Further, the beauty of multivariant regression analysis is that it works on averages and produces variable coefficients that can also be mathematically assessed in terms of reliability and probability. Obviously, there could also be outlyer tests conducted on the data to assess if a team's result(s) are out of the normal range of expected outcomes.
Huh?
I still remember some of the basics from my econometrics studies. Multiple regression modeling is a very statistically sound method to analyze data of this sort. Not only to you derive the coefficients for the variables but there are "tests" to determine the confidence level of the individual coefficients. Then there are further "test" to assess outlyers and "goodness of fit" of the model.
