A box score counts plays. A staff counts drives. This reads the season one possession at a time — where it started, how long it lived, and whether it ended in points — and then takes the starting spot back out again, because most of what people call a scoring offence is a story about field position wearing an offence's name.

Every regular-season drive of 2025: 5,700 possessions that ran a play from scrimmage, read at drive grain rather than play grain. 45 further drives never ran a snap — kickoff returns for a score and end-of-half possessions — and a drive with no snap has no starting spot and no play-caller, so they are out of every figure below. Painted on the Monday weekly sync; there is no play-grain live feed in this product, so nothing here is live.

Getting close is not the same as scoring, so both counts are stated: 1,797 of those 5,700 drives (31.5%) reached the opponent's 20, and 1,014 of those 1,797 red-zone trips (56.4%) ended in a touchdown. Both divide by the same denominator as every rate above — drives with at least one offensive snap — and both are counted from nflverse play-by-play, painted Monday 04:00 ET.

The WP column prices a drive in win probability rather than points: nflverse's own play-by-play model, summed over the drive's snaps and averaged over the split, painted with everything else here on the Monday 04:00 ET sync. In every drive the league itself averages +1.0 a drive, not zero — a possession-signed swing slightly favours whoever holds the ball — so a team is read against the other 31 rows and not against a zero line. The model reaches 98.9% of scrimmage plays; a drive it never priced prints an em dash rather than a nought.