Teaching Regression Using American Football Scores

Document Type

Conference Proceeding

Publication Date


Published In

International Statistical Institute 56th Session


Scores in professional American football games follow a distribution that is noticeably skewed towards larger values. However, the difference between the home teams’ scores and the road teams’ scores (the point spreads) do follow a distribution that is very close to Normal. Furthermore, the residuals from linear regressions of the point spreads on the Las Vegas betting spreads and on partial game point spreads (e.g., the point spreads at half-time) are also very close to Normal, and suggest that a linear model is appropriate. These data are also suitable for logistic regression problems, such as estimating the probability that the home team wins given the betting spread, the half time score, or both. This presents an ideal setting for teaching simple and multiple linear regression and logistic regression procedures for any level of statistics course.


Extra Sum of Squares F Test, Regression t test, Normal Quantile Plot


International Statistical Institute 56th Session

Conference Dates

August 22-29, 2007

Conference Location

Lisbon, Portugal