A Statistical Data Mining Approach to Determining the Factors that Distinguish Championship Caliber Teams in the National Football League
This paper uses machine learning and data mining techniques to explore most of the performance measurements used in American football. The main goal is to determine/extract those factors that are most responsible for the success of the so-called great NFL teams. We consider a very large number of commonly used performance statistics and variables along with success indicators like winning percentage, playoff appearance, and championship wins. It is held by many football analysts/experts that defense wins championships. In this paper, we seek to establish if indeed there is ample evidence that the so called dominant teams are based on more defense than offense. Other football analysts strongly believe and declare that high third down conversion percentage is a very strong indicator of playoff/championship caliber teams. Using five years worth of data from 2006 to 2010, our application of techniques such as cluster analysis, principal component analysis, factor analysis, support vector machine and traditional logistic regression reveal compellingly interesting and consistent (over the years) elements of NFL greatness.
Department, Program, or Center
The John D. Hromi Center for Quality and Applied Statistics (KGCOE)
Fokoue, E., and Foehrenbach D.(2001). A statistical data mining approach to determining the factors that distinguish championship caliber teams in the National Football League. Paper presented at the 2011 Joint Statistical Meetings, Miami Beach, FL.
RIT – Main Campus
This is the pre-print of a paper presented at the 2011 Joint Statistical Meetings, Miami Beach, FL, 30 July - 4 August, 2011.
Note: imported from RIT's Digital Media Library running on DSpace to RIT Scholar Works on April 2014.