Introduction
Modern soccer analytics relies on the fusion of two distinct data streams: event data and tracking data. Event data provides a semantic log of actions (e.g., passes, shots, and tackles) manually recorded by humans. Tracking data, typically captured at 25 frames per second via optical systems, provides the precise coordinates of every player and the ball.
However, these streams are frequently misaligned due to human reaction times in event logging and differing internal system clocks. This lack of synchronization—often referred to as "constant bias"—hinders accurate analysis of spatial context, such as determining if a defender was truly out of position during a specific pass. ETSY (Event and Tracking data SYnchronization) is a rule-based algorithm designed to bridge this gap efficiently.
The ETSY Methodology
To address the synchronization challenge, the authors propose a two-step pipeline that prioritizes physical realism and computational speed.
1. Kickoff Synchronization
The first step addresses the global time offset between the two systems. By analyzing the first five seconds of a match, ETSY identifies the exact frame where the ball is closest to the kickoff player and exhibits maximum acceleration. This frame serves as a temporal anchor to shift the entire tracking dataset into alignment with the event log, significantly reducing the initial bias.
2. Event-Level Alignment and Physics-Informed Filtering
For subsequent events, ETSY searches a temporal window around each event's timestamp. It employs a scoring function based on the proximity of players and the ball to their expected locations. The "secret sauce" of ETSY lies in its action-specific physical filters, which ensure that matched frames adhere to realistic soccer dynamics.
As shown in the table above, different actions like "headers" or "shots" have specific height, distance, and acceleration constraints. This prevents the algorithm from incorrectly matching an event to a nearby player who was not actually involved in the action, such as matching a header to a player standing on the ground.
Performance and Results
The study evaluated ETSY on a dataset of 313 European first-division matches. The results demonstrate that ETSY offers a compelling balance of speed and reliability compared to existing methods like sync.soccer.
Speed and Efficiency
ETSY is approximately 3x faster than the state-of-the-art sync.soccer method, processing a full game in roughly 2.7 minutes compared to 8.6 minutes. This efficiency makes it highly suitable for large-scale historical analysis where processing thousands of matches is required.
Coverage as a Quality Filter
While sync.soccer achieves 100% coverage, it often does so by "forcing" matches even when data quality is poor. In contrast, ETSY’s 91.61% coverage reflects its role as a quality control mechanism; it refuses to return a match if physical constraints are not met, ensuring higher confidence in the resulting synchronized data.
Conclusion
ETSY demonstrates that a well-tuned, rule-based system can match the performance of complex bio-informatics-inspired algorithms while significantly reducing computational overhead. By leveraging domain-specific knowledge—such as kickoff alignment and physics-informed filtering—ETSY provides a robust, open-source solution for the critical task of soccer data synchronization, enabling more accurate spatial analysis for researchers and clubs alike.
Summary updated