What conditions represent a good matchup for a kicker? Here is a list of the most common recommendations I’ve seen:
- Go with the dome kicker. Kickers are perceived to be more accurate in domes.
- Look for a heavy defensive matchup. When playing good defenses, teams are forced to settle for more field goals and, hence, kickers score more.
- Go for a kicker in a high scoring game. An above-average score should result in above average scoring for the kicker.
- Try and get kickers that are playing on MNF or SNF. They seem to step up in those games.
The last item is silly. It is neither logical nor supported by the data.
The first item, I argue, is given far too much weight. Given that 3/4’s of kickers make 80% of their field goals or more, I don’t think this accuracy effect is large. What’s more, two of the most inaccurate kickers in the league this season play in domes (Wilkins and Mare). Higher accuracy also has no effect on the number of field goals attempted. Detroit, for instance, plays in a dome yet is on a pace to allow only 10 field goals this season.
Items 2 and 3 seem to be cutting to the heart of the matter. These two arguments are seemingly in conflict with one another. One argues that playing a tougher defense will result in more kicker points while the other suggests that playing a weaker opponent will yield a similar result. So which is better, a tough or weak opponent? Let’s get into the data.
Methodology
Using statistics from Pro Football Weekly and NFL.com I gathered the following data for all 32 NFL teams dating back to the 2003 NFL regular season:
- Points Allowed per Game
- Red Zone TD Efficiency (%)
- Kicker Points Allowed
In addition to red zone touchdown efficiency, third down efficiency may also be a factor; however, red zone touchdown efficiency is, by my estimation, a linear function of third down efficiency. As such, I opted to exclusively look at red zone efficiency as it should theoretically be the more descriptive of the two.
Kicker points allowed is based on the actual point total of all field goals and extra points scored. No modifications were made based on the distance of these field goals. Field goal distance is an exogenous explanatory variable given that the range of the kicker has nothing to do with the defensive opponent. I may revisit this topic in a future blog.
Data
I utilize the following population model for this study:
KPPG = β0 + β1(RZ%) + β2(PPG) + u
Where RZ% is the opponent red zone touchdown efficiency (%), PPG is total points per game allowed, and KPPG is kicker points per game allowed.
Here are the results:
Interpretation:
- An increase of 1 PPG is correlated with an increase in .25 KPPG.
- An decrease of .01 RZ% (red zone touchdown efficiency increase of 1%) is correlated with a decrease of -.05 KPPG.
Findings
Not surprisingly, the effect of PPG on KPPG is very large. Thus, we can confidently pick matchups where a kicker is playing a defense that yields a lot of total points. The impact of defenses RZ% suggests that a low percentage of touchdowns will result in more field goals; however, this effect is very small. So while a strong defense has some impact on kicker scoring, magnitude of the PPG effect is much, much larger. Thus, total points allowed by a defense should have far greater weight in your matchup decision than RZ%.
As you can see, the model turns out to be fairly predictive. The regression equation predicts the following matchup strength based on NFL statistics through week 8:
6 comments:
Awesome! This is what I've been saying all year, so it's great to see a statistical analysis supporting it. Thanks.
Thanks for your feedback! I'm glad you found the article helpful. I hope you find my future analysis just as insightful.
Hey there. I think you're analysis was pretty cool but on a week to week basis, it doesn't seem to help very much. This link is to week 10's points for and points against and total points vs. the actual production by kickers:
http://spreadsheets.google.com/pub?key=pOonhrXHwsoO7HPH2lHxmcg
If you graph it, points against correlated very poorly (r-squared of 0.003 vs. total PF and PA, 0.0001 vs. PA and 0.00001 vs. PF) with actual production for the week. Pittsburgh was the best on the total matchup of PF and PA and had 11 points and Dallas was next and had 8 points but then IND, NO both had 1 point and Detroit had 3 points. Then there's a bunch that did a lot better. GB, NYG, ARI, BUF and then finally shows the week leader at 21. Then a bunch of other guys before KC with 9, SD with 5 and STL with 15!
Any way to adjust your formula to improve on this?
I appreciate your comments. The one week look you are taking at the numbers is not a statistically valid comparison for the reason that the sample sizes are different. As your sample size increases, the level of precision in the model will also increase. The Central Limit Theorem tells us that a sample size of 30 is necessary for us to assume our sample is normally distributed. Most serious econometricians would be uncomfortable with anything less than 80. Because you have a much smaller sample, the goodness-of-fit of your model will necessarily be smaller and you will have a smaller R-Squared (because your residuals are larger). I would caution you against placing too much emphasis on r-squared as there are many many instances where r-squared will be smaller than alternatives yet you have a better model. Instead, you should be concerned with the precision of your predicted parameters. In this model, I have P-values which are statistically significant to the .999 level. Thus, either there is a relationship between TPPG and RZ%, or else this is one of those 1 in 1000 samples. Thats a very strong model.
Evaluated on a week-to-week basis, the predictive power may not always be evident. But if you track this over time, it should give you results similar to this model.
In applying this model, it is important to understand what it says. That is, higher scoring games will result in more kicker points. Baltimore was #16 in PPG heading into week 9 and yet they yielded a high number of kicker points to Jeff Reed. This, however, was quite predictable given the injuries sustained in the secondary. Also keep in mind that while a team like CIN is giving up a horrific number of TPPG, that doesn't mean a team like San Franciso is going to score a lot of kicker points. They are likely to score relatively more than their average, but if they are the leagues worst offense, they aren't going to score as much.
After looking at your spreadsheet, I'm not exactly sure what you are graphing. Why are there fractional points? You do realize that this model is based on actual points allowed by a defense and not fantasy points allowed by a defense?
If I run the numbers on this just using this weeks total points allowed and kicker production, I get a statistically significant correlation (P = .04).
http://spreadsheets.google.com/ccc?key=pi7f3GubqbeO0slTIa0MXKg&hl=en
Realize that this study is merely saying the higher scoring game will produce the most kicker points. Just playing Cincinnati isn't going to cause the leagues worst offense to score a ton of points any more than playing Pittsburgh will cause the leagues best offense to score fewer points.
When deciding on a matchup and choosing, for instance, Josh Brown against SFO or Phil Dawson against PIT, you would obviously want to go with what you anticipated to be the higher scoring game. In this case, SEA was ostensibly the best choice. Choosing the right player is not a perfect science but this article does prove, definitively, that defensive showdowns, MNF games, or any other voodoo method of choosing a kicker is wrong.
Post a Comment