Data scientist, physicist, and fantasy football champion

Week 10 K predictions

Kickers are tricky…

These models all did a little better than FantasyPros top 20 K experts in weeks 8 and 9, but it’s still not fabulously accurate. Model C was the best in week 9, getting an average of 1.0 points closer to the actual score than the average of the top 20 Pros.

Model A:

 

Initially I suspected that the data wouldn’t fit a normal (gaussian) distribution because the kicker’s scores are basically bounded at 0 and can get fairly large (the highest I have so far is 28 pts from Cairo Santos last year). I still think that’s possible, but it may be only necessary to get accurate prediciton intervals which I don’t really care much about right now.

Model A uses a survival model with a Weibull distribution. It’s essentially a weibull regression (as opposed to linear models like C and D that assume normal distributions). Models A and B did well last week, but I don’t think the extra steps are necessary. That said I’m going to show those models here because they do well.

This is the model that uses most of the parameters I have right now. I threw it into JMP and found the following factors were all significant (p < 0.1 because 0.5 felt too restrictive). In decreasing order of significance, it uses:

  • Predicted team score
  • Predicted opponent score
  • Which team is the home team (stadium)
  • The player
  • The opponent

I don’t remember what the effect sizes are, though. That would have been a more helpful order. I’ll get to it though, so relax. It also only uses the data from 2016.

Model B:

 

Model B is the same as Model A but with both 2015 and 2016 data. This did well last week but A was a little better.

Model C:

 

Model C uses the same factors as Model A:

  • Predicted team score
  • Predicted opponent score
  • Which team is the home team (stadium)
  • The player
  • The opponent

But is a linear model (i.e., normal distribution of data). This was the best model is weeks 8 and 9, but not by a lot. It’s worth following it to see how it does week to week and if models A and B do any better.

Model D:

 

Model D was the worst of my models last week, but still better than Fantasy Pros (I think, more on that later). Same model as C but 2015 and 2016 data.

Week 10 model summary

RankModel.AModel.BModel.CModel.D

1Nick NovakCairo SantosNick NovakJustin Tucker

2Phil DawsonDustin HopkinsCaleb SturgisNick Novak

3Mike NugentNick FolkMike NugentStephen Gostkowski

4Chandler CatanzaroDan BaileyDustin HopkinsChris Boswell

5Dustin HopkinsGreg ZuerleinJosh LamboConnor Barth

6Jason MyersNick NovakJustin TuckerDustin Hopkins

7Josh LamboJosh BrownJason MyersJosh Brown

8Caleb SturgisStephen GostkowskiChandler CatanzaroNick Folk

9Steven HauschkaChris BoswellMatt BryantMike Nugent

10Dan BaileyGraham GanoBrandon McManusDan Bailey

11Josh BrownSteven HauschkaSteven HauschkaChandler Catanzaro

12Chris BoswellConnor BarthChris BoswellJosh Lambo

13Brandon McManusMike NugentNick FolkCaleb Sturgis

14Nick FolkJustin TuckerMason CrosbyMason Crosby

15Mason CrosbyJosh LamboStephen GostkowskiSteven Hauschka

16Stephen GostkowskiChandler CatanzaroJosh BrownCairo Santos

17Greg ZuerleinBlair WalshDan BaileyGreg Zuerlein

18Andrew FranksMason CrosbyGreg ZuerleinBrandon McManus

19Connor BarthJason MyersRyan SuccopMatt Bryant

20Matt BryantPhil DawsonCairo SantosJason Myers

21Justin TuckerBrandon McManusPhil DawsonBlair Walsh

22Blair WalshCaleb SturgisAndrew FranksPhil Dawson

23Graham GanoMatt BryantGraham GanoGraham Gano

24Cairo SantosAndrew FranksConnor BarthRyan Succop

25Ryan SuccopRyan SuccopBlair WalshAndrew Franks

If I listen to the top 2 accurate models (A and C) I should pick up Nick Novak this week (I think he’s available). I currently have Caleb Sturgis (2nd best in Model C, 8th in Model A) and he ’s done well this year. Then again, both he and Nick Novak have both kicked 2 or more FGs every week and Novak is going against JAX this week and they’ve given up the most points to kickers this year. Hopefully I’ll make a decision before I post this and anyone in my league reads this and acts on it.

I’m kidding, they won’t. They’re idiots. Especially Christian who, unrelatedly, is my opponent this week. He’s terrible.

 

Update: I picked up Nick Novak.  Come on buddy!  I believe in you!

Week 10 DEF - Evaluation

Week 10 DEF predictions