Understanding EPA/Play
Why traditional yardage stats are obsolete and how Expected Points Added changed the NFL.
If a running back gains 4 yards on 3rd & 5, he failed. If he gains 4 yards on 1st & 10, he succeeded. Traditional stats count both as "4 rushing yards." EPA (Expected Points Added) fixes this critical flaw by applying context.
The Baseline: Expected Points (EP)
To understand EPA, you must first understand Expected Points. Based on historical data (usually cited from nflfastR databases dating back to 1999), every combination of down, distance, and field position has a mathematical expected point value for the next scoring event.
Worked Example (Data: 2023 Season)
1st & 10 at your own 20-yard line has an EP of roughly +0.5 points. If the offense throws an incomplete pass, it becomes 2nd & 10 at the 20. The EP drops to -0.1 points.
EPA = New EP (-0.1) - Old EP (0.5) = -0.6 EPA
Common Mistakes in EPA Usage
- Ignoring Garbage Time: EPA must be filtered for win probability (usually WP between 20-80%) to avoid skewing data from blowout scenarios.
- Attributing EPA solely to the QB: While QBs have the highest correlation with EPA, offensive line play and receiver separation heavily influence the metric.
- Small Sample Sizes: Evaluating a single game's EPA variance can be highly deceptive due to massive swings from turnovers.
FAQ
- What is considered a "good" EPA/Play for a QB?
- In 2023, the elite threshold was anything above 0.200 EPA/play. Brock Purdy led the league at 0.338.
- Why is rushing EPA almost always negative?
- Passing is inherently more efficient than rushing in the modern NFL. An average pass play yields positive EPA, while an average run play yields negative EPA, making the decision to run heavily scrutinized by analytics departments.