Expected points added per play, cut by situation. A play is a dropback (sacks and scrambles count) or a designed run. On offense higher is better; on defense it is EPA allowed, so lower is better. Cells are tinted against the league in that same column, so a team is compared with what everyone does in that situation, not with its own average. Small numbers under each value are plays; faded cells have fewer than 30.

Home / away — offense

Click a column to sort; hover a cell for success rate and standard error. Click a team for every split at once.

Reading it. EPA/play is noisy: its standard deviation is about 1.4 points a play, so 100 plays carry roughly ±0.14 of noise and a single season's split of a few hundred plays still moves ±0.07–0.10 by chance. Situations also overlap with team quality: good offenses spend more time leading, and teams trailing late face defenses playing soft. Treat a split as a description of what happened, not a trait, until it holds across seasons. Source: nflverse play-by-play, rebuilt daily.