Expected Points Added — the change in expected scoring value from one play to the next.
Football analytics rests on one foundational idea: not all yards are equal. A three-yard gain on third-and-two is a drive-extending success; the same three yards on third-and-twelve is a defeat. Expected Points Added — EPA — converts every play onto a common scale by asking how much the play changed the offense's expected next-score value.
The setup is empirical. Years of historical play-by-play data are binned by down, distance, field position, and time. For each bin, you compute the average value of the next score in the half — positive for an offensive touchdown, negative for a defensive score, zero on a half-ending punt. That gives you the expected points at every state. EPA on a given play is the difference between the expected points after the play and the expected points before it.
A 12-yard completion on 3rd-and-10 from your own 30 might add about 1.5 EPA; the same 12 yards on 3rd-and-15 from your 30 might add 0.3, because you're still likely to punt. A successful two-point conversion can add over 1.5 EPA; a stuffed two-point try can subtract more than 1.5. Sacks, turnovers, and explosive plays move the number the most.
EPA per play has displaced yards per play as the standard team efficiency metric in football analysis. The reason is straightforward: EPA accounts for situation and is stable enough at the team level to be predictive across games. Offensive EPA per play in the first half of a season correlates with second-half offensive output far better than yards per play does. Most modern football models — including most of what NFL teams use internally — are built around EPA in some form.
Our team strength priors for NFL and CFB are EPA-based. Offensive and defensive EPA per play, adjusted for opponent and pace, are the core inputs to our spread and total models.