Data scientist, physicist, and fantasy football champion

Week 8 QB Predictions

I don’t really have anything for a preamble this week. We’re halfway through th eseason and Model H has been the best so far, though it’s not statistically significantly better than Model D or G. It’s probably not a coincidence that those 3 models use similar terms, but G and H use only 2017 data while D uses 2015-2017 data. I’m not sure which to go with if I were you. But as we’ll see in a second, I don’t want to believe H because it means I’m screwed this week.

Here are the models for this week:

Model Type Player Score OppScore ScoreOppScore HomeAway HomeAwayPlayer Opp DataYears
A LM X X X X X X 2015-2017
B LM X X X X X X X 2015-2017
C LM X X X 2015-2017
D Gibbs Sampler X X X X 2015-2017
E Gibbs Sampler X X X X X X 2015-2017
F PLS X X X X X X 2015-2017
G PLS X X X X X X 2017
H LM X X X X X 2017
I Many LMs X X X X X X 2015-2017
J Many LMs X X X X X X 2016-2017*

 

Model A:

.Model A-1.png

Terms:

Player
Score
OppScore
Score*OppScore
Home/Away
Home/Away*Player (Some players have a bit more of a problem playing at home or away. Looking at you, Roethlisberger…)
Data years: 2015 - 2017

Model B:

.Model B-1.png

Terms:

Player
Score
OppScore
Score*OppScore
Home/Away
Home/Away*Player
Opponent
Data years: 2015 - 2017

Model C:

.Model C-1.png

Terms:

Player
Score
OppScore
Data years: 2015 - 2017

Model D:

.Model D-1.png

Terms:

Player
Score
OppScore
Opponent
Data years: 2015 - 2017

Bayesian (Gibbs sampler)

Model E:

.Model E-1.png

Terms:

Player
Score
OppScore
Score*OppScore
Home/Away
Home/Away*Player
Opponent
Data years: 2015 - 2017

Bayesian (Gibbs sampler)

Model F:

.Model F-1.png

Terms:

Player
Score
OppScore
Score*OppScore
Score^2
OppScore^2
Home/Away
day (Thu/Sun/Mon)
Data years: 2015 - 2017

Model details: Partial least squares fit

Model G:

.Model G-1.png

Terms:

Player
Score
OppScore
Score*OppScore
Home/Away
Opp
Data years: 2017 only

Model details: Partial least squares fit

Model H:

.Model H-1.png

Terms:

Player
Score
OppScore
Home/Away
Opp
Data years: 2017 only

Model details: Linear model

Model I:

.Model I-1.png

Terms may include:

Player
Score
OppScore
Score*OppScore
Home/Away
Home/Away*Player
Opponent
Data years: 2015 - 2017

Model details: This model tries to predict all of the components of scoring separately (pass yards, pass TD, rush yards, interceptions, sacks, etc.). Each component comes from a linear model containing only the significant terms (p < 0.05) from the list of possible terms above

Model J:

.Model J-1.png

Terms may include:

Player
Score
OppScore
Score*OppScore
Home/Away
Home/Away*Player
Opponent
Data years: 2016-2017 for most people, 2017 only for Alex Smith and Jared Goff

Model details: This model tries to predict all of the components of scoring separately (pass yards, pass TD, rush yards, interceptions, sacks, etc.). Each component comes from a linear model containing only the significant terms (p < 0.05) from the list of possible terms above. I also modified the data years; 2016 and 2017 for most players, 2017 only for Alex Smith and Jared Goff since there’s been so much improvement since last year.

Weekly model summary
 

Conclusions

[NOTES: I think I have all the right QBs in this week. Last week I didn’t model Hundley, this week Cutler and Palmer are both out, but their backups are known quantities.]

Because the dummies in my league LOVE their QBs, I’ve been trying to pawn off Matt Stafford for the last week for any useable WR. No takers. Now I’m stuck with him for this week and let’s see what Model H says… crap. Not dead last, but hanging out with Brissett and Trubisky. Maybe this will finally be the week where I drop him. I’ll make one last attempt at a trade, but I can’t play him this week.

Can I?

Shit.

Let’s look at the top streaming options this week:

  • Case Keenum (At CLE): He could honestly be a good option against Cleveland this week. Model H loves him, D and G put him in the middle of the pack. Reasonably high score (23.5 pts) against a bad defense. My only worry is that they might get out to a big lead and just run out the clock, but Minnesota hasn’t really been blowing out other teams this year. I might target his this week if I can unload Stafford.
  • Tyrod Taylor (at home vs. OAK): He had a great week last week against the terrible Buccaneers and is likely to do so again against the mediocre Raiders. Last week was the most he’s thrown this year and the 2nd most he’s rushed. He’s been a solid flex play more often than not this year. Model D loves him, G and H like him a lot as well.
  • Andy Dalton (at home vs. IND): Indiana has given up a lot of points. Dalton is having a rough year given what I had hoped from him, but he’s actually put together mostly fantasy-useable weeks.

And one probably-stay-away:

  • Matt Moore (at Baltimore): I really like Matt Moore. My models don’t because he had 2 bad games in 2015, but the 4 games he’s had since then have all been good (middle of the pack, but again, for a streamer that’s useable). Model H would like him more, but the points he scored were against the Jets, and therefore count less. Still, going up against Baltimore is tough for any QB. I’d love to be wrong about him, but we’ll see.

Siemian is coming off a really rough game against the Chargers, but gets Kansas City next. If I run out of options, I’ll probably grab him last.

If anyone is looking for a shiny, barely-used Matt Stafford, I’ve got one for you.

Week 8 DEF Predictions

Week 8 K Predictions