EPA implementation with documentation, test and training script

[903124]

Closed #49 since origin repo is deleted.

For simple explanation of expected point added (EPA): https://www.espn.com/nfl/story/_/id/8379024/nfl-explaining-expected-points-metric

The PR has

1. training script named EPA_model.py which ultilize NFLdb and not intended to run here, but just for possible future enhancement/reproduction of model, and expected point for each down produced by the model

2. Updated game.py which contains script to parse play-by-play data and calculate expected point added

3. A small test

Feel free to ask for more detail on EPA model

[derek-adair]

Oh ok! I'll leave this open for now.

I will be forking nfldb this year before the season and getting that up-to-date w/ py3 etc etc. At that point i'll ask you to re-submit this to said fork.

[derek-adair]

Oh boy, I fixed some bugs and now we have conflicts. May wanna hold off on this until the code stabilizes a bit. I have some plans brewing on adding nfldb so there will likely be a loooot more breaking changes coming. Stay tuned! Thanks for your submission.

[903124]

I only edit the game.py and live and schedule one are due to the change on schedule update over the new season. In game.py EPA is added as a property and it should be safe to merge

[derek-adair]

I understand. The problem is i'm looking at updating nfldb and restructuring these projects. It's likely that this PR belongs in NFLDB. I am planning on simplifying this project a bit and making it entirely about pulling / organizing data. I am unsure if extrapolating/deducing data like EPA should belong here, there or maybe even in some kind of plugin.

[AccidentalGuru]

A plugin sounds like a good idea, maybe a test/simulate/project type for those NFL Analytics nerds that are ruining football.