EPA per dropback

What it measures

Passing efficiency with nothing hidden. Because the denominator is dropbacks rather than pass attempts, the plays quarterbacks would rather not count still count: sacks, throwaways, and scrambles are all in. That is what separates it from old-school rate stats like passer rating, which quietly excuse sacks. A quarterback who avoids negative plays gets credit for it here; one who takes ten sacks a month pays for them.

League average lives a little above zero (passing beats rushing on average), the best seasons run past +0.20 per dropback, and struggling passers sit below zero.

How it is computed here

For player rows, the mean of qb_epa across every play where the passer dropped back (qb_dropback = 1), which credits scramble value to the quarterback and does not charge him for a teammate's fumble after the catch. At the season grain it is passing EPA divided by dropbacks. Team columns use plain epa over all dropbacks, and league_baselines carries the league-wide per-dropback average for reference lines.

How fast it stabilizes

Faster than completion-based stats, but a real sample still takes months. Below roughly 200 dropbacks the number is dominated by a few plays and a few opponents; that is four to six games for a starter. Full-season starter samples (450+) are trustworthy. Within situation buckets (red zone, two minute) even a full season can be under 50 dropbacks, so treat those splits as descriptions, not skill estimates.

How it gets misused

• Ranking quarterbacks without a volume floor. An efficient 80-dropback relief stint is not a season. Our builder's minimum-volume gate exists exactly for this.
• Treating it as quarterback-only. Pass protection, receiver separation, and scheme all live inside the number. It grades the passing operation with the quarterback at the center.
• Comparing it directly to EPA per play. Dropbacks are a subset of plays with a higher average, so a passing number and an all-plays number are on different scales.