FootballMLP

FootballMLP is a framework which stands for Football Machine Learning Pipeline. By testing different Machine Learning Algorithms as well careffully studying football statistics, FootballMLP is able to predict football games and perform some extra tasks as described above.

What I need to have?

Before using FootballMLP it is important to have the fixtures needed for the discriminating football results. Here is everything you need to install or perform previous executing the predictions.

Python packages

FootballMLP needs the next libraries to work:

• Matplotlib - 2D graphics library
• Numpy - Scientific computation package
• Scipy - Mathematics and science environment
• Sklearn - Machine learning library
• Seaborn - Data visualization library

Where is the data?

Everything you want to retrieve can be find by using API-Football. But FootballMLP has released a library for it apifootballpy

Dataset for Standings

This creates a whole dataset of teams' standings up to a season and a week (the last week played), the window parameter controls how many season before the current season wants to be saved.
python maps_teams_stats.py --country --league --season --week --window

python maps_teams_stats.py DE 78 2020 4 5

Note: by using a window of 5, it is expected to save the standings from 2016 to 2020.

Dataset for Fixtures

This creates a whole dataset of league's fixtures up to a season and a week (the last week played) mapped with the team's last standings, the window parameter controls how many season before the current season wants to be saved.
python maps_fixtures_stats.py --country --league --season --week --window

python maps_fixtures_stats.py DE 78 2020 4 5

Note: by using a window of 5, it is expected to save the standings from 2016 to 2020.

Automate Datasets

Here are several standings and fixtures commands of different leagues to be executed by one command.
.\maps_stats.sh --season --window

./maps_stats.sh 2020 1

Note: first mark the file as executable by running:

chmod +x maps_stats.sh

Dataset for Odds

This creates a whole dataset of all league's odds in a given season window, the window parameter controls how many season before the current season wants to be saved.
python maps_odds.py --season --window

python maps_odds.py 2020 1

Note: by using a window of 1, it is expected to save only odds for season 2020.

Testing FootballMLP

evaluation.ipynb contains the evaluation by league.

Predicting with FootballMLP

prediction.ipynb contains the prediction of the weekgame.
ultimate_prediction.ipynb contains the prediction of the weekgame using a voting classifier trained diversely.