Data scientist, physicist, and fantasy football champion

Week 3 QB Predictions

Week 2 was a bit more like we all expected, with Brady, Brees, Rodgers, Roethlisberger, and Carr all in the top 10. Siemian was unexpected, but maybe he’s finally coming into his own as a QB. Or maybe he’ll regress. My models literally have to predict that, so it’ll be a long time before they put they put him toward the top of any lists. Still, I’ll keep an eye out for him.

Let’s see the models this week:

 

ModelTypePlayerScoreOppScoreScoreOppScoreHomeAwayHomeAwayPlayerOppDataYears
ALMXXXXXX 2015-2017
BLMXXXXXXX2015-2017
CLMXXX 2015-2017
DGibbs SamplerXXX X2015-2017
EGibbs SamplerXXX XXX2015-2017
FPLSXXXXX X2015-2017

 

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
Home Team (which stadium they’re in)
Opp
Team (some QBs play better under different teams)
day (Thu/Sun/Mon)
Data years: 2015 - 2017

Model details: Partial Least Squares fit using the caret package. I’m just throwing data science at the wall here and seeing what sticks. I need to work on variable selection, and if you’re reading this note past week 1 it’s because I still need to work on variable selection. For now, just consider Model F to be experimental.

Weekly model summary

RankABCDEF
1Aaron RodgersAaron RodgersAaron RodgersCam NewtonCam NewtonAaron Rodgers
2Cam NewtonCam NewtonTom BradyAaron RodgersAaron RodgersDak Prescott
3Kirk CousinsDerek CarrCam NewtonKirk CousinsKirk CousinsCam Newton
4Derek CarrKirk CousinsKirk CousinsDerek CarrDerek CarrTom Brady
5Tom BradyTom BradyDerek CarrTom BradyTom BradyKirk Cousins
6Matthew StaffordMatt RyanBen RoethlisbergerMatt RyanMatt RyanDerek Carr
7Matt RyanDeshaun WatsonDak PrescottBen RoethlisbergerMatthew StaffordBen Roethlisberger
8Jameis WinstonJameis WinstonMatthew StaffordRussell WilsonRussell WilsonMatthew Stafford
9Deshaun WatsonMatthew StaffordDrew BreesMatthew StaffordJameis WinstonMatt Ryan
10Andy DaltonRussell WilsonMatt RyanDak PrescottCarson PalmerDrew Brees
11Carson PalmerAndy DaltonRussell WilsonDrew BreesAndy DaltonMarcus Mariota
12Marcus MariotaMarcus MariotaMarcus MariotaPhilip RiversMarcus MariotaCarson Palmer
13Dak PrescottCarson PalmerJameis WinstonMarcus MariotaPhilip RiversRussell Wilson
14Russell WilsonBlake BortlesPhilip RiversJameis WinstonBen RoethlisbergerJameis Winston
15Carson WentzDak PrescottBlake BortlesBlake BortlesBlake BortlesPhilip Rivers
16Philip RiversPhilip RiversAlex SmithCarson PalmerDak PrescottAlex Smith
17Sam BradfordJay CutlerTyrod TaylorAlex SmithCarson WentzBlake Bortles
18Ben RoethlisbergerSam BradfordEli ManningEli ManningSam BradfordJoe Flacco
19Blake BortlesCarson WentzJoe FlaccoAndy DaltonDrew BreesTyrod Taylor
20Jay CutlerBen RoethlisbergerCarson PalmerCarson WentzAlex SmithSam Bradford
21Drew BreesDrew BreesAndy DaltonSam BradfordJay CutlerEli Manning
22Tyrod TaylorJosh McCownSam BradfordJay CutlerEli ManningJay Cutler
23Alex SmithEli ManningJay CutlerJoe FlaccoDeshaun WatsonAndy Dalton
24Eli ManningAlex SmithCarson WentzTyrod TaylorTyrod TaylorCarson Wentz
25Trevor SiemianTyrod TaylorBrian HoyerTrevor SiemianTrevor SiemianTrevor Siemian
26Joe FlaccoBrian HoyerTrevor SiemianBrian HoyerJosh McCownBrian Hoyer
27Brian HoyerTrevor SiemianJosh McCownJosh McCownJoe FlaccoJosh McCown
28Josh McCownJacoby BrissettDeshaun WatsonDeshaun WatsonJacoby BrissettJared Goff
29Jared GoffJared GoffDeShone KizerMike GlennonBrian HoyerDeShone Kizer
30Jacoby BrissettJoe FlaccoMike GlennonJacoby BrissettJared GoffMike Glennon
31DeShone KizerMike GlennonJared GoffDeShone KizerMike GlennonJacoby Brissett
32Mike GlennonDeShone KizerJacoby BrissettJared GoffDeShone KizerDeshaun Watson

 

Conclusions

Model F was good the last two weeks. We’ll start with that one and see what’s there.

There are really no surprises in the top 10 QBs, but if you have Newton, Cousins, Stafford, or Roethlisberger there’s a chance you’re considering streaming from week to week. I’d fire them all up with confidence this week in good matchups (NO, OAK, ATL, and CHI, respectively).

Model F really doesn’t like Wentz or Siemian despite good performances from them this year. I actually wouldn’t mind playing one of them this week; Wentz is a tougher sell against the Giants, but both BUF and NYG give up mid-level points to QBs.

Model F also doesn’t care for Cutler or Dalton, likely from sub-par performances from them. Dalton should be playing from behind most of the game and Cutler is up against the Jets, so there’s something to like with both of those guys. The other models like them both a lot more this week, with the next-most accurate models putting Dalton up there in streaming territory, so don’t just toss them out yet.

Everyone else in the middle seems like a reasonable play, and QB fantasy points all tend to cluster around the middle. Arizona is predicted to score pretty low this week, but they’re up against Dallas, so Palmer is a streaming option despite lackluster performances in weeks 1 and 2. This model puts Mariota little higher than I would given that TEN is up against SEA this week and they give up some of the fewest points to fantasy QBs. Remember what we learned about Mariota going up against tough defenses

I have Stafford, so I’ll likely be playing Stafford. But I’ll be keeping an eye on Siemian this week. It might be worth it to look into a model with a shorter memory, or at least one with a less intense memory. This type of flexibility is basically what Bayesian models are designed for. I just have to learn how to encode it.

Week 2 K Results

Week 2 QB Results