With / Without

How it is counted

Each split compares whole games. A game counts as "with" when the player took at least 25% of his unit's snaps and "without" when he took none (injured, inactive or not yet in the lineup) and did not play for another club that week. Games where he left early or played a reduced role are left out of both. Key players are each team's snap leaders by position (quarterback, lead back, top three receivers, tight end, two tackles, two edge rushers, an interior lineman, a linebacker, two corners and a safety) who average a starter's share of the snaps. A split appears only with at least two games with him and one without. EPA per play and success rate count pass and run plays from the nflverse play-by-play; points come from the final score. Offense players are measured on the offense, defenders on what the defense allowed, so a positive number always means the team's results were better in the games with him.

Sample size. Every card prints the games and plays in each group. It carries a "small sample" badge when either group has fewer than 3 games or fewer than 100 plays: read those as a signal, not a verdict. More on sample size.

Context. Each card lists the opponents in both groups with their average current team rating (points above or below an average team on a neutral field, the same rating behind the playoff odds), the other key starters who were out in those same games, and the team's EPA per play in up to 4 games before the player's first absence against up to 4 games from it on, to show whether the team was already trending before he went out.

What it is not. These are observed differences between two groups of games, not the player's causal effect: opponents, other absences, weather, game script and form all change between the two groups.