Every Thursday, we get ready for the weekly slate of NFL games by looking at the most exciting trench matchups of the week. Which quarterback will be kept clean, and which one will have to run for his life? To answer this, we project survival curves of offensive lines — we project how often a quarterback will be pressured at a given time after the snap.
Going into each game, a machine-learning model incorporates a lot of features in order to come up with a prediction:
- The survival curve of the offensive line in previous games (or from last year)
- The survival curve of the defensive line in previous games (or from last year)
- The PFF pass-blocking grades of the five offensive linemen
- The PFF pass-rush grades of the pass-rushers, and where they are projected to line up
- How the blend of individual grades and team-level survival curves from the past are weighted toward predicting the cervical curve depends on continuity and how far in the past the information lies. Less continuity means the individual grades are weighted more heavily.
- Schematic factors such as how often the defense is projected to blitz and how often the offense is projected to use play action, designed rollouts or six-plus-man protections.
[Editor's Note: PFF's advanced statistics and player grades are powered by AWS machine learning capabilities.]
Each week, we will use our projections to highlight the biggest mismatches in the trenches. We will also identify an interesting matchup to watch.
Week 3 was a tough week for the model. The prediction that the Chicago Bears offensive line would be able to keep their quarterback(s) clean wasn’t entirely proven wrong — they were indeed above average — but nine teams had a better quick-pressure (pressures allowed within 2.5 seconds) rate.
The negative game script for the Bears — they were forced to pass 57 times — probably didn’t help, as must-pass situations certainly favor the defensive line. Adjustments for game script are certainly something we should consider in the future.
We also predicted that Deshaun Watson would be under fire the whole game, but the Pittsburgh Steelers pressured him within 2.5 seconds only at a slightly-above-average rate (14th-highest of the week). This was also caused by the Steelers blitzing at a much lower rate than in previous weeks, as they cut their usual number of blitzes almost in half. Instead of sending the house, they may well have thought it better to have Watson hold onto the ball.
Here are our projected mismatches for Week 4:
Favoring the offensive line: New Orleans Saints OL vs. Detroit Lions DL