Two minute drill

What it measures

Football with the clock as a second opponent. In the last two minutes of a half, offenses go up-tempo, defenses trade yards for time, and the value of every snap changes: an 8 yard completion in bounds can be a bad outcome, and an incompletion that stops the clock can be fine. The two_minute bucket isolates these snaps so hurry-up performance can be compared on its own terms rather than polluting (or hiding inside) full-game numbers.

Like every situation bucket, it overlaps the others: a two minute snap while trailing in shotgun counts in all three buckets.

How it is computed here

Offensive snaps with 120 seconds or less remaining in either half (from half_seconds_remaining) fall into the two_minute bucket. The situation_team view aggregates team EPA per play, success rate, yards per play, and pass rate over those snaps; situation_qb aggregates passing efficiency for qualifying quarterbacks. EPA itself is clock-aware through the underlying expected points model, so end-of-half desperation snaps are judged against appropriately low expectations.

How fast it stabilizes

It mostly does not. A team gets a handful of two minute snaps per game, so even a full season's bucket is one good afternoon's worth of plays. A quarterback's two minute dropbacks in a season can be fewer than 40. These splits are best used to describe what happened (who actually moved the ball in hurry-up this year) rather than to claim a repeatable clutch skill.

How it gets misused

• Building a "clutch" narrative on 30 snaps. The sample sizes here cannot separate skill from one good drive against a soft prevent defense.
• Comparing two minute EPA to base offense EPA as if the conditions were equal. Play-calling constraints, defensive shells, and clock effects make it a different game.
• Ignoring score state. Trailing two minute offense (throwing against the sideline) and leading two minute offense (kneeling) are opposite tasks that share a bucket.