EPA in Football Explained: The Metric That Redefined the NFL
Pickviz
6 min read
Traditional football statistics tell incomplete stories. A quarterback throwing for 300 yards looks dominant in the box score, but if most of those yards came in garbage time or on futile 3rd-and-long checkdowns, they did little to help his team win. In modern NFL analytics, coaches, front offices, and sharp bettors evaluate performance using Expected Points Added (EPA).
By evaluating the value of every play relative to the game state, EPA strips away the noise of raw yardage. Much like Expected Goals (xG) in soccer or Strokes Gained in golf, EPA provides the baseline context that defines efficiency in a high-variance sport.
The Mathematical Foundation: Expected Points (EP)
To understand Expected Points Added, you must first understand the foundation: Expected Points (EP).
Using historical data from thousands of NFL games, analysts calculated the average net point value of every possible game state. A game state is defined by three primary variables:
* Field Position: Distance to the opponent’s goal line.
* Down: 1st, 2nd, 3rd, or 4th down.
* Distance: Yards needed for a first down.
At any given moment, a team has an Expected Points value. If a team has 1st-and-Goal on the opponent’s 1-yard line, their EP is high (roughly +6.0 points) because they are highly likely to score a touchdown on that drive. Conversely, if a team has 3rd-and-15 on their own 2-yard line, their EP is negative (roughly -1.5 points) because they are statistically more likely to concede points (via a safety, interception, or punt return) than to score.
Expected Points Added (EPA) is simply the change in Expected Points from the start of a play to the end of a play:
$$\text{EPA} = \text{EP}{\text{after play}} – \text{EP}$$}
A Real-World Example
Imagine the Kansas City Chiefs have a drive starting at their own 20-yard line.
* Start State: 1st-and-10 at the KC 20. The historical database shows that a team in this position is expected to score +0.55 points on average.
* Play A (10-yard gain): Patrick Mahomes completes a pass to the 30-yard line.
* End State: 1st-and-10 at the KC 30. The expected points value for this new state is +1.10 points.
* EPA Calculation:
$$\text{EPA} = 1.10 – 0.55 = +0.55$$
The completion added over half a point of value to the drive.
Now consider a different play from the same start state:
* Play B (2-yard gain): The running back carries the ball for 2 yards.
* End State: 2nd-and-8 at the KC 22. The expected points value drops to +0.42 points because 2nd-and-8 is statistically harder to convert than 1st-and-10.
* EPA Calculation:
$$\text{EPA} = 0.42 – 0.55 = -0.13$$
Even though the play gained positive yardage, it resulted in a negative EPA because the team fell behind the chains.
EPA vs. Success Rate: The Crucial Companion Metric
While EPA is the best metric for measuring the overall impact of offensive play, it has one major vulnerability: it is highly sensitive to outlier plays. A single 75-yard touchdown pass or a catastrophic 90-yard pick-six can heavily skew a team’s total EPA for a game.
To balance this, analysts pair EPA with Success Rate.
Success Rate is a binary metric (Yes/No) that tracks consistency. A play is deemed “successful” if it meets the following yardage thresholds:
1. On 1st Down: Gains at least 40% of the yards needed for a first down.
2. On 2nd Down: Gains at least 50% of the yards needed for a first down.
3. On 3rd and 4th Down: Converts the first down.
EPA = Magnitude of Value (Explosive Plays vs. Turnovers)
Success Rate = Consistency of Performance (Staying Ahead of the Chains)
By comparing these two metrics, you can identify the exact offensive identity of a team. A team with a high EPA but a low Success Rate is a “boom-or-bust” offense that relies on deep explosive plays but struggles to sustain long drives. A team with a high Success Rate but a low EPA is a conservative, methodical offense that moves the chains but lacks big-play capability.
| Offense Profile | EPA | Success Rate | Analytical Outlook |
|---|---|---|---|
| Boom-or-Bust | High | Low | Volatile, high reliance on explosive plays; prone to turnovers. |
| Methodical | Low | High | Consistent, low turnover rate; struggles to score quickly. |
| Elite | High | High | Dominant; stays ahead of the chains and creates big plays. |
| Poor | Low | Low | Inefficient; struggles to move the ball or generate big plays. |
Source: Advanced NFL play-by-play analytics baselines, verified against standard EPA models as of July 2026.
Passing EPA vs. Rushing EPA: The Analytical Verdict
One of the most significant discoveries of modern football analytics is the extreme discrepancy between passing and rushing efficiency.
Across any NFL season, passing plays generate a positive average EPA (typically around +0.05 to +0.10 EPA per play), while rushing plays generate a near-neutral or negative average EPA (typically around -0.05 to 0.00 EPA per play).
This gap exists because passing gains significantly more yards per attempt and is more likely to generate explosive plays. Similar to how sabermetrics in baseball proved that avoiding outs is more valuable than stealing bases, football analytics proved that passing is the primary engine of scoring.
Bettor’s Note: Teams that run the ball frequently are not necessarily efficient; they are often running because they are already winning and want to burn the clock. This is a classic case of endogeneity. To evaluate an offense’s true capability, always strip out garbage-time plays and focus on early-down passing EPA in neutral game states (when the win probability is between 20% and 80%).
How to Build a Betting Model Using EPA
For sports bettors, EPA is the foundational input for creating power ratings and projecting game totals.
1. Focus on Passing EPA and Passing Def-EPA
Because passing is highly correlated with winning, your offensive power ratings should weight Passing EPA significantly higher than Rushing EPA. When projecting matchups, compare an offense’s Passing EPA against the opposing defense’s Passing EPA allowed. This single matchup is the most predictive element of game outcomes.
2. Identify Regression Candidates using Success Rate
If a team is winning games with a high EPA but a bottom-tier Success Rate, they are prime candidates for regression. They are relying on unsustainable big-play variance and are likely to struggle when they face disciplined defenses that eliminate explosive plays. Backing their opponents on the point spread is a high-value angle.
3. Adjusting Totals for Matchup Pace and Efficiency
Matchups between two high-EPA, high-Success Rate offenses will consistently exceed game totals set by sportsbooks, especially if both defenses have poor pass-defense EPA. Conversely, matchups between high-Success Rate, low-EPA offenses will trend toward the under, as they consume clock without generating quick scores.
Frequently Asked Questions
What is a good EPA per play?
In the NFL, an individual quarterback averaging +0.15 EPA per play over a season is playing at a Pro Bowl or MVP level. For team offenses, an average of +0.10 EPA per play is elite, while anything below -0.05 is poor.
Does EPA account for garbage time?
Standard raw EPA includes all plays, but advanced analytics sites (like Ben Baldwin’s rbsdm or nflfastR) allow analysts to filter out “garbage time”—typically defined as plays where a team has a win probability greater than 96% or less than 4%. This ensures the data only reflects competitive play.
How does a turnover impact EPA?
Turnovers are the single most damaging event in the EPA model. An interception on 1st-and-10 near the opponent’s red zone can easily result in an EPA of -4.5 or lower, as it completely strips away a team’s high scoring probability and hands it to the opponent.
