It’s really easy for me to want to re-invent the wheel. I like the scientific process and learning on my own. However, it is not an efficient way to approach projects, like the NFL play-by-play machine learning project I’m working on.
I tried to develop my own API that gets me the play-by-play data from NFL GameCenter. This was a futile effort, but one helluva learning experience. I learned to use Python libraries like BeautifulSoup and URLlib. Valuable tools I hope to use later for data mining. I also learned more about JSON objects, and that the NFL publishes all play-by-play data to JSON objects. Good to keep in mind for future reference, and a project I’ve got for reference on GitHub (https://github.com/mooneyj3/football-play-agger)
Alas, my goal is not to develop API’s, my goal is to analyze play-by-play data using machine learning. I reached out to a professor at Western Washington University early on in this project, and I gave him an update. He was kind to remind me to use my existing data to extrapolate machine learning methods first and develop a model — up-to-date play-by-play would come later.