Wednesday, September 11, 2013

NFL Week 2 Rankings

Deux jaguars du Pérou
Deux jaguars du Pérou
Here is your weekly check in on how the betting market ranks the NFL teams. With opening lines available for week 3 NFL games, I can now dispense with the Cantor Gaming lines to fill in the gaps in my regression analysis. The rankings below take the point spreads that Vegas sets for each NFL game and uses them to reverse engineer an implied team ranking. See here and here for more background.

The Jaguars Limbo Under a Low Bar

The Jaguars (somehow) found a way to be even worse than expected. A four point underdog against the Chiefs (at home, no less), and they lose by 26. Prior to their week 1 loss, the Jaguars were a 3 point underdog on the road against the Raiders in week 2. The market has since corrected, and the Jaguars are now a six point underdog, although part of that movement could be due to a perceived better than expected showing by the Raiders against the Colts.

The Jaguars were already ranked the worst team in my week 1 rankings, but their GPF (Generic Points Favored), dropped even further this week, from 7.6 to 8.9. They drop any lower and they rival the post-Manning, pre-Luck, Curtis Painter-led Colts of 2011 in terms of sheer numerical awfulness. An unfortunate start for the Jaguars new, supposedly more analytical, approach.

Here is the Week 2 ranking table. The Broncos remain on top. The two Pennsylvania-based NFL teams were like ships passing in the night, with the Eagles jumping from #18 to #11, and the Steelers sinking from #14 to #24.

Rank Team Week 1 GPF
1  DEN 1 7.2
2  SF 3 6.4
3  SEA 2 6.0
4  GB 6 5.5
5  NE 4 4.7
6  NO 5 3.8
7  HOU 9 3.4
8  ATL 7 2.4
9  CHI 8 2.3
10  CIN 11 2.0
11  PHI 18 1.9
12  WAS 16 0.8
13  DET 10 0.5
14  BAL 13 0.1
15  CAR 22 0.1
16  DAL 12 -0.1
17  IND 19 -0.1
18  KC 24 -0.2
19  NYG 15 -0.3
20  MIA 23 -0.4
21  TB 21 -1.5
22  STL 17 -1.8
23  MIN 20 -2.1
24  PIT 14 -2.4
25  TEN 28 -2.7
26  ARI 25 -2.9
27  SD 27 -3.4
28  CLE 26 -3.5
29  NYJ 29 -5.2
30  BUF 30 -5.6
31  OAK 31 -6.0
32  JAC 32 -8.9


  1. I'm a little surprised that I haven't seen more talk about the Giants who managed to stay within a single score despite the spectacular -5 turnover differential. Though they might feature later if you revisit the Turnover Index.

    Filling out my prediction sheet, I also noticed that the over/under for the Eagles-Chargers game has climbed from its 51 open to (more or less) 55.

    (The 'ATS Model' simply looks at how often the home team and away team have covered - or failed to - since 2001, and then picks if there's a 'significant' net history.)

    HFA Model - 5-4-1
    O/U Model - 2-0-0
    ATS Model - 2-5-0

    Picks for Week 2:

    HFA Model:
    Falcons -7
    Bears -7
    Ravens -7
    Saints -3.5
    Jaguars +6
    Broncos -5
    Seahawks -3
    Steelers +7

    O/U Model:
    Dolphins v Colts Under 42.5
    Chargers v Eagles Over 54.5
    Lions v Cardinals Over 47.5
    Broncos v Giants Over 54.5
    Steelers v Bengals Under 40.5

    ATS Model:
    Panthers -3
    Dolphins +3
    Ravens -7
    Saints -3.5
    Jaguars +6
    Seahawks -3
    Steelers +7

  2. This is just one guy's opinion, but: Nate, start your own own blog if you just want to post picks from a model that has nothing to do with this site or this post. I might suggest or for tracking system performance.

  3. It's alright to me as long as if 1) Nate had seen impressive results from backtesting his model (I know, that doesn't guarantee future success) 2) the comments in this website are not cluttered by people who post about their experiments.

  4. The reason that I was doing stuff here is because the hypothesis that NFL spreads misjudge teams' home field advantage was inspired by the Turnover Index stuff that Micheal was publishing here (e.g.

    I think the NFLstatheads sub-Reddit may be a better place for my musings than here.

  5. I don't have a strong opinion on it. Nate - your line predictions may be better suited for pickmonitor or nflstatheads. But comments on methodology/results/etc. are always welcome (and encouraged).

    1. No worries, I'll be sure to comment when I have something apropos to write. For example:
      As an alternative to filling in the blanks with the Cantor data, have you considered incorporating last season's results (as ~52 weeks ago) into the rating scheme to see if it improves the ratings? Brian Burke at ANS suggests that there is some season-to-season carryover in team strength.

    2. I imagine that approach would work as well, as I'm sure there is correlation between seasons. But it feels like I would be adding less information that way, as the regression could be done as soon as the Superbowl finishes. At least with the Cantor lines, I get an estimate of what the offseason trades, draft picks, and free agent acquisitions are worth.