Tuesday, May 15, 2012

2012 NFL Opening Lines - An Analysis

The 2012 NFL regular season is still 113 days away, but you can start making bets now. Point spreads for Week 1 of the NFL season are now available, courtesy of Football Locks. The purpose of this post is to review how these point spreads compare to my 2011 NFL point spread rankings. This will give us an early read on who the betting market thinks the off-season winners and losers have been so far.

Here is some background on how the rankings were developed.  The Methodology page also has a simplified example.

For the purpose of this analysis, I am going to use the 2011 Week 16 rankings I published at the Advanced NFL Stats Community site. The table below compares the predicted point spread for each game to the actual point spread. The bar in each row of the table is proportional to the magnitude of the miss and points in the direction of the team that the point spread miss favored (similar to the currently dormant Today's Games feature).  A negative point spread indicates the home team is favored (a convention that has always confused me).


2012 Week 1 Opening Lines


Some observations:

  • Atlanta at Kansas City - This was the biggest miss, with a 6.5 point swing in favor of Kansas City.  It's not too surprising, as the rankings as of week 16 last year reflected a Chiefs team that had lost its star running back, starting quarterback, and head coach.  The Chiefs added Peyton Hillis to the roster and they'll have Matt Cassel and Jamaal Charles back.  But 6.5 points seems like a lot.
  • Miami at Houston - The Texans were missing Matt Schaub as of week 16 last year, which probably explains a significant part of the 5.5 point miss.
  • Pittsburgh at Denver - The Broncos have clearly improved in the eyes of the betting market, most likely due to their signing of a high profile former Colt. I refer, of course, to Jacob Tamme.
  • Washington at New Orleans - I was surprised by how small the miss was here. The Saints are without head coach Sean Payton, defensive coordinator Greg Williams, and linebacker Jonathan Vilma. The Redskins added Robert Griffin III. But the Saints still have Drew Brees, which is probably why they're still a 9.5 point favorite.
  • Buffalo at New York (Jets) - This was a four point swing in favor of Buffalo. Is it because the market expects the Bills to improve? Or they expect the Jets to regress? Probably a little bit of both.


Home Field Advantage and Pre-season Uncertainty
Another thing that emerges clearly from the table is that the swings in the point spread are definitely biased towards the home team. I think this reflects the market's expectation that teams will regress to the mean (as demonstrated here and here). 

Team performance and strength can vary wildly from year to year, but you can always count on home field advantage. If you knew nothing about any of these teams or the moves that they made in the off-season, you would set the point spread at 2.5 points in favor of every home team. The bias shown above in favor of the home team is in fact an admission of ignorance.

Nobody really knows how each team's draft prospects and free agent signings will work out. In the face of that uncertainty, you fall back on what you know, which is that home teams tend to win more often than not.


5 comments:

  1. Just in case you didn't know, Cantor has lines for week 1-16 already. http://www.sportsmemo.com/blogs/view/?name=Cantor-Gaming-releases-odds-for-entire-NFL-season&blog_id=7209

    ReplyDelete
  2. Perhaps you'd want to use only the lines for week 1-3. Just a guess but perhaps the further the season, the more conservative the lines are to factor in the probability of starting QBs getting injured, et cetera.

    ReplyDelete
  3. Oops, wrong link for the 2012 lines. The correct one: http://beyondthebets.com/2012/05/09/cantor-gaming-plans-to-release-point-spreads-for-all-256-nfl-2012-regular-season-games/

    ReplyDelete
  4. Thanks for that. This is fantastic.

    I agree that it would make sense to use weeks 1-3 to derive a ranking. But I'd be interested to see if your hypothesis is correct. I think it would be pretty straightforward to test.

    I think I'll write it up the results in a separate post.

    ReplyDelete
  5. Great idea. Looking forward to it!

    ReplyDelete