Welcome back to the final installment of Advanced Stats for Packers Fans. Today, I’m going to be providing an overview of the most prevalent advanced stat found in discussions around the league today: Expected Points Added, or EPA. 

EPA is based on a simple idea that we’ve already covered a number of times through this series: not all yards gained are equal. When we discussed success rate in part one, yards gained was compared against a simple pass/fail formula, based on how many yards you were gaining towards a first down (or touchdown). In part two, we found that DVOA compared the yards a team gained in a play, compared it to a league average and adjusted for a number of variables, and then returned a quantifiable number. EPA operates in a similar way. EPA finds its foundation in a very simple concept inherently known by all football fans: it’s better to have the ball closer to the endzone than further from the endzone. Hear that? That’s the sound of hundreds of Cheesehead TV readers all rolling their eyes at once. Stick with me though.

Let’s begin with the concept of Expected Points (EP). In 1970, the starting quarterback of the Cincinnati Bengals was Virgil Carter, and he wrote a research paper with a professor at Northwestern, Robert Machol. Together the two quantified that intuitive concept, by adding together the point value and probability of all potential outcomes of a possession. What they found was that, of course, as teams approach the opponent’s endzone, the probability of a touchdown or field goal increases. For example, they found that having the ball at the 50 yard line meant the probability of scoring a touchdown was 30%. Because a touchdown is seven points, the EP in that situation is 2.1 (0.3 * 7). Adding together this value for all potential outcomes yields the total Expected Points. 

Put simply, Expected Points describes how many points a team is expected to score, given their field position. Modern EP probability has grown to include more context into this equation, such as down and distance, and time remaining in the half. 

Notice how part of the EP value line dips below 0? This means that, based on all probable factors, the opposing team is actually more likely to score the next points, despite not having the ball. This creates an easy to understand difference between positive and negative EP, thus allowing users of the formula to easily understand whether their team is in an advantageous position or not. This also means that you can have the same EP from different spots on the field, based on down and distance. For example, 1st and 10 from your own 30 yard line has the same EP as a 3rd and 10 from midfield. 

When comparing the EP chart from the same spot on the field but with different downs, you can see that a 1st and 10 will usually have the same EP as a 2nd and 2, meaning that according to the metric, a team needs about eight yards per play in order to maintain the same EP throughout the drive. So we can now understand that there is a difference in a team’s EP during the drive, and that your EP changes as the drive progresses. This is what the A in EPA is tracking: Expected Points Added.

Consider a twenty yard pass, from a team’s own 20 yard line to their 40 yard line. Before the play, the team had an EP of 0.7, and after the play their new EP was 2.0. This means that the 20 yard pass had an EPA of +1.30. If a team were to gain no yards, or even lose yards, the EPA would be negative. So in essence, while points are obviously not scored on every single play, EPA is capable of attributing points to every play. 

Say a team drive starts on the opponent’s 45 yard line. There’s a run of six yards, a pass of 30 yards, and then a pass of nine yards for a touchdown. The run has a small EPA, the explosive pass has a large EPA, and the touchdown pass has a respectable EPA. While all of the plays together led to the touchdown, EPA allows a quantification of how helpful each individual play was along the way. Here’s an important idea to keep in mind, however: when charting the total EP gained over that drive, the total might not necessarily add up to 7. Because that drive began on the opponent’s side of the field, the offense is already starting with a positive EP. So our example team already started with an EP of, say, 2.8, and gained 4.2 along the way to their seven real life points. 

Again, while there is a simple formula for following along with a team’s EP during a drive, each play’s EPA is affected by a multitude of factors. This includes whether or not the play was a run or a pass, play type (play action, bootleg, running back draw, ect.) It’s by combining this play classification with point attribution, that the modern NFL analytics community derives a majority of its data. 

Remember the big swing towards passing heavy offenses in the 2010s? A big part of that was the quantification of running plays vs. passing plays, where according to EPA an average passing play will add 0.03 EPA, while an average running play will yield -0.05 EPA. So why would you ever run the ball in the first place? This is the first major drawback of EPA (in my opinion). EPA cannot quantify the intangible benefits for certain actions on the field. Running the ball will benefit the offense in ways like running the clock out, setting up the passing game, and tiring out a defensive line, but EPA will only care if it contributed to its goal: points scored. 

Similarly to DVOA, EPA is also tracked on the defensive side of the ball, by simply tracking the inverse of the play’s result. +0.50 EPA for the offense would be -0.50 for the defense. EPA is also tracked at an individual level, though only among quarterbacks. This is only possible because of the large and obvious impact that the quarterback has on the game, and attempting to track any other individual’s effort in the ultimate team game would be ultimately futile. It’s still a useful tool for analyzing QBs, most popularly used in the form of EPA per dropback, which could be viewed as a superior form of passer rating or yards per play. QBs are also commonly tracked according to stats like EPA lost to sacks, turnovers and penalties. 

If EPA sounds suspiciously similar to DVOA, that's because they are. There are a lot of similarities to the way that EPA and DVOA are calculated, but the big difference between the two comes from what exactly is being tracked. Remember, DVOA is an efficiency metric, that compares a team’s efficiency to a league average based on down and distance. EPA is a metric that tracks a team’s probability of scoring. Both metrics will take similar variables into account, such as down and distance, play type, and time remaining on the clock, but they are taken into account for different reasons, and only DVOA is able to take an opponent's strength into account. DVOA is extremely flexible and dynamic, while EPA follows a relatively simple progression of events and probabilities.  

It’s not perfect however, and as I’ve been doing throughout this series, I will stress again the importance of not viewing these advanced stats in a vacuum, but instead using them in tandem with other advanced stats. Success rate and EPA are intrinsically linked, and you might even see versions of success rate that assign a pass/fail metric based simply on whether or not a play had a positive or negative EPA. Further shortcomings include EPA’s inability to capture information outside of its dataset, such as scheme or intention, but also attributing the smaller details of a play. If a deep pass downfield picks up a chunk of EPA, was that because of a well thrown ball or a well run route? EPA couldn’t tell you. 

For fans wishing to find EPA data for themselves, I highly recommend NFELO, for highly accurate, easy to understand metrics. During the season I will check their power rankings and EPA tier charts about once a week, but they also have the best info for QB EPA, team tendencies, and a W.I.P wide receiver EPA metric. NFELO also does a great job in separating EPA into categories such as passing EPA vs rushing EPA. 

Let’s take a quick look at where the Packers fit in in all this at the end of the 2024 season. Green Bay was firmly a top ten team when viewed through the lens of EPA, with the ninth highest offensive EPA, and eighth in both passing and rushing, while defensively the team was fifth in total defensive EPA, fifth in passing,and seventh in rushing. Their position in the EPA tier chart I linked above also puts them right around the highest performing teams in the league. 

There is an interesting nugget here when looking at their Sack EPA (second) vs Turnover EPA (sixteenth), where most fans probably would have guessed the team was a little better in the turnover margin than the sack. The opposite is true however, and it seems, as far as EPA is concerned, the Packers got more value out of creating and avoiding sacks than they did in creating and avoiding turnovers.

As I wrap up this article, and the series, I want to say again that the object of these posts was not to convince you, dear reader, to pour your heart and soul into these metrics, and trust them totally. I advocate for these metrics to be a tool that one could use to understand what you are watching on the field. Instead, perhaps the most important thing we can take away is the advantage these advanced stats have over more “conventional” stats: the dynamic way they can include the broader context of what happens during the ultimate team game. 

I’d like to thank everyone for following along with my series on Advanced Stats for Packers Fans 101, and I hope that everyone reading along got some valuable information out of it along the way!