Tuesday, January 1, 2013

2012 NFL Final Regular Season Rankings

Since September 29 of this year, I have posted my betting market rankings for the NFL on a daily basis.  The purpose of these rankings is to determine who the betting market thinks are the best and worst teams in the league (see the Methodology page for more background).  Publishing these rankings also afforded me the opportunity to indulge in one of my current interests: data visualization.  My goal was to produce as "data-rich" a ranking table as possible that was still relatively simple to understand.

Now that the regular season is over, I am finalizing these rankings, with plans to now focus on the playoffs.  But before I move on to simulations and probabilities of advancing to each round, I thought I'd take a look at how each team performed from a win-loss perspective, compared to the implied betting market expectations.

Wins, Losses, and Luck

In mid-October, I added a Projected Wins column to the ranking table in which I simulated the remainder of each team's regular season and calculated their average projected wins (the win distribution was also provided in a bite-sized bar chart).

With all 256 games now in the can, there is nothing more to project.  But what I can do, is "replay" the season several thousand times, simulating wins and losses according to my final GPF (Generic Points Favored) rankings.  We can then see how each team's actual wins compare to the simulation average.  The difference between the two can be viewed as a measure of how "lucky" each team was.  I'm not the first to try something like this; you can find similar analyses at Advanced NFL Stats for prior seasons (2008 season and 2006 season).

Refer to the "Simulated Wins" and "Luck" column in the updated rankings table.  Here are the three luckiest teams this season, and how many wins could be attributed to "luck":

  1. Indianapolis - 4.2 lucky wins
  2. Atlanta - 3.3 lucky wins
  3. Minnesota - 2.7 lucky wins
And here are the three unluckiest teams:
  1. Detroit - 4.0 unlucky losses
  2. Philadelphia - 3.1 unlucky losses
  3. Kansas City - 3.0 unlucky losses

One of Many Possible Timelines

In the Simulated Wins column, I added a new bar chart that shows the win distribution summarized over the 10,000 simulation runs.  The red bar corresponds to the actual wins for that team this season.  For example, here is Miami's chart:


As you can see from the red bar, the Dolphins' seven wins were right at expectations. What is striking though is how wide the distribution of projected wins is.  In 8% of the simulations, the Dolphins win at least ten games and would have most likely been playoff bound.  And though each team's distribution varies somewhat, there appears to be about a 3% chance that any given team will outperform expectations by about 4 wins, or underperform with 4 extra losses.  In a 32 team league, that should happen about once a season (see the Colts and the Lions this year).

Despite our instinct to build narratives around each team's win-loss record (an instinct that is in overdrive right now for the media and the TV-movie ready Colts), a sixteen game season in a sport with a lot of high leverage plays means that there is a lot of randomness in each team's final record.  The NFL is not a coin flip league, but it is much closer to that than one would expect.

Additional Notes

These rankings will not be updated anymore and will remain "frozen" into place.  For the purposes of calculating the final rankings, I excluded the week 17 games as the point spreads are most likely distorted due to some teams having little or nothing to play for.

The Strength of Schedule column is now final.  The Colts have been lucky in many ways this season.  Not only do they have the luckiest win-loss record, they were able to draw the easiest schedule in the league.  I imagine Ken Whisenhunt could still have a job right now if the Cardinals (with the toughest schedule) got to play the Titans and Jaguars twice this season, as opposed to the Seahawks and the 49ers.  

Going by pageviews, my daily NFL rankings have been the most popular feature by far on this site.  I'm glad they found an audience and am looking forward to covering the 2013 regular season.

Also note that despite all this talk of Colts and luck, I managed to avoid any hacky references to the surname of a particular rookie quarterback out of Stanford.

1 comment:

  1. I really like those red bars. They're very illustrative, and make a good point.

    ReplyDelete