A win forecasting model for the NFL
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data
evaluate
notebooks
wincast
.gitattributes
.gitignore
LICENSE
Pipfile
Pipfile.lock
README.md
makefile
setup.py

README.md

wincast

A win forecasting model for the NFL

Note: This package is still under development. For now, you can play around with it via the command line.

Usage

The model will predict whether or not the offense team will win. An output of 1 means the model is forecasting a win for the offense team. A 0 means the model is forecast a loss (or tie) for the offense team.

$ pip install -r requirements.txt
$ python
>>> import numpy as np
>>> from wincast.train import Trainer
>>>
>>> model = Trainer()
>>> model.train()
>>> # Now you can make predictions. Input features are as follows:
>>> # (quarter, minute, second, points offense, points defense,
>>> #     t.o.l. offense, t.o.l. defense, down, yards to go,
>>> #     yards from own goal)
>>>
>>> # Here is an example of a call to predict, where the model
>>> # forecasts a win for the team on offense:
>>> model.predict([[4, 0, 5, 20, 7, 3, 2, 1, 2, 20]])
array([[1]], dtype=int32)

>>> # Get the probability of each class 0/1:
>>> model.predict_proba(np.array([[4, 0, 5, 20, 7, 3, 2, 1, 2, 20]]))
array([[ 0.00880867,  0.99119133]], dtype=float32)