Football Analytics
Expected Points Added (EPA)
The foundation of modern football analytics. EPA measures the real value of every play by asking one question: how much did this play change the offense's expected points?
What Is EPA?
Every down, distance, and field position in football has an expected point value. A 1st and 10 at your own 20 is worth about 0.4 expected points. A 1st and 10 at midfield is worth about 2.0. These values come from analyzing millions of NFL plays and tracking what happened next. On average, an offense with a 1st and 10 at their own 20 will eventually score about 0.4 points on that drive. The expected points model assigns a number to every game state.
EPA is the difference between the expected points before a play and the expected points after it. If an offense starts a play in a situation worth 1.5 expected points and ends in a situation worth 3.0, that play generated 1.5 EPA. A turnover might drop the expected points from 2.0 to negative 4.0, resulting in negative 6.0 EPA. Every play moves the needle, and EPA measures exactly how much.
How EPA Is Calculated
The foundation of EPA is the expected points model. Researchers build this model by looking at every combination of down, distance, and field position from years of NFL data and calculating the average points scored from each situation. A 1st and goal from the 1 is worth close to 6 expected points. A 3rd and 20 from your own 5 is worth almost nothing. The model creates a smooth curve across every possible game state.
Once you have the expected points for the pre-snap situation and the post-snap situation, EPA is just subtraction. A completed 15-yard pass on 2nd and 8 from the 40-yard line might move you from 2.5 expected points to 4.1 expected points, giving that play an EPA of 1.6. A sack on the same play might drop you to 1.1 expected points, giving the play an EPA of negative 1.4. The math is simple. The power comes from the context it provides.
Modern EPA models also account for score differential and time remaining, especially late in games when a team that is trailing will play differently than a team with a lead. Some models adjust for era, dome versus outdoor settings, and other variables. But the core logic is always the same: measure the change in expected points from one snap to the next.
Why EPA Matters More Than Yards
Yards treat every gain the same regardless of when it happens. A 4-yard run on 3rd and 3 converts a first down and keeps the drive alive. A 4-yard run on 3rd and 15 is effectively meaningless. Yards per play would rate both identically. EPA captures the enormous difference in value between those two plays because it accounts for down, distance, and what happens next.
This is why EPA has become the standard for evaluating quarterbacks, play callers, and offensive systems. A quarterback who consistently converts on third down and moves the chains will show up much better in EPA per dropback than one who racks up yards in garbage time or on plays that do not actually improve the team's scoring position. EPA tells you who is actually generating points for their offense, not just who is generating volume.
On the defensive side, EPA per play allowed is one of the best measures of a defense's true effectiveness. A defense that gives up yards but prevents points is doing its job. A defense that gives up short fields after turnovers and fails on third down will have a terrible EPA per play allowed even if the total yardage numbers look respectable. EPA strips away the noise and focuses on what actually decides games: points.
EPA per Play
The average expected points added on each snap. This is the baseline efficiency metric used to compare quarterbacks, play callers, and entire offenses.
Cumulative EPA
Total expected points added across a game or season. Shows the full picture of a player's point-scoring contribution over time.
EPA vs Yards
A 5-yard gain on 3rd and 3 is worth far more than a 5-yard gain on 1st and 10. EPA captures that context. Raw yards do not.
Context changes everything.
EPA captures what yards, completion percentage, and box score stats all miss. The value of a play depends entirely on when and where it happens.