
Understanding Machine Football's Attribute Scores

Fergus Sowery
Scroll to explore
Comparisons
Machine Learning

Within our model, we rank both teams and individual outfield players across three core areas of performance: defensive, build-up, and final-third play. These rankings are based on historical data spanning 1.1 billion on pitch events, enabling the comparison of 320,000+ players.
Machine Football Attributes
Team and player attributes are expressed as percentile rankings from 1-100, measuring how effectively individuals and squads performs across different areas of the pitch. All scores are calculated relative to every other team/player included in our model.
Player scores can be adjusted relative to:
Other players in the same position/all positions
Other players within the same league/the world
This allows for more accurate, like-for-like comparisons.
Within each attribute category, players are assessed across three to four underlying metrics, which combine to produce their overall score for that category.
Defensive attributes include:
Tackling
Defensive Headers
Ball Recovery
Build-up play attributes include:
Creativity
Passing Accuracy
Dribbling
Crossing
Final-third play attributes include:
Finishing (shots)
Finishing (aerial)
Attacking Positioning
For every player in our system, these attributes can then be viewed over time, allowing for precise tracking and analysis, updated on a game-by-game basis.