How ml-kuleuven/un-xpass works
un-xPass evaluates soccer-player creativity from StatsBomb 360 passing data. Its user workflow is to load and convert match events, build feature/label datasets, train and evaluate pass-selection, pass-success, and pass-value models, then combine trained components into creativity ratings. The graph preserves these domain stages and the SoccerMap model layer, plus the database backends. Some command and rating details are documented in the README rather than visible in the excerpts; those links cite the README, and unsampled implementation specifics are omitted.
Main entry points: __main__.py CLI commands · components/base.py Component training · databases/base.py Database interface · ratings.py Creativity ratings
Components 17
CLI workflow
- CLI commands
unxpass/__main__.py
Match data
- Event import
unxpass/databases/base.py - Database interface
unxpass/databases/base.py - SQLite store
unxpass/databases/sqlite.py - HDF store
unxpass/databases/hdf.py - Feature generation
unxpass/features.py - Label generation
unxpass/labels.py - Pass datasets
unxpass/datasets.py
Modeling
- Component training
unxpass/components/base.py - Pass selection
unxpass/components/pass_selection.py - Pass success
unxpass/components/pass_success.py - Pass value
unxpass/components/pass_value.py - SoccerMap network
unxpass/components/soccermap.py
Creativity ratings
- Creativity ratings
unxpass/ratings.py
Configuration and tracking
- Experiment configuration
unxpass/config.py
Other components
- Researcher
- MLflow tracking
Connections 25
- Researcher → CLI commands: runs commands
README.md - CLI commands → Event import: loads events
unxpass/__main__.py - Event import → Database interface: stores actions
unxpass/databases/base.py - Database interface → SQLite store: selects backend
unxpass/databases/__init__.py - Database interface → HDF store: selects backend
unxpass/databases/__init__.py - CLI commands → Pass datasets: creates dataset
README.md - Pass datasets → Feature generation: computes features
unxpass/datasets.py - Pass datasets → Label generation: computes labels
unxpass/datasets.py - Feature generation → Database interface: reads match data
unxpass/features.py - Label generation → Database interface: reads match data
unxpass/labels.py - CLI commands → Experiment configuration: composes settings
unxpass/__main__.py - CLI commands → Component training: trains component
unxpass/__main__.py - Component training → Pass datasets: initializes dataset
unxpass/components/base.py - Pass selection → Component training: extends training
unxpass/components/pass_selection.py - Pass success → Component training: extends training
unxpass/components/pass_success.py - Pass value → Component training: extends training
unxpass/components/pass_value.py - Pass selection → SoccerMap network: uses network
unxpass/components/pass_selection.py - Pass value → SoccerMap network: uses network
unxpass/components/pass_value.py - Pass success → SoccerMap network: uses network
unxpass/components/pass_success.py - CLI commands → Creativity ratings: computes ratings
README.md - Creativity ratings → Pass selection: loads model
unxpass/ratings.py - Creativity ratings → Pass success: loads model
unxpass/ratings.py - Creativity ratings → Pass value: loads model
unxpass/ratings.py - CLI commands → MLflow tracking: logs runs
unxpass/__main__.py - Component training → MLflow tracking: tracks training
unxpass/components/base.py