Data scientist, physicist, and fantasy football champion

Week 2 DEF Predictions

Week 1 came and went, and fantasy defenses all crushed. It was an especially high scoring week for defenses, and as such some of my models have poor accuracy scores, though they’re all just shifted together as a group because some of the best defenses went unpredicted. Last week I recommended (and ultimately played) BUF, and they ended up in the middle of the pack… with 11 points! That’s a solid day for a fantasy defense, but last week it was only good enough to tie for 10th.

Model C was the best last week, then Yahoo’s experts, then Model F. I still can’t run Model A since I don’t have any data for MIA and TB. I was really surprised how poorly the Gibbs sampler did. Maybe this week it will do better.

I’m trying to rush through these a little so I can get it done before waivers clear Tuesday night or Wednesday morning. Enough of the intro, on to the analysis:

 

ModelTypeScoreOppScoreOppHomeAwayDataYearsPointSystem
ALMXXXX2017IOUH/Standard
CLMXXXX2015-2017IOUH/Standard
FLMXXXX2015-2017TruScor
GGibbs SamplerXXXX2015-2017TruScor


The models are pretty homogenous this year. They all include the same terms but with different data, point systems, and methods of statistical inference. These models consistently beat the other ones that included other terms, so until I come up with new data to predict with this is about it for now.

On to the predictions:

Model A:


Coming soon (when I actually have enough 2017 data, MIA and TB haven’t played yet.)

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 2017
Standard points
Did well in 2016, but didn’t stand out


Model C:

.Model C-1.png


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 2015-2017
Standard scoring
Did well in 2016, but didn’t stand out


Model F:

.Model F-1.png


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 2015-2017
Truscor ignores defense TDs but weights interceptions and fumbles to try to account for it
I really liked this one in 2016. It was marginally better than A and C


Model G

.Model G-1.png


 Model G 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 G Notes:

Uses data from 2015-2017
Truscor ignores defense TDs but weights interceptions and fumbles to try to account for it
Uses Bayesian statistics (Gibbs sampler) to create a model
Would have done very well in 2016 based on simulation


Week 2 model summary

RankACFG
1NAARIOAKOAK
2NAOAKARIARI
3NACARCARCAR
4NABALBALBAL
5NATBSEASEA
6NAKCTBTB
7NALARDALDAL
8NASEAKCPIT
9NATENCINCIN
10NADALPITKC
11NAPHITENTEN
12NANYGLACLAC
13NAJACLARNYG
14NACINNYGLAR
15NALACHOUHOU
16NADENINDDEN
17NAWASPHIMIA
18NAPITDENPHI
19NAMIAMIAWAS
20NAHOUWASIND
21NAINDJACJAC
22NAMINBUFNE
23NABUFCLEBUF
24NACLEMINCLE
25NAATLGBGB
26NADETNEMIN
27NAGBCHIATL
28NACHIDETDET
29NASFATLCHI
30NANESFSF
31NANONONO
32NANYJNYJNYJ

 


Conclusions


All of my models say the same thing: we’ve got a big week coming up for ARI, OAK, BAL, and CAR. Not surprising, since there are teams heavily favored to win and keep their opponents scores down. ARI, OAK, BAL, and CAR are up against IND, NYJ, CLE, and BUF this week, and at least those first three are very juicy fantasy defense matchups. I’d be surprised if at least one of these wasn’t available in your league. I’d be only a little more cautious to play ARI if if Brissett gets the start, but ARI has a good defense and if it’s Tolzien at the helm I’d fire them up with all the confidence in the world. Last year when Brissett started 2 games for NE he didn’t throw any interceptions, but also didn’t throw and touchdowns. NE scores 27 points against HOU and 0 against BUF. Probably still safe to play ARI, but there’s the smallest chance it could end poorly for you.

The next grouping is where my models start to diverge. TB, KC, LAR, SEA, DAL, and TEN all show up here, with PIT and CIN a little higher or lower depending on the model.

JAC is in the mix, as are PHI and DET, but I don’t know if I’d expect the same performance we saw last week. JAC vs. HOU was a monumentally bad game, and Mariota and the Titans stand to be a much tougher offense. LAR crushed IND, but unless they get to play them again (spoiler alert: they don’t) I don’t think they’ll have quite as good a game against Cousins and WAS. But if they’re good enough for Model C they’re good enough for me.

I’m targeting one of those top 4 for the week enough to actually put in a waiver claim for them. Two are available in my league. Don’t waste your waiver order on fantasy defenses. If I can’t get one of those, I’ll save the claim and see whoever is left on Wednesday morning. I’d suggest you do the same.

Personal update

I made the rest of the people in my league look foolish last week. I had the most points overall, 25% higher than the next player, doubling about half of the players, and almost tripling the worst performer. I won’t say it’s because I played Buffalo, but I will say that it’s partially because I’m a brilliant and logical analyst and partially because the dummies in my league just make it so easy.

Up next: QBs and, if I have no self esteem left, kickers.

Week 2 QB Predictions

Week 1 K Results