floodlight is a Python package for streamlined analysis of sports data. It is designed with a clear focus on scientific computing and built upon popular libraries such as numpy or pandas.
Load, integrate, and process tracking and event data, codes and other match-related information from major data providers. This package provides a set of standardized data objects to structure and handle sports data, together with a suite of common processing operations such as transforms or data manipulation methods.
All implementations run completely provider- and sports-independent, while maintaining a maximum of flexibility to incorporate as many data flavours as possible. A high-level interface allows easy access to all standard routines, so that you can stop worrying about data wrangling and start focussing on the analysis instead!
floodlight simplifies sports data loading, processing and advanced performance analyses. Check out the example below, where querying a public data sample, filtering the data and computing the expended metabolic work of the active home team players is done in a few lines of code:
>>> from floodlight.io.datasets import EIGDDataset
>>> from floodlight.transforms.filter import butterworth_lowpass
>>> from floodlight.models.kinetics import MetabolicPowerModel
>>> dataset = EIGDDataset()
>>> home_team_data, away_team_data, ball_data = dataset.get()
>>> home_team_data = butterworth_lowpass(home_team_data)
>>> model = MetabolicPowerModel()
>>> model.fit(home_team_data)
>>> metabolic_power = model.cumulative_metabolic_power()
>>> print(metabolic_power[-1, 0:7])
[1669.18781115 1536.22481121 1461.03243489 1488.61249785 773.09264071
1645.01702421 746.94057676]
To find out more, see the full set of features below or get started quickly with one of our many tutorials from the official documentation!
We provide core data structures for team sports data, parsing functionality for major data providers, access points to public data sets, data filtering, plotting routines and many computational models from the literature. The feature set is constantly expanding, and if you want to add more just open an issue!
- Tracking data
- Event data
- Pitch information
- Teamsheets with player information
- Codes such as ball possession information
- Properties such as distances or advanced computations
- Tracab/ChyronHego: Tracking data, Teamsheets, Codes
- DFL/STS: Tracking data, Event data, Teamsheets, Codes
- Kinexon: Tracking data
- Opta: Event data (F24 feeds)
- Second Spectrum: Tracking data, Event data
- Skillcorner: Tracking data from the Skillcorner Open Dataset (new)
- Sportradar: Event data
- StatsPerform: Tracking data, Event data (with URL access)
- StatsBomb: Event data
- EIGD-H (Handball tracking data)
- IDSSE (Football tracking and event data) (new)
- StatsBomb OpenData (Football event data)
- Spatial transformations for all data structures
- Lowpass-filter tracking data
- Slicing, selection and sequencing methods
- Plot pitches, player positions and model overlays
- Approximate Entropy
- Centroids
- Distances, Velocities & Accelerations
- Metabolic Power & Equivalent Distances
- Voronoi Space Control
The package can be installed easily via pip:
pip install floodlight
You can find all documentation here.
Check out Contributing.md for a quick rundown of what you need to know to get started. We also provide an extended, beginner-friendly guide on how to start contributing in our documentation.
If you've used floodlight in your scientific work, please cite the corresponding paper.
@article{Raabe2022,
doi = {10.21105/joss.04588},
url = {https://doi.org/10.21105/joss.04588},
year = {2022},
publisher = {The Open Journal},
volume = {7},
number = {76},
pages = {4588},
author = {Dominik Raabe and Henrik Biermann and Manuel Bassek and Martin Wohlan and Rumena Komitova
and Robert Rein and Tobias Kuppens Groot and Daniel Memmert},
title = {floodlight - A high-level, data-driven sports analytics framework},
journal = {Journal of Open Source Software}
}
Why do we need another package that introduces its own data structures and ways of dealing with certain problems? And what's the purpose of trying to integrate all different data sources and fit them into a single framework? Especially since there already exist packages that aim to solve certain parts of that pipeline?
Our answer is - although we love those packages out there - that we did not find a solution that did fit our needs. Available packages are either tightly connected to a certain data format/provider, adapt to the subtleties of a particular sport, or solve one particular problem. This still left us with the essential problem of adapting to different interfaces.
We felt that as long as there is no underlying, high-level framework, each and every use case again and again needs its own implementation. At last, we found ourselves refactoring the same code - and there are certain data processing or plotting routines that are required in almost every project - over and over again just to fit the particular data structures we're dealing with at that time.
This project has been kindly supported by the Institute of Exercise Training and Sport Informatics at the German Sport University Cologne under supervision of Prof. Daniel Memmert.