Data scientist, physicist, and fantasy football champion

Week 12 DEF Predictions

I’m keeping the same five models this week as I did last week. Models A, B, C, and E did well last week while Model D did not. I’ve run A through C more than once and I expect that level of accuracy, but let’s see what happens with D and E two weeks in a row.

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

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

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:

  • New model this week
  • Uses only data from weeks 5 through 10 (inclusive) for 2016
  • What have you done for me lately?

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:

  • New model this week
  • Uses data for all of 2016 (through week 10)
  • includes week + Team:week and Opponent:week terms for each team

Week 12 model summary

Rank

Model.A

Model.B

Model.C

Model.D

Model.E

1

ATL

PHI

KC

PHI

PIT

2

PHI

BUF

NYG

BUF

ATL

3

BUF

ATL

DEN

HOU

SD

4

HOU

HOU

BUF

NYG

BUF

5

KC

SD

PHI

ATL

BAL

6

NE

NE

PIT

BAL

KC

7

SD

KC

ATL

NO

TB

8

DEN

MIA

NE

SD

MIA

9

CHI

PIT

SEA

KC

DAL

10

NO

NO

HOU

PIT

NYJ

11

PIT

NYG

OAK

MIA

HOU

12

NYG

CHI

BAL

SEA

CLE

13

BAL

DEN

NO

DET

PHI

14

MIN

MIN

SD

MIN

CAR

15

MIA

BAL

CHI

TB

MIN

16

SEA

SEA

TEN

GB

DET

17

OAK

OAK

MIA

CHI

TEN

18

SF

SF

MIN

DEN

DEN

19

GB

TEN

DET

NE

CHI

20

DET

DET

TB

LA

LA

21

CIN

LA

ARI

CIN

SEA

22

TEN

CIN

GB

OAK

NE

23

LA

GB

DAL

TEN

OAK

24

TB

TB

WAS

CAR

CIN

25

ARI

ARI

CAR

JAC

GB

26

CAR

CAR

LA

SF

NYG

27

DAL

DAL

CIN

DAL

ARI

28

CLE

CLE

SF

IND

JAC

29

WAS

WAS

JAC

ARI

NO

30

JAC

JAC

IND

CLE

IND

31

NYJ

NYJ

CLE

WAS

WAS

32

IND

IND

NYJ

NYJ

SF

I’m seeing a lot of ATL, PHI, BUF, KC, and HOU in models A, B, and C. PHI, BUF, and KC aren’t really surprising to me but HOU and ATL are somewhat. HOU has had a tough time in the last few weeks with only 1 game above 5 points in their last 6 games. They’re going up against SD who have given up the 6th most points (according to Yahoo), so that’s a point in their favor. ATL is going against ARI who have given up a surprising number of interceptions and sacks this year. They’re not an especially strong defense (fantasy-wise), but If they can’t get 2 sacks and 1 interception out of Palmer and the Cardinals this year I’d be a little surprised.

The Giants are another team to consider this week. If Pittsburgh’s performance last week taught me anything it’s to never bet against betting against the Browns this year. Rounding out the top tier in Model B is San Diego, which is another fine choice.

Joey Bosa, so how right now. Joey Bosa.

Those are all available in my league, so I’m going to pick up any of those if I can this week. You should do the same (unless you’re in my league, in which case just go ahead and fail)

Thoughts for future analysis

Stop reading here if all you want to know are predictions for week 12. I’m going to let you all into my mind for a few minutes while I think about improving accuracy.

I’m at an accuracy impasse. I tried adding combinations of the variables that I have including interactions (i.e., expected score * Team, Home/Away * Opponent, etc.) to models B and C but it didn’t improve my accuracy score for week 11. That may mean that just week 11 is weird, but it’s more likely that it’s a sign that adding new terms like that would be overfitting. If that’s the case, then I need to come up with a way to evaluate teams other than just linear modeling.

I’ll play around a little this week if my family doesn’t mind my ignoring them at Thanksgiving. I may try to find a way to account for TDs other than just with when a team has scored TDs in the past. A quick fit to TDs vs Fumbles and TDs vs Interceptions looks like a team has about a 50% chance of scoring a TD for every 4 fumbles and 50% chance of a TD for every 3 interceptions. The problem is that predicting fumbles and interceptions is only a little more possible than predicting TDs. Maybe I can create a new point system where a fumble is worth 2 points + (0.25*6) which would be the fantasy score plus the chance of converting a fumble into a TD. I could model fumbles and interceptions, but they’ll be dependent on team and opponent like the rest of the model so the factors will be very strongly confounded. Still, worth a shot. I refuse to believe that I’ve reached the limit of accuracy already.

Week 11 K Evaluation

Week 11 DEF Evaluation