Saturday, September 14, 2013

NCAA FBS Performance 2008 to 2012

Over the last few weeks I have wanted to look at how NCAA FBS teams have fared year to year.  After spending just a little time putting the data together, I have calculated the correlation coefficient (called r, which measures the degree to which two variables vary together or apart) and the coefficient of determination (called r2, which measures the proportion of variance in common between the two variables).  So I will present both statistics - though I will only be interpreting r2 as that is the statistic that looks at what I am interested in explaining.

Since the 2008 NCAA FBS season, the variation in common of winning percentage in one year to the next explains from 36% in the 2008-2009 seasons to 23.5% for the 2011-2012 seasons.

Winning Percent r r2
2008-2009 0.5999 0.3599
2009-2010 0.5257 0.2764
2010-2011 0.5527 0.3055
2011-2012 0.4851 0.2354

So given that NCAA FBS previous seasons winning percent explains less than half of the next seasons winning percent, how does the Complex Invasion College Football Production Model's ranking relate year to year?  As you can see in the table below the variation in total NCAA FBS Production Rank from one year to the next explains between 28% and 43% of the variation in total rank.

Total Rank r r2
2008-2009 0.6536 0.4272
2009-2010 0.5362 0.2875
2010-2011 0.6551 0.4291
2011-2012 0.5262 0.2769

In terms of offensive ranking, the previous seasons offensive rank explains from 19% to 34% and in terms of defensive ranking, the prior seasons defensive rank explains from 28% to 34%.

Off. Rank r r2
2008-2009 0.4830 0.2333
2009-2010 0.4853 0.2355
2010-2011 0.5861 0.3435
2011-2012 0.4439 0.1970



Def. Rank r r2
2008-2009 0.5833 0.3402
2009-2010 0.5296 0.2805
2010-2011 0.5511 0.3037
2011-2012 0.5334 0.2845

In each case, the previous season is not a great predictor of the next season in that the variation in common is less than 50%; which is what makes college football so interesting - that for the "league" as a whole we really cannot predict what will happen.