Filed under:

# A treatise on the punter’s contributions to the defense

Late in the spring, as the realities of off season hit, I started a project. The goal was to either find a useful metric to rank punters or find a means of quantifying their contributions to the defense. This article is a continuation of that pursuit. The original attempt was very much a work in progress, now this is an attempt to reduce some of the limitations and explore the assumptions of earlier analysis.

### Adjusting points scored per game

One of the first problems that arose from the previous attempt was using number of punt placement and comparing it to points allowed on defense. While this did yield a statistically significant result, the methodology was a bit flawed. After all, not all scoring drives initiate as a result of a punt.

To address this point, I flagged every drive in the NFL database that started immediately after a punt. As before, I extracted out from the play descriptions the placement of the kick, ignored all penalties, punts returned for touchdowns, and if a play occurred twice for any reason, such as replay review, only counted the second instance. I then tallied all of the points scored from every drive that started as a punt return and assigned it to the defensive team and punter responsible for all years available, 2009 to 2017.

The relationship of punt placement to points allowed at the game level is not strong, but it is statistically significant. However, in order to reduce variance and get a clearer picture, we can use this information to construct season averages and look at the same relationship. Punters with fewer than 10 punts in a given year are excluded.

How strong should it be and what would make sense? If you assumed that only the 11 players on the field for the defense influenced points per game allowed under the following assumptions:

• All opponents same quality
• Play calls have no influence
• No substitutions during the game
• All players contributed equally on average

You would expect that player to explain just over 9% of the variability in points allowed over the course of the season. In this case, we can see in the below regression output that our punt placement metric is statistically significant and correlates to the number of points allowed on defense after a punt. How much? It could explain about 6.4% of our variance, not as much as our hypothetical 9% defensive player, but a contribution nonetheless.

### Quantifying the offense in punt placement

A natural question to ask is: Can we exclude, or at least quantify, how much of a punt’s placement is due to the contribution of the offense? After all, a punter could be unduly helped or hindered by the performance of his offense. In order to measure this, I took the 100 yard field position at the time of a punt, paired it with the punt placement, and constructed the yearly averages as before.

The location at which the offense gives the ball to the punter is clearly related to where the punter can place the ball. Our sample indicates it may explain about 34% of the position where a ball lands when punted. This result may be surprising to some who expect the offense to contribute more, or perhaps it’s a confirmation of what you thought before. In any case it seems pretty clear that punt placement belongs mostly to the punter, not the offense.

### Conclusion:

Before starting this latest attempt, I expected to find a stronger relationship between punt placement and defensive points allowed than when we used all scoring drives — but that turned out not to be the case. In my mind, even finding and reporting unexpectedly weak results is important, as is trying to quantify the contributions of every member of a team.

To summarize, in expectation we see a one-yard change in punt placement correlates to about a 0.4 point per game change in defense points allowed. Overall, this correlation seems to explain 6.4% of variance in defensive points per game allowed. The offense’s contribution to punt placement appears to explain about 34% of the overall placement, of the remaining 66% the vast majority should belong to the punter barring pressure from the defense and near blocks. For the next punting article I will be looking at something a little more near and dear to your hearts, is there statistical evidence that Michael Dickson’s punts cause more muffed punt catches or turnovers? Find out only here at Field Gulls.

## Modeling Constraints, Limitations, and Assumptions in this piece:

### Constraints:

One of the biggest limitations to any sort of punting analysis is a lack of hang time tracking data. Without this information, it becomes impossible to comb through a database containing over 4600 punts and judge which punter out kicks their coverage, or develop metrics such seconds/10 yards to judge consistency over distance.

### Limitations:

By and large the NFLScrapR data sets are very consistent and clean, however, we cannot certify they are entirely error free. As such, any analysis done with this set may differ from that performed using the NFL’s internal database, or outside commercial trackers such as Pro Football Focus or Sports Information Systems.

### Assumptions:

I assume in the preceding analysis that any errors within the data set are not so egregious as to distort the overall trends and findings within this piece.

Citation: This, like many other articles I’ve written, wouldn’t be possible without the efforts of everyone involved making the NFLscrapR package a great source of free, open, NFL play by play data.