Wednesday, January 8, 2014

Year to year volatility in NFL standings

Question:   To what extent was team record in the NFL in 2012 related to team record in the NFL in 2013?

Short Answer:  There was a fair amount of change in the performance of many teams from the 2012 to 2013 seasons.

Data:   The raw data used for this analysis is presented below.

 Team 2012 2013 2012 minus 2013 AT 0.813 0.25 0.563 AZ 0.313 0.625 -0.312 BA 0.625 0.5 0.125 BU 0.375 0.375 0 CA 0.438 0.75 -0.312 CH 0.625 0.5 0.125 CI 0.625 0.688 -0.063 CL 0.313 0.25 0.063 DA 0.5 0.5 0 DE 0.813 0.438 0.375 DR 0.25 0.813 -0.563 GB 0.688 0.531 0.157 HO 0.75 0.125 0.625 IN 0.688 0.688 0 JA 0.125 0.25 -0.125 KC 0.125 0.688 -0.563 MI 0.438 0.5 -0.062 MN 0.625 0.344 0.281 NE 0.75 0.75 0 NO 0.438 0.688 -0.25 NYG 0.563 0.438 0.125 NYJ 0.375 0.5 -0.125 OA 0.25 0.25 0 PH 0.25 0.625 -0.375 PI 0.5 0.5 0 SD 0.438 0.563 -0.125 SE 0.688 0.813 -0.125 SF 0.719 0.75 -0.031 SL 0.469 0.438 0.031 TB 0.438 0.25 0.188 TE 0.375 0.438 -0.063 WA 0.625 0.188 0.437

Note that some teams, which did very well in 2012, did very poorly in 2013.  These teams with a disappointing year in 2013 included Atlanta, Detroit, and Washington.

Other teams (Arizona, New Orleans, Philadelphia, and Kansas City) improved a lot and exceeded expectations.

Some Analysis:  Well how closely is 2012 performance associated with 2013 performance?  One measure of this association is the correlation coefficient.   A correlation near +1 indicates that on average teams that performed well in 2012 also performed well in 2013.  A correlation near  -1 means a good year in 2012 is associated with a poor year in 2013.

I used Excel to calculate the correlation of percent of victories in 2012 with the percent of victories in 2013.   The correlation between these two variables is  -0.003, pretty close to zero.  Essentially the correlation statistic reveals no relationship – either negative or positive -- between record in 2012 and record in 2013.

My next approach involved categorizing team records for the two years into two categories -- less than or equal to 500 and greater than 500.
I have included the 500 clubs in the lower group.

Theses two-by-two tabulations are presented below.

 Observed 2012 and 2013 outcomes <=500 in 2013 >500 in 2013 <= 500 in 2012 11 7 18 >500 in 2012 8 6 14 19 13 32

The results indicate that

7/18 (38.9%) of teams with a record that is 500 or below in 2012 had a record over 500 in 2013.  This is pretty close to the (13/32) 40.6% of teams in the entire sample with records greater than 500.

8/14 (57.1% a majority) of teams with a record of over 500 in 2012 had a record that is less than or equal to 500 in 2013.  This is reasonably close to the 19/32 (59.4%) of teams in the entire sample with records that were 500 or lower.

My bottom line is that between 2012 and 2013 there was a fair amount of year-to-year volatility in NFL standings.

Additional Work:  How has the volatility in year-to-year records of the NFL changed over time? My perception is that volatility has increased.

How does the volatility of championship outcomes vary?   The Baltimore Ravens (the defending Super Bowl Champions) did not make the playoffs.  But several teams that went fairly far in the playoff last year – Denver, New England, San Francisco and Seattle – are still in the hunt.  It is possible that the 2013 NFC championship game – Seattle versus San Francisco – will be a repeat of the 2012 championship game.

Finally, how does year-to-year volatility in standing and championship outcomes differ across sports?  My sense is that football is more volatile than baseball or basketball.