NFL Penalties 2023 – The Cost of Penalties in Expected Points Added

As a follow up to the NFL 2023 – Referees & Penalties – Midseason Review, I did a little google research and found an interesting paper by Jack Lichtenstein on the value of NFL Penalties in Expected Points Added.  The good news is he analyzed all the penalties from 1999 – 2020 as the dataset.  Then he generated the EPA gained or lost by the penalties.  The bad news. only the top 16 penalties by amount called were analyzed and graphed.  Here is a link to the Research: Exploring Stability and Predictive Power of Penalties in the NFL

The Assumptions & Estimations of Penalty EPA

This left out two of the more recent, and frequent penalties of defensive holding and illegal contact.  Both of these penalties are the automatic 1st down passing penalties we covered in the first article.  So here is the big assumption we are going to make for this article:

The EPA value for defnsive holding and illegal contact will be set equal to the EPA value for defensive illegal contact.  The semantics of these three calls can be argued but they are used fairly interchangeably and at times suspectly by NFL officials so I dont think we are going to far over the edge here.

All 3 penalties are 5 yard, automatic first down penalties, so their value should be about even, the only difference being what the penalty is deemed and how far it was to go for the 1st down that was now mitigated.  If by some chance, illegal use of hands has a completely different value than the real values for defensive holding and illegal contact, then much of what follows would be nonsense.  I think the chances of that are extremely low so I am going to continue albeit with the warning and transparency noted.

I also had to make some assumtions on the averages of each of the shown ranges.  And to keep thing fairly simple and avoid miniscule arguments, I am limiting EPA to 0.25 increments for penalties.  There will be no penalties with values like 1.47 or -2.83.  So here is the list of values I came up with for each type of penalty.

Offensive Penalties EPA Lost

Offense Unsportmanlike Conduct = 0
False Start = -0.75 EPA
Intentional Grounding = -1
Ineligible man downfield = -1
Delay of game Offense = -1
Offense Illegal Block = -1.25
Offense Illegal Use of Hands = -1.5
Offense Unecessary Roughness = -1.75
Offense Personal Foul = -1.75
Illegal Shift = -2
Offensive P.I. = -2.5
Offensive Holding = -2.5

Defensive Penalties EPA Gained

Defense Unsportmanlike Conduct = 1
Defense Illegal Block = 1.25
Defense Offsides = 1.5
Delay defense(12 men) = 1.75
Defense Personal Foul = 2
Defensive Holding = 2
Defensive Illegal Contact = 2
Defense Illegal Use of Hands = 2
Defense Unecessary Roughness = 2.25
Roughing the Passer = 3
Defensive Pass Interference = 3.5

The EPA Oddities

The Gorilla in the corner is the zero value for Offensive Unsportsmanlike conduct.  Since I have no statistics to back it up, I will give you my best guess.  Most times when this is called, it is called as offsetting if it involves an offensive call.  There will be times only an offensive player is called(rare), and only a defensive player is called(common), and this imbalance and offsets lead to a net near zero EPA even though it is supposedly a 15 yard flag against the offense.

We can see that the value for Defensive Unsportmanlike conduct is also a very low 1 EPA, even though roughing the passer, another 15 yard penalty, is 3 EPA.  This points to the unsportmanlike conduct calls offsetting again.  You can’t offset Roughing the Passer and hence its much higher EPA gain.  We also see Unneccessary roughness falling around the +/-2 EPA range, meaning it is probably less likely to be offset than unsportsmanlike conduct, but more offset than roughing the passer.

Illegal shift being a value of -2 EPA seems high for a 5 yard pre-snap penalty.  My guess here would be that this foul occurs more often when in the red zone, and thus a 5 yard penalty in that area of the field could cause wider EPA swings, particularly if it turns a TD into a FG.

Notably the offensive drive killers are Offensive Pass Interference, a rare call, and offensive holding, a common one.  Both are more damaging than unnecessary roughness and personal fouls, again perhaps due to the chance of offsetting, or post big gain(taunting).

On defense, Defensive pass interference is the most damaging penalty of all at 3.5 EPA, followed by Roughing the Passer at 3 EPA.  Unnecessary roughness is next, and then our 3 interchangeable calls of defensive holding, illegal use of hands, and illegal contact come in at 2 EPA

Converting Net Auto First Down Calls For & Against Into EPA

Here are the same 3 charts posted in the first article about 2023 NFL Penalties.  It shows the auto 1st down penalties for and against for each team, and then the NET of those values.

Now let’s apply the EPA value estimates into the three charts.  I will do one final adjustment so the EPA gained or Lost is shown on a per game basis since we have teams that have played different amount of games.  All we are going to do is multiply each occurance of an auto 1st down passing penalty with its EPA value.

And now let’s just sort from the team that has been hurt the most from these type of penalties by EPA,  down to the team that has benefitted the most the most.

NFL Penalties 2023 – NET EPA on Auto 1st Down Penalties

The Effect of NFL Penalties Measured By EPA – 2023

As you can see,  New Orleans has been tortured by these calls to the tune of -3.40 EPA per game, the only team losing more than 3 points per game(Rams are second punished at -2.83 EPA).   At the other end, Jacksonville has clearly benefitted the most from these types of flags, adding 4.33 EPA per game.

The good news?  For most teams these penalties cost only 1 -2 points per game on average, but therein lies the rub.  These penalties are normally not evenly spaced at all, they are bunched.  This means the EPA per game may be something like: -4, 0, 2 , -1, 0, 0, 6, 4, 1, 1….It is going to be when you see the big number in a single game that the game itself has been influenced by these type of calls.

Over the course of the season, or in the case half season, it balances out to a fairly small amount per game.  It is just that you could be easily fooled into thinking it has no influence if you deceive yourself into believing the penalties are perfectly evenly spaced out such that they truly are just a little bump in every game.  The true nature of the calls is more like a series of small or no bumps, with a few huge potholes thrown in here and there.

What is also interesting is that EPA prouces a completely different list than Yardage Lost/Gained.  Yardage has New Orleans #1, Atlanta #2, & the Giants at #3 as most punished.  EPA has the punitive list as New Orleans, LA Rams, Detroit.

At the lucky end, the lists are also different.  Yardage says teams most benefitted are Jacksonville, Green Bay & Dallas.  EPA says the luckiest teams have been Jacksonville, Carolina, & Minnesota(albeit not with injury luck).

It will be interesting to run this data again at the end of the season to get a true read on what the referees were calling, when they were calling them, and who they were calling them against the most and the least.

As with all the ratings and analysis, feel free to chime in via the post on reddit.com/r/nfl, Practicalist on Reddit, or by sending an email.  Until next week, enjoy the football and best of luck to your team.Until then, enjoy the second half of the season.