SPADL

SPADL is a standardized representation for soccer match event data designed to facilitate analysis and model-building across datasets. Standing for Soccer Player Action Description Language, SPADL provides a formal schema for describing discrete on-ball actions in a match, decoupled from raw video or tracking data. The goal is to enable cross-source comparability and to support machine learning workflows built on a common representation.

A SPADL record encodes when and where an action occurred, who performed it, and what happened as a result. Typical fields include time, team, player, action_type, start_location, end_location, and outcome, plus optional attributes such as body_part, pass_type or shot_type, and situational tags like zone or distance. The action_type vocabulary covers passes, shots, dribbles, tackles, interceptions, fouls, and clearances, among other standard in-play events. The model emphasizes high-level actions to enable scalable analysis.

In practice, SPADL is used to transform data from broadcasts, tracking systems, or labeled datasets into a uniform format, commonly stored as CSV or Parquet files. It supports research on action prediction, tactical analysis, and performance evaluation, and is frequently applied alongside related datasets that provide labeled match events and segments.

Limitations include reliance on consistent labeling across sources, the need for a common coordinate system, and challenges in representing complex sequences or ambiguous events. As a standardized schema, SPADL evolves with new data sources and research needs, and users should verify schema versions and field definitions when integrating datasets.