Correlations Between Punting and Defense and Punting and Winning

[kccrow]

Ok, so you weren't wrong to say 8 of the top 10 in PPG were top 15 in Punts inside the 20%. However, this theory doesn't really hold up in other years in the data set. Also, I've shown no strong correlation. There also isn't a strong correlation between YDS/G and Punts inside the 20 %.

Despite the fact that we haven't found any strong indicator that punting affects defense in a meaningful way this did prompt me to do one more data set, and that is Win % versus Punt IN 20 % and that has given the strongest correlation of all data.

Spoiler!

2017 Correlation Coefficient of W% to %IN 20 0.469
2016 Correlation Coefficient of W% to %IN 20 0.316
2015 Correlation Coefficient of W% to %IN 20 -0.042
2014 Correlation Coefficient of W% to %IN 20 0.638
2013 Correlation Coefficient of W% to %IN 20 0.401

This is surprising, because one would expect that if they did affect W% they'd also affect Defensive performance. That really doesn't seem to be the case.

[kccrow]

Number of teams in the top 10 in % of Punts IN 20 that were also in top 15 in PTS/G

2017: 8
2016: 6
2015: 5
2014: 6
2013: 5

What this tells me is that if you have a punter that is in the top 30% of the league in % IN 20, then you've got about a 60% chance of being in the top half the league in PTS/G. But, if you have one of the 70% of the rest of the punters, you still have a 40% shot... However, 2017 skews the data and if we were to regress that value more towards the mean (6), we'd end up closer to a 55%/45% split.

More to follow...

[kccrow]

Number of Punters above and below one standard deviation from the mean by year:

2017: >1STDEV = 5, <1STDEV = 3
2016: >1STDEV = 6, <1STDEV = 4
2015: >1STDEV = 6, <1STDEV = 6
2014: >1STDEV = 7, <1STDEV = 8
2013: >1STDEV = 2, <1STDEV = 5

Here are the approximate ranges within 1STDEV of the mean, so the data I worked with were for punters above and below these:

2017 1STDEV = 42.98% to 30.42%
2016 1STDEV = 44.69% to 29.99%
2015 1STDEV = 42.29% to 27.85%
2014 1STDEV = 41.68% to 28.90%
2013 1STDEV = 41.41% to 28.59%

I combined these into 1 list to get some data, and so I ended up with 26 punters in total in each category.

>1STDEV
Average %IN20 = 46.4% (Min 42.5%)
Average W% = 59.6%
Average PTS/G = 20.9
Average YDS/G = 344.4

<1STDEV
Average %IN20 = 26.1% (Max 30.0%)
Average W% = 35.1%
Average PTS/G = 24.2
Average YDS/G = 352.6

If I take the data set as a whole I get the following correlations:
IN20% to YDS/G = -0.140
IN20% to PTS/G = -0.430
IN20% to W% = 0.565

What I've learned. It is better to have a great punter in terms of %IN20 than an absolutely shitty punter. That said, overall that means you need to find a punter that is within 1STDEV of the mean or better. It may indicate as well that if you have a punter that is better than one standard deviation from the mean, he is an advantage.

In case you're curious, here's the best and worst through the past 5 years in terms of %IN 20.

Spoiler!

Best Punters 2017
1. Chris Jones, DAL
2. Sam Koch, BAL
3. Johnny Hekker, LAR
4. Dustin Colquitt, KC
5. Thomas Morstead, NO

Worst Punters 2017
1. Brad Wing, NYG
2. Justin Vogel, GB
3. Britton Colquitt, CLE

Best Punters 2016
1. Bryan Anger, TB
2. Johnny Hekker, LAR
3. Dustin Colquitt, KC
4. Matt Bosher, ATL
5. Sam Koch, BAL
6. Jeff Locke, MIN

Worst Punters 2016
1. Britton Colquitt, CLE
2. Colton Schmidt, BUF
3. Bradley Pinion, SF
4. Kevin Huber, CIN

Best Punters 2015
1. Dustin Colquitt, KC
2. Marquette King, OAK
3. Jordan Berry, PIT
4. Brad Wing, NYG
5. Johnny Hekker, LAR
6. Ryan Allen, NE

Worst Punters 2015
1. Mike Scifres, SD
2. Tim Mathay, GB
3. Shane Lechler, HOU
4. Britton Colquitt, DEN
5. Jacob Schum, TB
6. Colton Schmidt, BUF

Best Punters 2014
1. Jon Ryan, SEA
2. Donnie Jones, PHI
3. Dustin Colquitt, KC
4. Pat McAfee, IND
5. Sam Koch, BAL
6. Drew Butler, ARI
7. Sam Martin, DET

Worst Punters 2014
1. Michael Koenen, TB
2. Bryan Anger, JAX
3. Spencer Lanning, CLE
4. Jeff Locke, MIN
5. Pat O'Donnell, CHI
6. Marquette King, OAK
7. Tim Masthay, GB
8. Tress Way, WAS

[kccrow]

Looking into Net Average, in the same way I looked at %IN 20, I came away with results that indicate Net Average is not a factor.

Number of Punters above and below one standard deviation from the mean by year:

2017: >1STDEV = 5, <1STDEV = 4
2016: >1STDEV = 5, <1STDEV = 4
2015: >1STDEV = 5, <1STDEV = 5
2014: >1STDEV = 5, <1STDEV = 6
2013: >1STDEV = 5, <1STDEV = 3

Here are the approximate ranges within 1STDEV of the mean, so the data I worked with were for punters above and below these:

2017 1STDEV = 42.30 to 39.14
2016 1STDEV = 42.21 to 38.44
2015 1STDEV = 41.68 to 38.16
2014 1STDEV = 41.30 to 37.73
2013 1STDEV = 41.55 to 37.48

I combined these into 1 list to get some data, and so I ended up with 25 punters above and 22 punters below 1STDEV.

>1STDEV
Average Net = 42.9
Average W% = 50.75%
Average PTS/G = 22.4
Average YDS/G = 348.2

<1STDEV
Average Net = 37.2
Average W% = 47.44%
Average PTS/G = 22.8
Average YDS/G = 337.7

If I take the data set as a whole I get the following correlations:
NetAve to YDS/G = 0.164
NetAve to PTS/G = -0.080
NetAve to W% = 0.065


After seeing all of this data, I can pretty well conclude that the only real factor you need to look at with a punter is % inside the 20. That seems to be the only factor that has any significant impact on games and it seems to do so in terms of PTS/G and W%. Now the question is, at what point is there a cutoff where it no longer matters? This would tell me a reasonable break in what should define a punter "worth paying" versus one that isn't. My next question would be, at what point is the cutoff for bad punters? I'm working on formulating how I want to tackle these questions. Basically, I have to continue to include punters by %IN 20 in tiers until I reach a 50% mark for win percentage, IMO. If anyone has a suggestion, feel free to chip in.

[Chiefshrink]

Originally Posted by kccrow

After seeing all of this data, I can pretty well conclude that the only real factor you need to look at with a punter is % inside the 20. That seems to be the only factor that has any significant impact on games

Considering that Peters,Hali and Johnson are gone and Ford,KPass are still question marks, with Amerson and Fuller getting acclimated with Fuller being the only real starting CB with no hard hitting safety, with inconsistent d-line play; our defense is still very suspect and why a good punter helps because when our offense gets stopped it will now be between the 40's with Mahomes(where a finesse punt is needed more often) as opposed to being stopped more often than not inside our 40 with Smith at QB(where you need a booming punt).

With Mahomes now at QB with still a very suspect defense you need finesse punting more than ever pinning back the opposing offense giving your suspect defense more of a chance for a stop and why Colquit got paid. Veatch recognizes his defense needs help in the field position game for sure.

[kccrow]

Originally Posted by Chiefshrink

Considering that Peters,Hali and Johnson are gone and Ford,KPass are still question marks, with Amerson and Fuller getting acclimated with Fuller being the only real starting CB with no hard hitting safety, with inconsistent d-line play; our defense is still very suspect and why a good punter helps because when our offense gets stopped it will now be between the 40's with Mahomes(where a finesse punt is needed more often) as opposed to being stopped more often than not inside our 40 with Smith at QB(where you need a booming punt).

With Mahomes now at QB with still a very suspect defense you need finesse punting more than ever pinning back the opposing offense giving your suspect defense more of a chance for a stop and why Colquit got paid. Veatch recognizes his defense needs help in the field position game for sure.

I was just talking with a friend about this very thing.

We also discussed looking at how offensive performance affects Punt Inside 20 %. I think there's a solid chance it is a factor. Just looking at the data I've poured through and the teams I see having punters at the worst end of that, like Buffalo, Cleveland, New York, etc, I just want to see how it looks. I've put way too much time into this tonight and I'm not any further.

I'm looking for more beyond Dustin Colquitt though in all of this. I want to see if there really is a reason to pay a punter good money and when that situation should exist from a statistical perspective. I think I'm close to that answer now, but I don't know if I can take it much further.

It's been forever since I got my undergrad degree in mathematics. I haven't used it for a very long time, especially anything beyond basic algebra because of the field I'm in. I don't' have the time nor do I remember much on multivariate analysis. I'd love to see someone that does take this to another level. It'd be interesting to see it.