I am in the middle of a senior project and I am lucky enough to be able to use basketball stats as part of my analysis. I have been looking into how regression models work and some of the data I have been checking have been very interesting. I started with point differential as the independent variable and wins as the dependent variable from a two year sample, and as expected the correlation was very high and the models all fit extremely well.
In an effort to find data that did not work very well I tried team assists as the independent variable and season wins as the dependent variable. The data came out very messy.
In the end the model for Least-Squared Regression was:
Season Wins=-24.93+3.067(Team Assists)
The R-squared of the model was .102. To put that is prospective 0 means no correlation, or completely random. If the R-squared is below .5 the model generally is not used. Basically the model is useless though it has a small positive correlation. It should also be noted that the Kernel Regression did not fit any better, but that is a little too much detail to explain and have you all still be reading.
What this means for the Kings. Well, in the NBA team assists do not equate to wins. The Kings are dead last in team assists, but surprisingly that is not as big of an issue as it seems at first glance.
I thought this was very interesting and thought it would be a good idea to share what I found with some other basketball fans. If anyone has any ideas onto why I would love to hear your feed back.