What is EPA in the NFL? All about Expected Points Added
Expected points in the NFL is the amount of points a team is expected to score on the current drive. Expected points added or EPA in football measures how a team’s expected points (EP) changes on an individual play.
The EPA formula in football is:
To figure out the expected points added, you look at the difference between the expected points before a play and after a play.
Expected points added measures how well a team performs relative to expectation. Positive values mean a team increased their expected points on a play; it was a good play. Negative EPA means a play was detrimental to a team’s winning chances.
The plot below is a histogram of the expected points added across all plays in the 2022-2023 NFL season.
In the rest of this article we’ll take a deeper dive into this stat to help understand what it actually means.
Expected Points in the NFL
Everything is predicated upon the idea of expected points. Expected points is the average points scored on all drives that have the same down, distance, and field position.
For example, a 1st and goal from the 1 yard line has an expected points of nearly 7 – a touchdown is almost guaranteed! On the other hand, a 4th and 10 from the opponents’ 20 yard line will have an expected points of about 3 – almost all of these scenarios end up with made field goals.
In the plot below we show the expected points based on down and field position (to make the plot we averaged over “yards to go”).
A lot of the information in this chart is “obvious”. Being closer to the endzone leads to higher expected points. Earlier downs lead to more expected points. Expected points range from nearly 6.5 in extraordinarily favorable positions to actually negative values in unfavorable positions.
NFL EPA
EPA is all about transitions. If a play results in our expected points being increased, then that was a good play. If a play results in our expected points decreasing, it was a bad play.
EPA assigns values to individual plays based on how much expected points changes. Here is an example. A 2nd and 10 form roughly a team’s own 45 yard line has an expected points of about 2.0. Suppose our team completes a pass for 40 yards down to the opponents’ 25. This 1st and 10 from the new field position has an expected points of 4.9. The EPA for this play was +2.9
There are many ways to interpret this stat. My favorite: if we go on to score a touchdown, then a full 2.9 of those eventual 7 points are due to this one play.
So far we’ve described how EPA is calculated, but not why it is useful. In the next section we’ll look at the why behind this stat.
Why Use EPA?
In the NFL, points scored aren’t the result of just one play. They are the result of long drives, along the way getting closer and closer to scoring. A running back punching it in from the 1 yard line shouldn’t get credit for all 6 points for the TD, he just finished the drive. A kicker hitting a 25 yard field owes some credit to his team for giving him the easy kick.
EPA lets us break down scoring drives into individual play contributions. The 7 points from a touchdown happen gradually over the course of a drive. Using EPA lets us figure out which plays were the most impactful on a drive.
Involved Players and EPA
By looking at which players were involved on which plays, we can also get a guess at the value contributed by individual players. If a receiver has a big play but gets tackled at the 1 yard line, everyone knows he really gets credit for the inevitable TD to come. EPA will recognize this fact while the boxscore stats won’t
Even more interestingly, EPA can help us value players with different styles. Who is more important to a team: a big play receiver or a consistent, short yardage guy? The big play receiver gets all the highlights. But maybe the short yardage guy contributes more by keeping drives alive. EPA can tell us who actually is more important.
Similarly, it can be hard to compare receiving (AKA 3rd down) running backs to workhorse 1st and 2nd down running backs. By comparing the EPA on 1st and 2nd down versus third down, we might be able to get a sense on which back is better.
Overall, expected points added is a very valuable tool to assign point values to individual plays.
Expected Points Added and RE24
Expected points added is one a modern NFL statistic, one of many that you might call “NFL sabermetrics”. The funny thing is, though, that EPA in the NFL is almost identical to baseball’s RE24 and the run expectancy matrix.
The goal of this football stat is to figure out the incremental value of individual plays. It does this by measuring the expected points scored before and after a play. The difference is the expected points added or subtracted due to the outcome of the play in question. Computing EPA relies on estimating expected points given field position, down, and yards to go.
In baseball, the stat RE24 measures the amount of runs added or subtracted via a player’s at-bat. To do this, RE24 computes the expected number of runs in an inning both before and after an at bat. The difference between these two values is a measure of a player’s contributions. Computing RE24 relies on the run expectancy matrix.
Recognizing this relationship between EPA in football and RE24 leads to some interesting thoughts.
Expected Points Added and Context
One of the great debates in sports relies on whether context matters. Is an RBI double more impressive than a regular double that doesn’t score a run? Is a 3 pointer to take a late lead (“in the clutch”) actually that much more impressive than the exact same shot made in the second quarter?
Many people – myself included – think that baseball stats should be measured context free. That is, a double is a double regardless of who is on base, how many outs there are, and what the score is. For this reason, I prefer context-independent stats like wRAA or WOBA over their context-dependent counterparts like RE24.
Just like RE24 in baseball, EPA in football depends on context. A 1 yard run on 4th down from the 1 yard line is worth more than a 1 yard run on 1st and 10 from your own 20. However, unlike in baseball, I actually think that these outcomes should be worth different amounts.
In the NFL, teams play differently based on context. In baseball, a pitch is (by and large) the same as every other pitch. Continuing the example from above, yards are harder to gain on 4th and 1 from the 1 because all 11 defenders will be right on or near the line of scrimmage. On 1st and 10, the defensive backs have to cover potential long pass plays as well as runs, making their job harder.
Another example: on 3rd and 15+, it is not uncommon for the defense to give up easy 7-10 yard plays in order to escape to fourth down. This means that a gain of 10 yards in this setting is much less impressive than a gain of 10 on 1st and 10.
Because of this, I don’t think there is much value to context-independent EPA stats in the NFL. However, I do think there is one way to improve how we use EPA.
Mean v. Median
Most times, EPA is presented as “average EPA per play”. Sometimes this is broken down to the offensive or defensive side of the ball. This is all a way of measuring mean EPA.
However, mean EPA is by-and-large the same thing as margin of victory. If you add up the EPA of a drive, it will be very, very close to the points scored on that drive (the only difference is taking into account relative starting field position). If you add up and take the mean of EPA over the course of a game, you will just get the margin of victory.
I think there is an opportunity to get more value out of the stat by looking at median EPA as well as other quantiles. Sometimes you’ll be watching a game and “feel” like one team is dominating, only for a small handful of plays to swing the game. A pick 6, a stop deep in the redzone, a missed FG, etc.
If we can somehow take into account the overall game flow and diminish the impacts of big swing plays, we might get a better sense of how good or bad a team is. This is something that we’re going to explore in the future.
To receive email updates when new articles are posted, use the subscription form below!
