Using Scikit-learn Multiple Regression to Predict NBA Rookies' Third Year Rating from Athletic Test

The target of this paper is to help NBA teams make wise decisions during the draft. Before the draft, the relatively complete statistics of rookies is an official athletic test holds by the NBA. Thus, the paper works on predicting the future rating of the rookies due to their statistics from athletic tests. Our prediction is based on a set of multiple regression models. We filtrate, split, and integrate the data. Then, we do tests to pick out the best set of independent variables that would fit the model best. Finally, our model works with r square in range 0.25-0.45, which is relatively good for models predicting human behaviours. Though, the result provided by our model is not accurate enough to make accurate rating predictions. However, the result of our model can provide an insight for the teams about the potential of the players under regular circumstances. Team cannot pick rookies based on our results. However, considering our results as a reference would be a great choice.