This document explores the year-over-year stability of eight separate punting statistics on a per-punt basis, finding strong year-over-year correlations for punt yards and punt return yards allowed (among other statistics) and very little correlation for statistics such as touchback rate and out-of-bounds punt rate.

It also includes the fitting of a model using xgboost for punter fantasy performance projection for the 2022-23 NFL season on a per-punt basis, and a brief analysis of the most valuable overall punters in fantasy football for the 2022-23 season.

This project was borne from my desire to forecast punter performance for next year’s NFL season, and the desire to gain an edge on my leaguemates in our punters-only fantasy league (scoring rules are explained further along in the document). It evolved into an analysis of punter statistic stability, which I believe to be more broadly useful to the general reader than my hyperspecific projection of scores according to a highly unscientific scoring system.

While this is an RMarkdown document, I’ve omitted most of my actual R code for readability reasons (this is an article, not a demo). However, if you are interested in how any of this analysis was conducted, you can contact me via email at .

Year-over-year correlational analysis for each statistic:

All statistics given are for the relationship between a given statistic in year n and that statistic’s value in year n-1.

Touchback rate: defined as the percentage of a punter’s attempted punts resulting in touchbacks for a given season.

## Year-over-year R-Squared for touchback rate: 0.096
## P-value for single-predictor model significance: 0.12723

Fair catch rate: defined as the percentage of a punter’s attempted punts resulting in fair catches by the opposing team for a given season.

## Year-over-year R-Squared for touchback rate: 0.304
## P-value for single-predictor model significance: 1e-05

Out-of-bounds rate: defined as the percentage of a punter’s attempted punts resulting in a kick out of bounds for a given season.

## Year-over-year R-Squared for touchback rate: 0.112
## P-value for single-predictor model significance: 0.07578

Return rate: defined as the percentage of a punter’s attempted punts resulting in an attempted return for a given season.

## Year-over-year R-Squared for touchback rate: 0.216
## P-value for single-predictor model significance: 0.00053

Return yardage rate: defined as the average return yardage obtained by the opposing team per punt by a given punter in a given season. Note: this includes all kicks, not just those where a return was attempted.

## Year-over-year R-Squared for touchback rate: 0.349
## P-value for single-predictor model significance: 1e-05

Punt inside 20 rate: defined as the percentage of a punter’s attempted punts resulting in the opposing team starting their drive inside the 20-yard line (before penalty yardage is assessed) for a given season.

## Year-over-year R-Squared for touchback rate: 0.275
## P-value for single-predictor model significance: 1e-05

Punt inside 10 rate: defined as the percentage of a punter’s attempted punts resulting in the opposing team starting their drive inside the 10-yard line (before penalty yardage is assessed) for a given season.

## Year-over-year R-Squared for touchback rate: 0.269
## P-value for single-predictor model significance: 1e-05

Punt yardage rate: defined as the average net punt yards on punts by a given punter for a given season (excluding blocked kicks).

## Year-over-year R-Squared for touchback rate: 0.354
## P-value for single-predictor model significance: 1e-05

Which statistics are the most significant predictors year-over-year? Which are the least?

Fitting a model for prediction of next-season punter fantasy points: Punter fantasy points are computed via the following formula:

Just like the analysis above, this will focus on per-punt stats, instead of total stats, to avoid bias as a result of games played or punts attempted. Note that this model will NOT take into account a few important factors: most importantly, how many times a punter punts per game (controlled by a variety of factors, including head coach fourth-down aggressiveness and offensive performance). Understanding punter availability is a key part of projecting punter fantasy scoring, just as target and carry projections are key to WR and RB evaluation, but this model is not suitable for its projection.

Instead, pay close attention to team roster news regarding punters (I recommend setting up a Google alert, especially for combing the midseason waiver wire) and utilize other analysis from sources like PFF for head coach aggressiveness data and team offensive projections. Even the most talented punters on a per-punt basis won’t be league winners if their coach won’t give them a chance to shine.

The next step is projecting punter performance for next season using this data. I fit two models to do this — one using a multiple linear regression model and the other using xgboost — and have plotted their performance on the testing set below. Visually, it’s clear that the xgboost model performs better (average distance from the actual = predicted line is smaller).

While the xgboost model is more of a black box than a linear regression, we can still visualize the most important features for prediction using a feature importance plot.

Using the xgboost model to forecast the most effective punters for 2022-23 in terms of fantasy points earned per punt:

Players with high expected points-per-punt who are on teams without projected high-level offenses (i.e. more likely to punt) are my favorite targets here. Their team from last season is listed in the table below, but since punter movement is very common in the NFL, many players have switched allegiances in the offseason. In the blurbs below I have included up-to-date roster info on each punter.

Some names that will be at the top of my draft board because of their combination of poor projected offense and high-level points-per-punt projection:

Michael Dickson, SEA: Dickson was the best punter on a points-per-punt basis last year, though the model has him pegged for regression in 2022-23, as he’s ranked just seventh in the projections. However, he might be the highest-value player in the draft on account of a perfect storm of opportunity. The Seahawks finished second in the NFL in punts last season, and are primed to punt even more frequently this year as their offense takes a step back and their conservative playcalling situation in terms of fourth down aggressiveness remains unchanged.

Thomas Morstead, MIA: Morstead spent time with both the Jets and the Falcons last season, putting up excellent statistics with each team and finishing first in points-per-punt. Your ranking for him should depend on how much you believe in a revamped Miami offense to get drives going. The Dolphins ranked fifth in the NFL in punts last season.

Logan Cooke, JAX: Cooke is projected to be the fifth-best punter on a per-punt basis next season after putting up solid numbers on high volume in Jacksonville last season. No matter how you slice it or project Trevor Lawrence’s development, the Jaguars will probably be a top-half team in punts this season, although their high turnover numbers meant fewer punts than you might expect from such a putrid offensive team (they ranked just ninth in the NFL in punts last year).

rank name team Proj. PPP
1 Thomas Morstead NYJ 7.85
2 AJ Cole LV 7.85
3 Jake Bailey NE 7.77
4 Bryan Anger DAL 7.12
5 Logan Cooke JAX 7.12
6 Corey Bojorquez GB 6.78
7 Michael Dickson SEA 6.56
8 Tress Way WAS 6.55
9 Johnny Hekker LA 6.27
10 Sam Martin DEN 6.18
11 Cameron Johnston HOU 6.15
12 Tommy Townsend KC 6.13
13 Andy Lee ARI 6.12
14 Jack Fox DET 6.12
15 Kevin Huber CIN 6.12
16 Brett Kern TEN 6.06
17 Michael Palardy MIA 6.04
18 Rigoberto Sanchez IND 6.03
19 Matt Haack BUF 5.78
20 Sam Koch BAL 5.69
21 Bradley Pinion TB 5.58
22 Pat O'Donnell CHI 5.31
23 Mitch Wishnowsky SF 5.15
24 Jamie Gillan CLE 4.81
25 Riley Dixon NYG 4.67
26 Ty Long LAC 4.14