https://github.com/kahnvex/wincast

A win forecasting model for the NFL
https://github.com/kahnvex/wincast

deep-learning forecast keras nfl

Last synced: 15 days ago
JSON representation

A win forecasting model for the NFL

• Host: GitHub
• URL: https://github.com/kahnvex/wincast
• Owner: kahnvex
• License: mit
• Created: 2017-01-10T04:30:28.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2018-06-13T23:45:09.000Z (over 6 years ago)
• Last Synced: 2024-12-20T16:39:09.890Z (about 2 months ago)
• Topics: deep-learning, forecast, keras, nfl
• Language: Python
• Homepage:
• Size: 22.7 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# 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.

```sh

$ 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)

```