Repository for building customizable passing networks with Matplotlib as part of the "Friends of Tracking" series. The code is prepared to use both eventing (StatsBomb) and tracking data (Metrica Sports).

The sample data can be found in the following repositories:

• StatsBomb: https://github.com/statsbomb/open-data
• Metrica Sports: https://github.com/metrica-sports/sample-data

Also, further information about the VAEP metric can be found in this repository: https://github.com/ML-KULeuven/socceraction

Video tutorial: https://www.youtube.com/watch?v=JZ6Jc-cvKX8

Please, follow these steps to be able to run the code:

1. Install the dependencies listed in requirements.txt
2. Download the sample data into the data/eventing and data/tracking folders.
3. (Optional) Run the prepare_vaep.py script which trains a model with StatsBomb's data and computes and sets the VAEP metric for each action in the eventing dataset. It might take some minutes.

Now you are ready to go!

The script run.py allows you to choose the following arguments:

• -m (--match-id) specifies the match ID.
• -t (--team-name) is the name of the team that will be plotted from the previous match.
• -s (--source) must be either eventing or tracking.
• -k (--plot-type) can be basic, pass_value, or tracking.

For eventing data, only basic and pass_value plot types are allowed. The basic one uses the number of passes as the metric for both the size and color of nodes and edges. On the other hand, pass_value uses the number of passes for the size, whereas the color range depends on the value of the passes (computed with the VAEP metric).

For tracking data, only basic and tracking plot types are allowed. basic will plot the nodes in the average locations where each player makes his passes. tracking will plot the players in their average location. This second plot type can be customized with the following optional arguments:

• -b (--ball-location). If present, it filters the location of the player to those frames when the ball was in the team's half (own_half) or in the opponent's half of the pitch (opponent_half).
• -c (--context). If present, it filters the location of the player to those frames when the selected team was either attacking or defending.

In addition, the colors and sizes of the elements in networks can be configured by changing the values in the visualization/plot_config.json file.

StatsBomb: python3 run.py -m 7576 -t Portugal -s eventing -k pass_value

Metrica: python3 run.py -m 1 -t Home -s tracking -k tracking -c attacking -b opponent_half

The resulting images will be saved onto the plots folder.

For further information, please contact me on Twitter: @SergioMinuto90.