I'm not familiar with EPA so I did a semi-deep dive on it.
It seems actually like a pretty good metric (it doesn't seem to be ESPN's, for the record, but rather a pretty respectable stat).
And, at least according to my Grok research, I think this has been pretty widely misinterpreted in regards to what it says about our performance with and without Parsons, including by me.
Apparently a negative defensive EPA is what you want. So in other words, the defense was actually much better without Parsons on the field than with him.
Which is surprising because even guys like Sturm are repeating this, and Sturm seems to know what he's talking about.
I don't know who Bill Barnwell is but he seems to be connected with ESPN in some way (I don't watch ESPN at all). This is the tweet Sturm is quoting from Barnwell...
So I asked Grok about this post and it said:
"The post highlights Micah Parsons' critical impact on the Dallas Cowboys' defense, showing a stark contrast in EPA per play (a metric measuring a team's performance against expectations) from +0.12 when he plays to -0.18 when absent over 1,039 plays, suggesting his presence adds roughly 0.3 points per play, a difference backed by EPA's reliability in player evaluation as noted in Best Ball Stats' 2024 analysis."
Then I asked what EPA is, to which it replied:
"EPA per play, or Expected Points Added per play, is a key metric in football analytics that measures how much a team or player contributes to the expected points outcome of a game on a per-play basis. It’s derived from the concept of Expected Points (EP), which estimates the average number of points a team is likely to score on a given drive based on factors like down, distance, field position, time remaining, and game situation (e.g., home/away, score differential). EPA then calculates the change in those expected points resulting from a specific play.
How It Works:
Expected Points (EP): Before each play, a statistical model (built from historical NFL data) assigns an EP value to the current situation. For example, starting a drive at the 50-yard line might yield an EP of about 2.5 points, while a 3rd-and-10 at your own 20-yard line might be closer to 0.1 points.
EPA Calculation: After the play, the EP is recalculated based on the new situation (e.g., a 10-yard gain might increase EP from 2.5 to 3.5). The difference (3.5 - 2.5 = 1.0 EPA) reflects the play’s contribution to the team’s scoring potential.
Per Play Normalization: To account for varying play counts, EPA is divided by the number of plays to get EPA per play. This normalizes the data, making it easier to compare teams or players across games or seasons. For instance, a team with 10 EPA over 50 plays has an EPA per play of 0.2.
Application to Defense: For defenses, EPA per play is negative when they perform well, as their goal is to reduce the opponent’s expected points. A lower (more negative) defensive EPA per play indicates better performance. In the context of Bill Barnwell’s post about Micah Parsons, the Cowboys’ defense shows a +0.12 EPA per play with him (indicating they allow more points than expected, a poor defensive mark) and a -0.18 EPA per play without him (indicating they prevent points better), highlighting his disruptive impact."
In other words, I think a lot of us got this stat backwards, including myself and Sturm and Barnwell.
I asked Grok to clarify and got in part:
"Conclusion: The Cowboys performed better without Micah Parsons, as measured by EPA per play. The -0.18 EPA per play without him is a stronger defensive metric than the +0.12 EPA per play with him, showing a difference of about 0.3 points per play in favor of the defense when he’s not playing. This counterintuitive result might reflect factors like opposing teams adjusting strategies against Parsons, his occasional injuries (e.g., the 2024 absence noted in DallasNews.com), or the overall team’s reliance on him leading to breakdowns when he’s targeted.
Caveats:The 1,039-play sample without Parsons is significant, but the with/without split might include different opponents or game contexts (e.g., blowouts vs. close games).
This doesn’t diminish Parsons’ individual impact (e.g., sacks, pressures); it suggests the team’s defensive scheme or personnel struggles to complement him effectively.
So, yes, the data indicates the Cowboys' defense statistically performed worse with Parsons on the field over this period, which is a surprising twist worth digging into further!"
As a second opinion, I asked Google's AI which said:
"a negative EPA per play is good for a defense because it indicates that the defense has reduced the offense's expected points, limiting their ability to score and moving them further from a scoring opportunity. A positive EPA for the defense means the offense is performing better than average, which is a negative outcome for the defense."
Then I asked Google's AI what the team's EPA per play was with and without Parsons, and it contradicted itself:
"When Micah Parsons was on the field from 2021-2024, the Dallas Cowboys had the NFL's best defense by EPA per play, with a 0.08 EPA/play. However, when he was off the field during that same period, the Cowboys defense ranked dead last in the NFL by the same metric, with a -0.05 EPA/play. The difference in EPA per play was a significant 0.13 points (0.08 - (-0.05) = 0.13), highlighting Parsons' invaluable impact on the team's defensive performance."
So I went back to Grok and it said:
"Is Google Wrong?Yes, Google’s AI appears to be wrong in labeling 0.08 EPA/play as the “NFL’s best defense.” That value reflects a below-average to poor defense, and the narrative doesn’t match the metric. Barnwell’s post also seems inconsistent internally, likely due to a miscommunication of “best” with a positive EPA. The correct reading, aligning with your insight, is that negative EPA is better for defenses, and the data shows the Cowboys’ defense statistically improved without Parsons."
So based on my current understanding of this, aided by Grok, (and maybe I'm wrong because I still only have a rudimentary understanding of it) it seems like they were significantly better without Parsons, and a lot of people seem to be misinterpreting it.
I asked myself but only got this mess which is simply citing ESPN, CBS Sports, Facebook 
, and Reddit. 
