Free Throws, Blocks and Personal Fouls

[Goujer]

I noticed that free throws, blocks and personal fouls are not gathered by getSeasonData().
I'm not 100% familiar with the algorithm yet but I figured I'd add them to the list and see what happens. I am curious though why these were emitted from the start and I figured you guys might have a reason.

On another note much of the missing columns like TOV, Opp., ORB, MP and others have been added to sports-reference, I'm working on importing the data into the CSVs and I am seeing a slight accuracy boost.

[adeshpande3]

No particular season they were emitted. At the beginning, I just wanted to get more of the basic features before adding other ones. Adding more features definitely does provide more information, it would create vectors of high dimensions which generally does make it a bit harder for the models to learn, so I'm not sure if a stat like minutes played would help that much. Although I could be wrong and like you mentioned, if you're seeing a slight performance boost, it could be worth it.

[Goujer]

The increase is very slight like 0.3% average increase in accuracy. It's small but consistent in that increase.
I'm not the most familiar with sklearn or the models it uses so when you say it makes it harder, is that harder in a way that it could get things wrong or harder in that it takes more time?

[adeshpande3]

Harder in the sense that increasing the number of features generally increases the number of training samples you may need to get a sufficiently trained model.

https://en.wikipedia.org/wiki/Curse_of_dimensionality
http://www.visiondummy.com/2014/04/curse-dimensionality-affect-classification/