Expected points added (EPA)

Expected points added measures the value of a single American football play by how much it changes the offense's expected number of points from the current drive, based on down, distance, field position and time remaining. A play is assigned an EPA value equal to the expected points after the play minus the expected points before it.

The underlying expected-points model is built from historical play-by-play data: for any down-distance-field position situation, it estimates the average net points the possessing team will score before the next score, by either team. A first down deep in opponent territory carries a high expected-points value; a third-and-long near one's own end zone carries a low or negative one. EPA is continuous and per-play, unlike success rate, which only records whether a play cleared a fixed threshold, not by how much.

EPA is the foundation for most modern football analytics: summed or averaged per play, it ranks quarterbacks, offenses and defenses, and it underlies fourth-down and two-point-conversion decision models built on win probability model outputs. The main pitfall is comparing raw EPA totals across different game contexts without adjustment; garbage-time plays in blowouts, and differing opponent quality, can inflate or deflate EPA totals in ways that don't reflect true team strength.