Data scientist, physicist, and fantasy football champion

Week 1 K Predictions

Predicting kickers is no better than randomly guessing. None of my models did well last year, so I’m largely throwing them out and trying again. Model D was good enough to keep around, but I’m dropping the rest for now. Since all of the Bayesian models were worse than randomly guessing I’m dropping most of them, too.

Let’s make this fast. I have a 9 week old baby in one arm who just REFUSES to sleep and typing one handed is slow, so I’m not going to waste any more time on kickers than I have to.

[Note: There was a weird shuffling of kickers during the offseason. 6 or 7 of them moved teams, Lambo is out, Gano might be as well, and there are some new guys out there like Elliott and Yoo. I can’t model rookies until they have at least one week under their belts, so if you see something weird or if I’m recommending a player that isn’t starting, just ignore it. It’s more work to exclude players in a week than it is to model nonsense, but I’m not going to waste any more time on kickers than I have to. This weirdness will be sorted out a little better in week 2 when I know who is actually starting and when I have a little more data on them.]

ModelTypePlayerScoreOppScoreScore2OppScore2OppHomeTeamDataYears
DLMXXX XX2015-2017
ELMXXXXX 2015-2017
FLMX 2015-2017
GGibbs SamplerXXXXX 2015-2017
HLM X 2015-2017


Model D was the best last year, but barely better than randomly guessing.

Model E includes the most statistically significant terms and only those terms with p > 0.05. I’m not trying to make a statement here about p-values, it just seemed as good as threshhold as anything.

Model F simply asks: who has done well in the past? Play that guy.

Model G turns Model E into a Bayesian model

Model H is based solely on who is expected to score the highest that week (the most statistically significant term). It’s a little rude, but it assumes that kickers are interchangeable. Probably not true, but it should capture something. It’ll be neat to compare this to a random guess.

For now all of these use all of the data that I have (2015-present). I may eventually make new models with limited data, but I see no reason to think that a kicker drastically changes between years. They sometimes change teams between years, so I’ll likely try a model with only 2017 data in a few weeks when I have enough data.

Model D:


Terms:

The Player
Score
Opponent Score
Opponent
Home Team
Data years: 2015 - 2017

.Model D-1.png

Model E:


Terms:

The Player
Score
Opponent Score
Score^2
(Opponent Score)^2
Data years: 2015 - 2017

.Model E-1.png

Model F:


Terms:

The Player
Data years: 2015 - 2017

Model G


Terms:

The Player
Score
Opponent Score
Score^2
(Opponent Score)^2
Data years: 2015 - 2017

Bayesian (Gibbs sampler)

.Model G-1.png

Model H:


Terms:

The predicted score
Data years: 2015 - 2017

.Model H-1.png

Random guess!

.random guess-1.png


Weekly model overview

RankDEFGHRandom
1Dan BaileyMatt BryantJustin TuckerMatt BryantMatt BryantDan Bailey
2Justin TuckerStephen GostkowskiMatt BryantStephen GostkowskiStephen GostkowskiNick Novak
3Nick NovakJustin TuckerStephen GostkowskiGraham GanoChris BoswellJason Myers
4Brandon McManusGraham GanoWil LutzDan BaileyMason CrosbySteven Hauschka
5Matt PraterDan BaileyGraham GanoSteven HauschkaGraham GanoChandler Catanzaro
6Matt BryantSteven HauschkaDustin HopkinsChris BoswellRyan SuccopMason Crosby
7Robbie GouldNick NovakCairo SantosJustin TuckerDan BaileyChris Boswell
8Stephen GostkowskiChris BoswellSteven HauschkaNick NovakKai ForbathSteven Hauschka
9Steven HauschkaBrandon McManusSteven HauschkaBrandon McManusPhil DawsonRyan Succop
10Dustin HopkinsNick FolkCaleb SturgisCaleb SturgisSebastian JanikowskiKai Forbath
11Josh LamboCaleb SturgisBrandon McManusNick FolkCaleb SturgisRandy Bullock
12Wil LutzDustin HopkinsDan BaileyDustin HopkinsSteven HauschkaDustin Hopkins
13Caleb SturgisMatt PraterChris BoswellKai ForbathSteven HauschkaAdam Vinatieri
14Greg ZuerleinSteven HauschkaAdam VinatieriMason CrosbyBrandon McManusMatt Prater
15Steven HauschkaKai ForbathNick NovakMatt PraterDustin HopkinsSebastian Janikowski
16Phil DawsonWil LutzMatt PraterSteven HauschkaMatt PraterGreg Zuerlein
17Mason CrosbyMason CrosbyChandler CatanzaroPhil DawsonNick NovakAndrew Franks
18Randy BullockPhil DawsonMason CrosbyWil LutzRandy BullockCairo Santos
19Graham GanoGreg ZuerleinSebastian JanikowskiGreg ZuerleinGreg ZuerleinRobbie Gould
20Chris BoswellRobbie GouldRobbie GouldSebastian JanikowskiNick FolkPhil Dawson
21Nick FolkSebastian JanikowskiJason MyersRobbie GouldRobbie GouldPhil Dawson
22Cairo SantosAdam VinatieriNick FolkRyan SuccopWil LutzMatt Bryant
23Sebastian JanikowskiRyan SuccopJosh LamboAdam VinatieriConnor BarthGraham Gano
24Kai ForbathRandy BullockKai ForbathRandy BullockPhil DawsonNick Folk
25Jason MyersJosh LamboRandy BullockJosh LamboJosh LamboCody Parkey
26Adam VinatieriCairo SantosConnor BarthJason MyersAndrew FranksWil Lutz
27Ryan SuccopJason MyersRyan SuccopPhil DawsonCairo SantosJustin Tucker
28Phil DawsonPhil DawsonGreg ZuerleinCairo SantosJustin TuckerStephen Gostkowski
29Cody ParkeyConnor BarthPhil DawsonConnor BarthCody ParkeyJosh Lambo
30Connor BarthCody ParkeyPhil DawsonCody ParkeyAdam VinatieriBrandon McManus
31Chandler CatanzaroChandler CatanzaroCody ParkeyChandler CatanzaroJason MyersCaleb Sturgis
32Andrew FranksAndrew FranksAndrew FranksAndrew FranksChandler CatanzaroConnor Barth

 

Conclusions


Models E, F, and G give similar results while Models D and H are almost opposites of each other in some places; D and H have very different opinions on Gano, Boswell, Tucker, and others.

Most of these models like Gostkowski, Bailey, and Bryant. Hell, even the random guess likes Bailey. I like Gano and as of writing he’s available in my league so if he plays I’ll probably play him. Janikowski, Forbath, Lutz, and Sturgis consistently show up in the top half of my models, so maybe go with one of them?

I’m sorry, I can’t even. I’m just so sick of kickers already and the season hasn’t even started. Andrew Franks and Cody Parkey will probably kick 10 field goals each this week. Screw kickers, screw my models, I’m out!

I’m also so tired. Just so tired. Baby, you need to sleep. Your dad needs to rest so he can crush your uncle and his idiot friends.

Week 1 DEF Results

Week 1 QB Predictions