The Future

[schorrm]

So, a few things I want to ask the community here:

1. Do we want to go back on to PyPI? If so, what should the new name be?
2. Do we have particular goals to work on? I've recently been doing a lot of stuff with retrosheet, mainly with the prebuilt parquet databases from boxball, that could be something, absolutely open to ideas. Do we want to add modelling / projections beyond scraping? We could do something like integrate Marcel the Monkey or something similar (perhaps something more advanced?). Hidden Markov?
3. Any new datasets that have come out that we don't have yet?

[andersonfrailey]

1. I'd love for this to be up on PyPI. I don't have a name off the top of my head, but I can try and think of something and will post if inspiration strikes.
2. Adding projections would be a great feature. It's possible there are already some libraries out there we could integrate. I'm personally interested in creating more visualizations as well.

[tqwhitetech]

1 ) Agree with the person above, definitely should be on PyPi, but I too don't have a cool name.

Off topic a bit, but do we know where we stand with the issues from the original repo? If there are still unresolved issues in there should we get them brought over here so we can start working through them? Its tough to tell what issues are stuck in pull requests and what just never got done.

[schorrm]

I think we have almost all of them covered

[schorrm]

Given that nobody has a good name idea, what if we did a Pillow kind of thing? Like if we listed it under pybaseball2 or something, so it'd be pip install pybaseball2, but then import pybaseball?

[andersonfrailey]

That works for me

[schorrm]

This is up as pybaseball2 on pip, the package is still used as pybaseball. I have the CI set up for publishing on a GH Release.

[schorrm]

And another idea: what about implementing OpenWAR here?

[andersonfrailey]

I think that'd be a cool feature to have. I'd offer to help, but we're starting a new semester on Tuesday and my free time will be fairly limited.

[schorrm]

A major update: James LeDoux reached out to me, and offered to have me take over the original repo. I think it would be better to do that, we'll maintain the stars etc, and back under the original name.

[schorrm]

Obviously, I would migrate all of our improvements over too.

[TAThor]

During this off-season, should I have the time, I would like to work on some baseball-related machine learning projects, but I am not sure if adding ML functionality directly to this repo makes sense, or if it would make more sense to create a new repo (e.g., pybaseballML) which implements pybaseball but is dedicated to machine learning projects. I'd like some input on which option you guys think is more preferable. I give a brief example project below.

Pitch Prediction Given the entire pitch-by-pitch history (potentially alongside other informative data) of a pitcher, train a neural network that attempts to predict the next pitch based on the prior pitch sequence (e.g., in the current at-bat, the pitcher has thrown the sequence [1. Fastball 2. Changeup 3. Changeup]. The neural network gives a prediction of what it thinks will be the next pitch).