Saturday, September 8, 2012

NFL Team Rankings - Week 1

Dolphin-trashcan-july09Waaaaaaay back in May, I published my first look at the 2012 Betting Market rankings for the NFL, using opening lines courtesy of Cantor Gaming.  With Opening Sunday nearly upon us, I can take an updated look at the rankings using the latest betting lines for both Week 1 and Week 2.  The rankings published back in May were, in effect, just reverse engineered rankings according to the experts at Cantor Gaming.  I can now see where the actual money is flowing to get an updated view.

See the Methodology page for a simplified example of my ranking approach.  For more detail, you can take a look at my first post on this topic at Advanced NFL Stats Community.

As usual, my source for betting information is Killer Sports.  They have NFL lines updated for both Week 1 and Week 2 of the season.  Since my methodology usually needs at least three weeks of data to get a proper set of rankings, I am using the Cantor lines from my May post to "fill in the gaps", so to speak (more technically, my ranking is a weighted linear regression, with weeks 1 and 2 -the real money- weighted at 1, and weeks 3-16 -the Cantor lines- weighted much smaller at 0.0000001).

See the table below for the updated rankings.  Here is a glossary of terms:

  • GPF - Stands for "Generic Points Favored".  It is what you would expect a team to be favored by against a league average opponent at a neutral site.
  • Opened - The Cantor Gaming-based GPF from my May post.
  • Change - How much each team's ranking changed (in terms of GPF) from the Cantor ranking to the updated "live money" version
Some observations:
  • If the nation's gamblers were watching Hard Knocks, they must have not liked what they saw.  The Miami Dolphins moved the most by far, dropping 5 points to the bottom of the rankings.
  • Fellow Hard Knocks alum, the New York Jets, did not fare so well either, dropping two points.
  • On the positive side, Seattle, St. Louis, Chicago, and Atlanta all improved by about 2 points.
  • The Packers and the Patriots still sit atop the rankings, although the gap between those two and the rest of the league has shrunk somewhat.

The Rankings


  1. I'm new to your site and was wondering how would you convert you results to predict current week lines?

  2. Sure. To calculate an expected point spread for a given matchup, you take the difference of each team's GPF and then add 2.5 points advantage to the home team.

    For example, in week 3, the Giants play at Carolina. The Giants' GPF is 1.5 and the Panthers' GPF is 0.7. Predicted point spread = 0.7 - 1.5 + 2.5 = 1.7. So, the rankings above would expect the Panthers to be favored by about 1.5 points.

  3. Thanks for the detailed explanation. Really curious how your ranking system could be leveraged with a tradition one based on offense and defense. Im applying your weekly updates to my model to see how they canbe used together in the future. One question, last yr home field advantage was roughly 3.65. Your 2.5 is derived from how many years worth of data. The only reason I ask is due to the fact the league has become more of a spread and attack style. Which seems to garner more pts on a week to weekbasis.

    1. One more thing, sorry to be a pain, how often do you update your rankings?

    2. I will update before the Week 2 games, but my eventual goal is to update daily via the Ticker feature.

      Thanks for the interest shown in the site. I would be interested to see any results you can come up with using the metrics here.

  4. At a high level, you can view these rankings as "competitive intelligence". If you're betting on football, you're competition is the point spread and these rankings are an attempt to understand the dynamics driving the point spread. More specifically, there is my Today's Games feature which calculates an "expected" point spread and then compares it to the actual point spread. The idea being that significant deviations are indications of a late-breaking development that may not be reflected in a stat-based model (e.g. injury to a starting quarterback). The feature is available for Baseball for now, but I plan on adding one for the NFL.

    On your home field advantage question, I'm assuming that 3.65 points is the average actual scoring differential for home teams last year? If you look at this differential over time, it bounces around quite a bit, so I don't think you can read too much into the high number last year. For example, it was 1.7 points in 2010, 2.4 points in 2009, and 3.5 points way back in 1998. But my 2.5 points is not based on actual scores but on the point spreads themselves. The actual differential may jump around year to year, but Vegas still seems to think that home field is worth about 2.5 points.