To stat geeks like me, there are three little letters that represent the most explanatory, most predictive, and most fascinating stat in all of sports: EPA. The Environmental Protection Agency might have a stronghold on the acronym among the general public, but in the world of sports analytics, EPA is all about expected points added.
With most of the stats I discuss, the aim is typically to uncover how closely perception matches reality—how easily things could have gone differently. I want to find out how a team will perform in the future, emphasizing the process of how they’re doing things instead of the results.
In terms of future points, past points is only a moderately good predictor. Let’s face it: points are relatively fluky. We see offenses score because their defense give them the ball in opponent territory or fail to score after driving 80 yards all the time. So we want to know how well offenses (and defenses) are really performing, and how many points they should be scoring relative to their performance.
EPA captures that. Using historic game data, sites like Advanced NFL Stats can very accurately predict the number of points a team can be expected to score on any given drive. A first-and-10 at your own 20-yard line has historically been worth a net of +0.34 points to an offense, i.e. they could be expected to outscore the opponent by 0.34 points, given that particular game situation. An average offense in such a situation would have a 15 percent chance of scoring a touchdown and an eight percent chance of connecting on a field goal.
So if we know the point expectation before a play and we know it after a play, we can measure the difference to see how well a team is playing. That’s EPA.
The Cowboys currently rank fifth in the NFL in points with 83. That’s awesome, but it doesn’t tell us much about the Cowboys’ offense because the play of a team’s offense, defense, and special teams are so connected. They’re tied together in such a way that we could never separate them studying points scored or allowed. For years, Dallas has ranked really high in yards but really low in points because their defense failed to put them in advantageous situations. Points scored is just a really poor barometer for offensive strength.
But EPA adjusts for that. Think about what happens if your defense forces a fumble and gives the offense the ball at the opponent’s one-yard line. The offense will likely punch it in for a score, adding seven points to their total just the same as if they drove 99 yards. That doesn’t tell us much.
In terms of EPA, though, the jump would be minimal. At the opponent’s one-yard line, the offense’s expectation would already be nearly seven points. The EPA would increase when the offense scores a touchdown, but not by much. EPA accounts for game situations in a way that traditional stats cannot, and thus it does a much better job of isolating offensive play. It leads to better data, better insights, and most important, better predictions.
The Cowboys, Chargers, and Offensive EPA
Like I said, the ‘Boys rank fifth in points. But the takeaways they got against the Giants in Week 1 have really skewed that rank to suggest the offense is playing better than it is. If we look at EPA to account for game situations, the Cowboys rank 15th in the NFL with +1.71 EPA (which basically means they’ve been just slightly better than average, which would be a net EPA of zero). The Broncos lead the NFL in EPA by a wide margin at +54.2, while the Jaguars are last, also by a wide margin, at -55.2.
The fact that Dallas ranks in the middle of the pack in EPA suggests that, as the randomness of defensive takeaways reverts back to the mean, they’ll be unlikely to continue scoring 27.7 points per game. Offensively, they just aren’t playing well enough to post that sort of number. They’re playing closer to the league average of 22.6 points per game.
Meanwhile, San Diego has seen just the opposite effect. They currently sit at ninth in scoring with 26 points per game. But they rank fourth in the NFL with +29.5 EPA. Based on that total, their most likely output has been much closer to 30 points per game, i.e. the offense has played well enough to score a whole lot of points, but the defense hasn’t been as great.
There’s lots of data that goes into any prediction, and both teams’ EPA has certainly been influenced by their opponents. But if we want to predict which team will score more points moving forward and all of the information we have is EPA and points scored, we’d be justified in picking San Diego to score more points, even though they’ve actually scored fewer than Dallas thus far. That’s the power of EPA.
- This quickly turned into a rant on EPA—hopefully a useful one—so I’ll leave you with a few quick stats for both squads.
- Philip Rivers has attempted fewer play-action passes than anyone, doing so on 12.1 percent of his dropbacks. Tony Romo ranks 28th at 14.8 percent.
- Romo’s play-action passer rating is 110.2. Rivers’ is 152.6. The San Diego quarterback is 10-for-13 for 146 yards, two touchdowns, and no picks. But yeah, both teams should continue to ignore the numbers and run play-action at bottom-dwelling rates, right?
- Rivers has been pressured on just 28.0 percent of his dropbacks—a top five rate—while Romo is just above him at 29.5 percent. Certainly the offensive lines have played better, but the real reason for the reduced pressure is the quarterbacks. There’s a reason that Peyton Manning and Tom Brady always have “elite” offensive lines that don’t allow much pressure; they get rid of the football in a hurry.
- In 2013, both Rivers and Romo have thrown the ball in 2.46 seconds or less, on average, according to Pro Football Focus. Both rank in the top six. Brady checks in at 2.51 seconds and Manning at 2.38. So the majority of the praise for the Cowboys’ offensive line through three games should really go to Romo for getting the ball out quickly.