NBA Logistic Regression

Analysis on an NBA player dataset to determine:

• How to define a good rebounder, Good Rebounder, (> 8 total rebounds per game).
  • This index was categorical, so I converted it to a numerical column called RebounderNumeric in order to make a prediction on a binary value.
• What variables in the dataset can be used to make a prediction of whether a player will be a good rebounder.
• How accurate a prediction can be, and how to tune the model to avoid overfitting.

In order to test the model, I created 3 players with hypothetical statistics based upon the chosen predictor variables:

• Pos - Position
• MP - Minutes Played
• PS/G - Points Scored Per Game
• AST - Assists Per Game
• STL - Total Steals Per Game

I stacked the parameters vertically using Numpy into a new test set and ran a prediction using sklearn on RebounderNumeric.