Score state

What it measures

Game script, made queryable. Our situation views bucket every offensive snap by the score at the snap, from the offense's perspective: leading, trailing, or neutral (tied). Score state shapes everything downstream: trailing teams throw more, face softer two-deep coverage, and rack up yards that arrive too late to matter; leading teams run clock and happily punt on efficiency. Splitting by score state lets you ask what an offense looks like when the game is actually in doubt.

The neutral bucket is the closest thing to a laboratory: play-calling is unconstrained, defenses play honest, and tendencies (like early-down pass rate) reflect philosophy instead of desperation.

How it is computed here

The score differential at the snap assigns each play to leading, trailing, or neutral, where neutral means tied. Like all situation buckets these overlap with the rest of the vocabulary (a trailing snap in the two minute drill counts in both), and the situation_team and situation_qb views aggregate the same efficiency columns over each bucket.

How fast it stabilizes

The buckets are big compared to red zone or two minute samples: a team collects hundreds of snaps across the three states in a season, though the split is uneven for very good and very bad teams. The practical wrinkle is composition rather than size. A dominant team's trailing sample is small and skewed toward its worst afternoons, so cross-state comparisons within one team can mislead.

How it gets misused

• Garbage-time inflation. Season totals compiled while trailing big are the classic empty-stats trap; checking a player's trailing versus neutral splits is the two-click audit.
• Calling a team "front-running" from raw splits. Everyone is more efficient while leading; the defense is playing the clock, not the down.
• Forgetting that score state is endogenous. Good teams spend more snaps leading because they are good. The split describes context, it does not explain quality.