adjustments needed in measuring pressure rate (in a simulated fashion using the Stan GLM model). We fit this model with several PFF charted variables to understand the play and game level context:
- manufactured pressure indicator (stunt or blitz),
- total pass rushers were involved in the play,
- players in the box pre-snap,
- gap the pass rusher primarily rushed the passer,
- player’s primary position,
- indicator if the play is third and long,
- leveraged situation in the game being in the last two minutes of the half,
- score differential at time of play,
- play action indicator,
- screen indicator,
- opposing offensive strength (model generated by SumerSports evaluating relative collegiate offensive performance on a rolling basis):
-
- below average (30th percentile and below)
-
- average (30-70th percentile; not shown given most will represent this value)
-
- good (70-90th percentile)
-
- elite (>90th percentile)
- And surrounding defense strength:
-
- below average (30th percentile and below)
-
- average (30-70th percentile; not shown given most will represent this value)
-
- good (70-90th percentile)
-
- elite (>90th percentile)
Values greater than 1 in this study indicate a higher likelihood that a variable will contribute to generating pressures. Players marked with these values are typically more effective at creating pressure. Specifically:
- Manufactured Pressures: Players involved in actions like blitzes or stunts are more likely to generate pressure. Interestingly, when these types of plays are excluded, the “non-manufactured” pressure rate reveals a slight increase in the correlation from college to professional success, measured at 0.454.
- Play Contexts: Players are particularly effective during third and long situations due to the obvious need for quarterbacks to hold onto the ball longer, enhancing the opportunity for pressure.
Moreover, being part of a strong defensive team—from good to elite levels—can also enhance a player’s ability to exert pressure. This cooperative effect confirms a trend observed in the pressure maps, where pressures are more likely as players move from inside to outside positions, with the left tackle (LT) typically allowing a lower pressure rate than the right tackle (RT) at the collegiate level.
These variables bring about the notion that play at the college level is very “noisy” and context at a play-level matters. Many scouts who watch and study the film on a given player know the game level context (time left, score differential, quality of opponent faced) and the play-level context (what type of play occurred, where was their pre-snap alignment) are important when going through their process of providing a grade on a player.
With our GLM model, we can infer the appropriate weights on each game/play indicator to properly adjust how difficult the pass rush rep was for a given player. This can help our college to pro evaluation process as we get a better picture of how a pass rusher performed.
Application: Evaluating NFL Draft Prospects
Applying this model to the 2024 draft class, we can get a “Pressure Rate Over Expectation (OE)”, which is pressure rate adjusted for game/play level context. In this class, we can see that Byron Murphy II (Seattle Seahawks), Jared Verse (Los Angeles Rams), and Laiatu Latu (Indianapolis Colts) lead the way in adjusted pressure rate relative to play/game level context (Pressure Rate OE). Latu, Verse, and Chris Braswell (Tampa Bay Buccaneers) led the class in pressure rates in non-blitz or stunt situations (non-manufactured). Braswell, Latu, Bralen Trice (Atlanta Falcons), and Chop Robinson (Miami Dolphins) lead the class in non-third long pressure rate (obvious passing situations). Lastly, Verse, Latu, and Robinson lead the class in pressure rate against good to elite offenses.
The notion of play/game level adjustment is imperative when looking at statistics at the collegiate level given the high variance of play. Understanding what can make the outcomes of the metrics more successful provides a framework or mental weighting that can be done when confirming a player’s ability through watching their film.
With regards to pressure rate, there are many other variables we could have included in the pressure rate model to understand player context but we need to be cautious with including all variables before we start to see multicollinearity (one or more variables are highly correlated with each other, which can cause irregularity in the model outcome and inflation performance).
One improvement to the model is to understand the player’s age at the time of the play given pass rushers will develop their arsenal of pass rush moves the more games they play in. We also learned that alignment at the play-level matters. Laiatu Latu was used on both sides but had a wider alignment at the D gap, which has a higher relative pressure rate than C gap (more traditional DE alignment). Byron Murphy was used at A-B gap equally (with almost equal success). Jared Verse almost exclusively rushed the passer from the D gap as a “hand in the dirt” DE, which is a wider alignment than a typical DE. One thing that holds true is their stability in generating pressure on the passer regardless of the play/game state. It will be interesting to see what techniques and alignment they will be schemed into at the next level given their collegiate success in generating pressures.
