SoccerNet 2026 Challenges Results

The 2026 edition focused on a diverse set of problems:

• Ball Action Anticipation (BAA): Predicting the timing and class of future ball-related actions based on a 30-second observation window.
• Player-Centric Ball Action Spotting (PCBAS): Temporally localizing actions while identifying the specific player (via team and jersey number) responsible.
• Novel View Synthesis (NVS): Generating images from unobserved camera angles in multi-view scenes.
• Spiideo SoccerNet Synloc: Mapping athletes to real-world pitch coordinates using single static-camera images.
• Visual Question Answering (VQA): Answering multiple-choice questions about broadcast content using multimodal inputs.

Across the tasks, winning submissions demonstrated several recurring strategies. In Ball Action Anticipation, top teams moved beyond simple baseline models by increasing input resolution and using multi-model ensembles to account for the inherent uncertainty of future events. In Player-Centric Ball Action Spotting, the most successful approaches prioritized player-ball interaction cues and refined temporal modeling for individual player tracks. A common theme across all tasks was the shift toward richer contextual representations, where models are increasingly required to synthesize visual data with tactical game-state information.

Standardized benchmarks like SoccerNet are essential for moving sports analytics from specialized, proprietary systems to reproducible, open-source research. By providing unified evaluation protocols and high-quality annotated datasets, these challenges allow the community to compare methods fairly. The techniques developed here—such as improved player tracking, action anticipation, and multimodal reasoning—have direct applications in automated game summarization, tactical analysis, and athlete performance monitoring.