This application offers an in-game football substitution recommendation system. It is intended to be used as a decision-making aid during a football game.
The application is deployed at this address.
Alternatively, to run it locally:
-
Clone this github repository or upload all of its files to the folder where you want to place this project.
-
Install the necessary packages from the requirements.txt file. In the terminal, replacing path with the path of your dedicated folder:
pip install -r path/requirements.txt
- Launch the application:
streamlit run path/↔️_Substitution_Suggestions.py
This application offers the following three features, each on one page:
- The "
↔️ Substitutions Suggestions" page features a substitution recommendation system based on StatsBomb's On-Ball Values. Every minute, a model tests different combinations of possible substitutions and brings out those that would be the most profitable. It therefore offers two features in one:- identifying players who are underperforming during the game in progress;
- suggesting a potential substitute to these players - based on past player performance and game context.
- The "📈 All players" page offers an overview of the performance of all players on the pitch.
- The "⏱️ Opponent Substitution Prediction" page presents the probable future substitutions of the opposing team, based on historical data and the scenario of the game, to possibly allow the coach to anticipate.
- Consider not only the cumulative On-Ball Values but also a Time Series of their evolution during the match.
- Include data relating to off-ball events, in particular through tracking data (pressing, etc.)
This application was created as part of a hackathon organized by Hays and Man City Football Club.