# Intention Driven Identification of In-Possession Match Phases in Association Football through Temporal Graph Learning

arXiv: https://arxiv.org/abs/2606.09289 · HTML: https://mlsanity.com/paper/2606.09289/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 59 references that are in the map.
- First posted on arXiv 8 Jun 2026.

## Builds on (most cited first, 7 of 7)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Human Motion Trajectory Prediction: A Survey](https://mlsanity.com/paper/1905.06113/) (2019)
- [SR-LSTM: State Refinement for LSTM towards Pedestrian Trajectory Prediction](https://mlsanity.com/paper/1903.02793/) (2019)
- [Crowded Scene Analysis: A Survey](https://mlsanity.com/paper/1502.01812/) (2015)
- [The ActivityNet Large-Scale Activity Recognition Challenge 2018 Summary](https://mlsanity.com/paper/1808.03766/) (2018)
- [OpenSTARLab: Open Approach for Spatio-Temporal Agent Data Analysis in Soccer](https://mlsanity.com/paper/2502.02785/) (2025)
- [Precise Temporal Action Localization by Evolving Temporal Proposals](https://mlsanity.com/paper/1804.04803/) (2018)

## Abstract (arXiv)

Understanding tactical organisation in association football requires identifying in-possession match phases that are shaped by evolving tactical intentions rather than by spatial patterns alone. This study proposes an intention-driven framework for identifying phases from tracking data. Seven German Bundesliga matches recorded at 25 Hz with TRACAB were analysed. A hierarchical model was defined with three tactical intentions (Invade Opponent Space, Keep Possession, Scoring) and six phases (Build Up, Progression, Counter Attack, Maintenance, Sustained Threat, Finishing). A Temporal Graph Attention Network (T-GAN) combined frame-level player-interaction graphs, contextual features, and Transformer-based temporal modelling. Performance was evaluated using frame-level F1 and temporal Intersection over Union (tIoU). T-GAN achieved macro-average frame-level F1 scores of 0.87 at the intention level, 0.76 for invasion-related phases, and 0.79 for scoring phases. After filtering, mean class-wise tIoU increased from 0.44 to 0.67 for intentions and from 0.40 to 0.57 for phases, showing that sequence-level evaluation captured fragmentation and boundary errors missed by frame-level metrics. Model comparisons indicated that Transformer-based sequence modelling drove coherent segmentation, while graph-based relational modelling was most beneficial for Counter Attack. Misalignment analysis revealed common sequence level misalignments, mainly Build Up/Progression ambiguity, Build Up/Maintenance confusion, Counter Attack/Progression inconsistency, and shortened pre-shot Finishing segments. Overall, the framework translates tracking data into tactically interpretable phase representations for automated annotation, tactical analysis, and playing-style profiling.