Data scientist, physicist, and fantasy football champion

Week 14 DEF Prediction

Last week Models A and F did the best of my models, with B and C performing well and D and E bringing up the rear.

Interestingly, Yahoo’s experts beat me, but since that’s the first time that this happened since I’ve been documenting my attempts at modeling I’m not too worried. It still feels weird to make claims about accuracy when I’m grading myself. I feel okay about the accuracy algorithm that I’m using, but I’d still love to actually be considered a FantasyPros Expert and get them to rate me every week. Until then I feel a little odd saying confidently that I’m better than anyone. There’s a difference between confidence and arrogance, and I’m much better at the latter.

This week I’m continuing to use the same six models, partially because I want to see a little more consistency and partially because adding or removing models from the Markdown file takes time and I’m a busy man. Here are this week’s models with a summary below. As usual, if you want to know more check out the Methodology and Data Exploration sections.

Model A:

Model A parameters:

  • The team that’s playing
  • Who their opponent is
  • Their predicted score
  • Their opponent’s predicted score
  • Whether the team is home or away

Model A Notes:

  • Uses only data from 2016
  • Best model from weeks 8 and 9, 3rd best in week 10
  • Terrible in week 12 but great in Week 13

Model B:

Model B parameters:

  • The team that’s playing
  • Who their opponent is
  • Their opponent’s predicted score

Model B Notes:

  • Uses only data from 2016
  • Comparable results to Model A while reducing overfitting
  • Terrible in week 12

Model C:

Model C parameters:

  • The team that’s playing
  • Who their opponent is
  • Their predicted score
  • Their opponent’s predicted score
  • Whether the team is home or away

Model C Notes:

  • Uses data from 2015 and 2016
  • OK performance weeks 8 and 9, best model in week 10

Model D:

Model D parameters:

  • The team that’s playing
  • Who their opponent is
  • Their predicted score
  • Their opponent’s predicted score
  • Whether the team is home or away

Model D Notes:

  • Uses only data from the last 6 weeks

Model E:

Model E parameters:

  • The team that’s playing
  • Who their opponent is
  • Their predicted score
  • Their opponent’s predicted score
  • Whether the team is home or away
  • NEW: “Team:week” term (does the team have momentum?)
  • NEW: “Opponent:week” term (does the opponent have momentum?)
  • NEW: “week” term (basically a bookkeeping term if I want to use the other two)

Model E Notes:

  • Uses data for last 8 weeks
  • includes week + Team:week and Opponent:week terms for each team
  • New terms try to account for momentum

Model F:

Model F parameters:

  • The team that’s playing
  • Who their opponent is
  • Their predicted score
  • Their opponent’s predicted score
  • Whether the team is home or away

Model F Notes:

  • Uses data from 2015 and 2016
  • Best model in weeks 11*, 12, and 13
  • Ignores defense TDs but weights interceptions and fumbles to try to account for it

Week 13 model summary

Rank

Model.A

Model.B

Model.C

Model.D

Model.E

Model.F

1

MIN

MIN

MIN

MIA

CIN

MIN

2

CAR

CAR

DEN

CIN

DET

CIN

3

DEN

DEN

MIA

MIN

MIN

DEN

4

MIA

SD

CAR

CAR

ARI

TEN

5

SD

SF

CIN

TB

CLE

MIA

6

SF

MIA

SD

DET

MIA

CAR

7

CIN

ATL

TEN

SD

TB

SD

8

ATL

CIN

DET

SEA

NE

WAS

9

TB

TB

IND

DEN

IND

IND

10

ARI

ARI

ATL

IND

KC

SF

11

DET

PHI

SF

PIT

PIT

DET

12

WAS

DET

WAS

CLE

DEN

ARI

13

TEN

WAS

SEA

GB

NYJ

PIT

14

SEA

TEN

ARI

TEN

LA

SEA

15

PHI

KC

KC

NE

SD

TB

16

HOU

HOU

PHI

ARI

NYG

ATL

17

KC

SEA

HOU

KC

TEN

NE

18

PIT

IND

PIT

ATL

DAL

KC

19

IND

PIT

TB

NYG

HOU

HOU

20

NE

NE

NE

HOU

CAR

OAK

21

GB

BUF

NYG

SF

OAK

PHI

22

BAL

GB

GB

CHI

GB

GB

23

BUF

NYJ

LA

LA

NO

NYJ

24

CLE

CLE

OAK

WAS

CHI

BUF

25

NYJ

NO

CLE

OAK

ATL

LA

26

OAK

LA

BAL

NYJ

SEA

CLE

27

NO

OAK

DAL

JAC

WAS

NYG

28

LA

BAL

NO

BAL

BUF

DAL

29

NYG

NYG

BUF

PHI

JAC

NO

30

JAC

JAC

JAC

NO

SF

JAC

31

DAL

DAL

NYJ

BUF

BAL

BAL

32

CHI

CHI

CHI

DAL

PHI

CHI

Since Models A and F seem to be the best I’ll stick with them this week and point out a few interesting choices.

Anyone surprised to see MIN and DEN near the top really hasn’t been paying attention this year. I would say that those are the only two teams up here that I would consider sure-fire starts. My brother likes “tiers”, so let’s call them “Tier 1”. Since he has MIN he can stop reading here (thereby dropping my current weekly readership by 100%).

Tier 2 (ugh, that feels gross) contains a lot of good teams with good matchups. CIN, TEN, and MIA are going up against CLE, DEN, and ARI, respectively. Each of those defenses have had good and bad weeks but are favored in low-scoring games this week. ARI is definitely wiley (I had WAS last week), but CLE has given up an enormous number of points the last 3 games with all 3 games in double-digits and 2 of those games with more than 20 points. Also in this… group… are CAR and SD facing each other. Both defenses have scored in the upper-middle of the pack on the year and both have given up points to team defenses in the upper-middle. Vegas is predicting an upper-middle scoring affair (around 23-24 points each as of this writing). Given their histories this season I feel like you can expect a shootout in which case I don’t know if either team will be a great pick, but the models don’t lie and since Newton and Rivers both get sacked a lot there’s a good chance that the defenses will still rack up enough sacks and turnovers to score well.

The next (group? cluster?) are middling defenses with middling matchups. ARI is a decent choice (top 10 in points up against Miami who have give up the 14th most points) as is Detroit (16th in points against CHI who is 11th is points against). The rest of this group could really go any way. Model A likes ATL and TB a little more than Model F does, and Model F likes WAS and IND a little more than Model A. Personally, I wouldn’t be excited about either. Sand Francisco’s defense has scored some of the fewest fantasy points but they’re up against NYJ, currently giving up the most points per defense. According to ESPN the Browns, Jets, and Rams have all given up more than 10 fantasy points per game to defenses. That makes ATL a potentially nice choice this week as well (against LA), but they’re very inconsistent this year as well.

Whatever you decide this week, I wish you luck. I’m out of my fantasy playoffs, but I’ll continue this through the rest of the year despite the fact that it no longer benefits me. I’m just that benevolent. And it’s fun.

Week 13 K Evaluation

Week 13 DEF Evaluation