Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Machine-learning predictions for Draft Kings (NBA) :)

It scraped basically every single player's performance in every single NBA game ever. I tried XGBoost and Keras, and the Keras model outperformed the XGBoost model. Was about to incorporate real-time injury data, so if a player was injured or out that game it would not select them.

In the end it didn't perform too well. I think the limitation was my lack of domain knowledge, and not really knowing what features to select that would predict a players performance. Also data. I hear MLB is more consistent than NBA because there's just more data.



I'm sitting on MLB data (converting it to a BigQuery warehouse), am also interested in this space. The challenge w/ MLB is randomness plays a larger element - most folks in the MLB gambling space prefer other leagues where an information advantage more directly translates to profits.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: