arxiv: 2604.11786 · v1 · submitted 2026-04-13 · 💻 cs.AI · cs.MA

Recognition: 3 theorem links

· Lean Theorem

GenTac: Generative Modeling and Forecasting of Soccer Tactics

Jiayuan Rao , Tianlin Gui , Haoning Wu , Yanfeng Wang , Weidi Xie

Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:39 UTC · model grok-4.3

classification 💻 cs.AI cs.MA

keywords generative modelingsoccer tacticsdiffusion modelsmulti-agent trajectoriestactical forecastingcounterfactual simulationteam sports analytics

0 comments

The pith

A diffusion-based model learns to generate varied, realistic soccer trajectories from tracking data while keeping team structure intact.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents GenTac as a way to treat open-play soccer as a stochastic multi-agent process that can be sampled from historical data. Instead of producing one fixed forecast, the system draws multiple plausible future paths that respect both spatial movement and discrete tactical events. This matters because real matches branch constantly, and single deterministic predictions miss the range of realistic options coaches and analysts actually face. The framework adds conditioning for opponent behavior, team style, and strategic goals, then grounds the continuous paths in a 15-class event vocabulary. Evaluations show the outputs stay geometrically accurate, preserve collective team shape, reproduce stylistic differences across teams and leagues, support guided what-if changes, and yield usable forecasts when rolled out.

Core claim

GenTac is a diffusion-based generative framework that models soccer tactics as a joint stochastic process over continuous player trajectories and discrete semantic events; by learning the distribution of movements from tracking data it can sample diverse long-horizon futures, accept rich contextual conditioning, and ground spatial dynamics in a 15-class tactical event space.

What carries the argument

Diffusion-based generative sampling of multi-player trajectories conditioned on context and grounded in a discrete 15-class event space.

If this is right

Generated trajectories maintain collective team structure even while varying individual paths.
The model distinguishes specific team styles and league conventions when conditioned appropriately.
Offensive or defensive guidance changes measurable quantities such as spatial control and expected threat in the simulated futures.
Future tactical outcomes can be read directly from statistics computed on the sampled rollouts.
The same training procedure transfers to basketball, American football, and ice hockey tracking data.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Coaches could use the controllable conditioning to test specific formation adjustments before matches.
The generative rollouts offer a natural way to build probabilistic opponent models for training AI agents in simulation.
If the event grounding proves robust, the same architecture could support real-time tactical adjustment tools during live games.

Load-bearing premise

Historical tracking data contains a learnable distribution that is rich enough to produce diverse yet realistic long-horizon trajectories respecting actual branching possibilities in matches.

What would settle it

Generate rollouts on held-out matches and check whether the fraction of trajectories that break real team formations or produce event sequences never observed in the data exceeds the rate seen in genuine matches.

read the original abstract

Modeling open-play soccer tactics is a formidable challenge due to the stochastic, multi-agent nature of the game. Existing computational approaches typically produce single, deterministic trajectory forecasts or focus on highly structured set-pieces, fundamentally failing to capture the inherent variance and branching possibilities of real-world match evolution. Here, we introduce GenTac, a diffusion-based generative framework that conceptualizes soccer tactics as a stochastic process over continuous multi-player trajectories and discrete semantic events. By learning the underlying distribution of player movements from historical tracking data, GenTac samples diverse, plausible, long-horizon future trajectories. The framework supports rich contextual conditioning, including opponent behavior, specific team or league playing styles, and strategic objectives, while grounding continuous spatial dynamics into a 15-class tactical event space. Extensive evaluations on our proposed benchmark, TacBench, demonstrate four key capabilities: (1) GenTac achieves high geometric accuracy while strictly preserving the collective structural consistency of the team; (2) it accurately simulates stylistic nuances, distinguishing between specific teams (e.g., Auckland FC) and leagues (e.g., A-League versus German leagues); (3) it enables controllable counterfactual simulations, demonstrably altering spatial control and expected threat metrics based on offensive or defensive guidance; and (4) it reliably anticipates future tactical outcomes directly from generated rollouts. Finally, we demonstrate that GenTac can be successfully trained to generalize to other dynamic team sports, including basketball, American football, and ice hockey.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

GenTac uses diffusion to generate conditioned soccer trajectories with 15-class event grounding, but lacks visible metrics and risks inconsistency from coarse discretization.

read the letter

GenTac applies diffusion models to generate plausible multi-player trajectories in open-play soccer, grounded by a 15-class event space and conditioned on context like team styles and strategic goals. The new part is the specific setup for stochastic, branching tactics rather than single forecasts or set pieces. It adds controllable conditioning for counterfactuals and claims generalization across sports. The work does well in framing the problem and proposing TacBench to test geometric accuracy, style fidelity, counterfactual changes to threat metrics, and rollout-based prediction. Extending to basketball and other sports is a nice touch. The soft spots are in the evidence. No quantitative results or ablations appear in the abstract, so the strong claims on preserving structural consistency and accurate simulation rest on unshown data. The 15-class event grounding might not be fine enough to enforce team-level invariants during long rollouts, which could let the model produce inconsistent formations despite matching marginal stats. That assumption about learnable distributions in tracking data needs careful validation. This is for people in sports data science or multi-agent simulation. A reader looking for generative tools in dynamic team settings would find the ideas useful. It should go to peer review. The framework has potential, but the referee can push for the missing metrics and grounding details.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces GenTac, a diffusion-based generative framework that models soccer tactics as a stochastic process combining continuous multi-player trajectories with discrete semantic events in a 15-class tactical space. Trained on historical tracking data, it generates diverse long-horizon rollouts under contextual conditioning (opponents, team/league styles, strategic objectives) and is evaluated on the proposed TacBench benchmark for four capabilities: geometric accuracy with team structural consistency, stylistic nuance simulation, controllable counterfactuals that alter spatial control and expected threat, and direct anticipation of tactical outcomes from rollouts. Generalization to basketball, American football, and ice hockey is also demonstrated.

Significance. If the quantitative results and methodological details hold, this would be a meaningful contribution to generative modeling of multi-agent dynamic systems in sports analytics, offering stochastic sampling and controllability beyond single deterministic forecasts or set-piece focus. The TacBench benchmark and cross-sport extension are useful additions; the continuous-discrete grounding approach has potential for preserving invariants while enabling branching.

major comments (3)

[Abstract] Abstract: the claims of 'high geometric accuracy while strictly preserving the collective structural consistency' and 'reliably anticipates future tactical outcomes' are asserted without any reported quantitative metrics, error bars, baseline comparisons, or ablation results; the central performance assertions on TacBench cannot be evaluated from the provided text.
[Abstract] Abstract (grounding mechanism): the 15-class tactical event space is stated to ground continuous spatial dynamics, yet no definition of the classes, no equations for the joint continuous-discrete diffusion process, and no analysis of whether coarse high-level actions suffice to enforce long-horizon team invariants (e.g., formation spacing or passing networks) are supplied; this directly bears on whether generated trajectories can violate structural consistency not encoded in the discrete space.
[Abstract] Abstract (counterfactuals): the claim that controllable simulations 'demonstrably alter spatial control and expected threat metrics' lacks any reported effect sizes, statistical tests, or comparison of pre/post-guidance distributions; without these, the controllability result cannot be assessed as load-bearing for the framework's utility.

minor comments (2)

The generalization experiments to other sports are mentioned only in passing; specific metrics or qualitative examples for basketball, American football, and ice hockey would strengthen the claim.
Notation for the diffusion process and event embedding is not introduced in the abstract; a brief equation or diagram reference would improve clarity for readers.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the positive summary of our contribution and the constructive major comments. We address each point below with references to the full manuscript and indicate the revisions made to strengthen the abstract.

read point-by-point responses

Referee: [Abstract] Abstract: the claims of 'high geometric accuracy while strictly preserving the collective structural consistency' and 'reliably anticipates future tactical outcomes' are asserted without any reported quantitative metrics, error bars, baseline comparisons, or ablation results; the central performance assertions on TacBench cannot be evaluated from the provided text.

Authors: The full manuscript reports the supporting quantitative metrics, error bars, baseline comparisons, and ablation studies in Section 5 (Experiments and TacBench evaluation), including trajectory accuracy, structural consistency scores, and outcome prediction results. The abstract is a high-level summary of these findings. We have revised the abstract to incorporate brief quantitative highlights and references to the TacBench results so that the central claims can be evaluated directly from the abstract text. revision: yes
Referee: [Abstract] Abstract (grounding mechanism): the 15-class tactical event space is stated to ground continuous spatial dynamics, yet no definition of the classes, no equations for the joint continuous-discrete diffusion process, and no analysis of whether coarse high-level actions suffice to enforce long-horizon team invariants (e.g., formation spacing or passing networks) are supplied; this directly bears on whether generated trajectories can violate structural consistency not encoded in the discrete space.

Authors: The manuscript defines the 15-class tactical event space, formalizes the joint continuous-discrete diffusion process with the relevant equations, and provides analysis of long-horizon invariant preservation (including formation spacing and passing networks) in the Methods section. We have updated the abstract to include a concise definition of the event classes and a brief statement on the grounding mechanism. revision: yes
Referee: [Abstract] Abstract (counterfactuals): the claim that controllable simulations 'demonstrably alter spatial control and expected threat metrics' lacks any reported effect sizes, statistical tests, or comparison of pre/post-guidance distributions; without these, the controllability result cannot be assessed as load-bearing for the framework's utility.

Authors: The full manuscript reports the effect sizes, statistical tests, and pre/post-guidance distribution comparisons for the counterfactual simulations in the Experiments section. We have revised the abstract to include a short summary of these quantitative results to make the controllability claim self-contained and assessable. revision: yes

Circularity Check

0 steps flagged

No significant circularity in claimed derivation or predictions

full rationale

The paper introduces GenTac as a diffusion-based generative model that learns the distribution of player trajectories and events from historical tracking data, then samples new rollouts under conditioning. All four key capabilities are presented as empirical results on the proposed TacBench benchmark rather than closed-form derivations. No equations, fitted parameters, or self-citations are invoked that would make the sampled trajectories or controllability metrics equivalent to the training inputs by construction. The 15-class event space is an architectural choice for grounding, not a tautological redefinition of the outputs. The framework is therefore self-contained against external data and does not reduce its central claims to its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the central claim rests on the unstated assumption that diffusion can capture the joint distribution of trajectories and events.

pith-pipeline@v0.9.0 · 5567 in / 1069 out tokens · 56819 ms · 2026-05-10T16:39:45.483580+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

GenTac, a diffusion-based generative framework that conceptualizes soccer tactics as a stochastic process over continuous multi-player trajectories and discrete semantic events... grounding continuous spatial dynamics into a 15-class tactical event space.
IndisputableMonolith/Foundation/ArithmeticFromLogic.lean LogicNat unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

autoregressive sliding-window rollout... causal window of 0.2s... 8 attention heads... M=4 attention layers
IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

TacBench-Event: 423 trajectory segments... 5 major event types and 15 fine-grained subtypes

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

TacticGen: Grounding Adaptable and Scalable Generation of Football Tactics
cs.AI 2026-04 conditional novelty 7.0

TacticGen generates realistic, adaptable football tactics via a multi-agent diffusion transformer trained on 3.3M events and 100M frames, supporting rule-, language-, or model-based guidance at inference time.

Reference graph

Works this paper leans on

44 extracted references · 3 canonical work pages · cited by 1 Pith paper

[1]

Soccernet: A scalable dataset for action spotting in soccer videos

Silvio Giancola, Mohieddine Amine, Tarek Dghaily, and Bernard Ghanem. Soccernet: A scalable dataset for action spotting in soccer videos. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2018

2018
[2]

Soccernet-v2: A dataset and benchmarks for holistic understanding of broadcast soccer videos

Adrien Deliege, Anthony Cioppa, Silvio Giancola, Meisam J Seikavandi, Jacob V Dueholm, Kamal Nasrollahi, Bernard Ghanem, Thomas B Moeslund, and Marc Van Droogenbroeck. Soccernet-v2: A dataset and benchmarks for holistic understanding of broadcast soccer videos. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2021

2021
[3]

Scaling up soccernet with multi-view spatial localization and re-identification.Scientific Data, 2022

Anthony Cioppa, Adrien Deliege, Silvio Giancola, Bernard Ghanem, and Marc Van Droogenbroeck. Scaling up soccernet with multi-view spatial localization and re-identification.Scientific Data, 2022

2022
[4]

Soccernet-tracking: Multiple object tracking dataset and benchmark in soccer videos

Anthony Cioppa, Silvio Giancola, Adrien Deliege, Le Kang, Xin Zhou, Zhiyu Cheng, Bernard Ghanem, and Marc Van Droogenbroeck. Soccernet-tracking: Multiple object tracking dataset and benchmark in soccer videos. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2022

2022
[5]

Towards universal soccer video understanding

Jiayuan Rao, Haoning Wu, Hao Jiang, Ya Zhang, Yanfeng Wang, and Weidi Xie. Towards universal soccer video understanding. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2025

2025
[6]

Matchtime: Towards automatic soccer game commentary generation

Jiayuan Rao, Haoning Wu, Chang Liu, Yanfeng Wang, and Weidi Xie. Matchtime: Towards automatic soccer game commentary generation. InProceedings of the Conference on Empirical Methods in Natural Language Processing, 2024

2024
[7]

Soccernet-caption: Dense video captioning for soccer broadcasts commentaries

Hassan Mkhallati, Anthony Cioppa, Silvio Giancola, Bernard Ghanem, and Marc Van Droogenbroeck. Soccernet-caption: Dense video captioning for soccer broadcasts commentaries. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2023

2023
[8]

Multi-agent system for comprehensive soccer understanding

Jiayuan Rao, Zifeng Li, Haoning Wu, Ya Zhang, Yanfeng Wang, and Weidi Xie. Multi-agent system for comprehensive soccer understanding. InACM Multimedia, 2025

2025
[9]

Soccernet game state reconstruction: End-to-end athlete tracking and identification on a minimap

Vladimir Somers, Victor Joos, Anthony Cioppa, Silvio Giancola, Seyed Abolfazl Ghasemzadeh, Floriane Magera, Baptiste Standaert, Amir M Mansourian, Xin Zhou, Shohreh Kasaei, et al. Soccernet game state reconstruction: End-to-end athlete tracking and identification on a minimap. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition...

2024
[10]

Camera calibration and player localization in soccernet-v2 and investigation of their representations for action spotting

Anthony Cioppa, Adrien Deliege, Floriane Magera, Silvio Giancola, Olivier Barnich, Bernard Ghanem, and Marc Van Droogenbroeck. Camera calibration and player localization in soccernet-v2 and investigation of their representations for action spotting. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2021

2021
[11]

A universal protocol to benchmark camera calibration for sports

Floriane Magera, Thomas Hoyoux, Olivier Barnich, and Marc Van Droogenbroeck. A universal protocol to benchmark camera calibration for sports. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2024

2024
[12]

Soccermaster: A vision foundation model for soccer understanding.Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2026

Haolin Yang, Jiayuan Rao, Haoning Wu, and Weidi Xie. Soccermaster: A vision foundation model for soccer understanding.Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2026

2026
[13]

Identifying team style in soccer using formations learned from spatiotemporal tracking data

Alina Bialkowski, Patrick Lucey, Peter Carr, Yisong Yue, Sridha Sridharan, and Iain Matthews. Identifying team style in soccer using formations learned from spatiotemporal tracking data. In2014 IEEE International Conference on Data Mining Workshop, pages 9–14. IEEE, 2014

2014
[14]

Consistency and identifiability of football teams: a network science perspective.Scientific reports, 10(1):19735, 2020

David Garrido, Daniel R Antequera, Javier Busquets, Roberto López del Campo, Ricardo Resta Serra, Silvestre Jos Vielcazat, and Javier M Buldú. Consistency and identifiability of football teams: a network science perspective.Scientific reports, 10(1):19735, 2020. |19

2020
[15]

Javier M Buldú, Javier Busquets, Ignacio Echegoyen, and F Seirul. lo. Defining a historic football team: Using network science to analyze guardiola’s fc barcelona.Scientific reports, 9(1):13602, 2019

2019
[16]

The impact of team quality, match outcomes, and match location on pitch-passing networks

Kongyun Huang, Changjing Zhou, Bo Gong, Xudong Yang, Javier M Buldú, and Miguel Ángel Gómez Ruano. The impact of team quality, match outcomes, and match location on pitch-passing networks. Scientific Reports, 15(1):23199, 2025

2025
[17]

Tacticai: an ai assistant for football tactics.Nature Communications, 2024

Zhe Wang, Petar Veličković, Daniel Hennes, Nenad Tomašev, Laurel Prince, Michael Kaisers, Yoram Bachrach, Romuald Elie, Li Kevin Wenliang, Federico Piccinini, et al. Tacticai: an ai assistant for football tactics.Nature Communications, 2024

2024
[18]

Taceleven: generative tactic discovery for football open play,

Siyao Zhao, Hao Ma, Zhiqiang Pu, Jingjing Huang, Yi Pan, Shijie Wang, and Zhi Ming. Taceleven: generative tactic discovery for football open play.arXiv preprint arXiv:2511.13326, 2025

work page arXiv 2025
[19]

Jointdiff: Bridging continuous and discrete in multi-agent trajectory generation.arXiv preprint arXiv:2509.22522, 2025

Guillem Capellera, Luis Ferraz, Antonio Rubio, Alexandre Alahi, and Antonio Agudo. Jointdiff: Bridging continuous and discrete in multi-agent trajectory generation.arXiv preprint arXiv:2509.22522, 2025

work page arXiv 2025
[20]

Skillcorner open data.https://github.com/SkillCorner/opendata

SkillCorner. Skillcorner open data.https://github.com/SkillCorner/opendata. Accessed: 2025-12-01

2025
[21]

An integrated dataset of spatiotem- poral and event data in elite soccer.Scientific Data, 12(1):195, 2025

Manuel Bassek, Robert Rein, Hendrik Weber, and Daniel Memmert. An integrated dataset of spatiotem- poral and event data in elite soccer.Scientific Data, 12(1):195, 2025

2025
[22]

Metrica sports sample data.https://github.com/metrica-sports/sample-data

Metrica Sports. Metrica sports sample data.https://github.com/metrica-sports/sample-data. Accessed: 2025-12-01

2025
[23]

Nba player movements: Public sportvu tracking logs.https://github.com/linouk23/ NBA-Player-Movements

linouk23 / Kostya Linou. Nba player movements: Public sportvu tracking logs.https://github.com/linouk23/ NBA-Player-Movements. Accessed: 2025-12-01

2025
[24]

Big data bowl.https://operations.nfl.com/gameday/analytics/big-data-bowl/

National Football League. Big data bowl.https://operations.nfl.com/gameday/analytics/big-data-bowl/. Accessed: 2025-12-01

2025
[25]

Big data cup 2026: Data release.https://github.com/bigdatacup/Big-Data-Cup-2026/releases/ tag/Data

Big Data Cup. Big data cup 2026: Data release.https://github.com/bigdatacup/Big-Data-Cup-2026/releases/ tag/Data. Accessed: 2025-12-01

2026
[26]

Analysing team coordination patterns from player movement trajectories in soccer: Methodological considerations

Roger Bartlett, Chris Button, Matthew Robins, Aviroop Dutt-Mazumder, and Gavin Kennedy. Analysing team coordination patterns from player movement trajectories in soccer: Methodological considerations. International Journal of Performance Analysis in Sport, 12(2):398–424, 2012

2012
[27]

Oscillations of centroid position and surface area of soccer teams in small-sided games.European journal of sport science, 11(4):215–223, 2011

Wouter Frencken, Koen Lemmink, Nico Delleman, and Chris Visscher. Oscillations of centroid position and surface area of soccer teams in small-sided games.European journal of sport science, 11(4):215–223, 2011

2011
[28]

Competing together: Assessing the dynamics of team–team and player–team synchrony in professional association football.Human movement science, 32(4):555–566, 2013

Ricardo Duarte, Duarte Araújo, Vanda Correia, Keith Davids, Pedro Marques, and Michael J Richardson. Competing together: Assessing the dynamics of team–team and player–team synchrony in professional association football.Human movement science, 32(4):555–566, 2013

2013
[29]

Introducing expected threat (xt).https://karun.in/blog/expected-threat.html, 2018

Karun Singh. Introducing expected threat (xt).https://karun.in/blog/expected-threat.html, 2018. Accessed: 2026-02-07

2018
[30]

Visualization of dominant region in team games and its application to teamwork analysis

Tsuyoshi Taki and JUNICHI Hasegawa. Visualization of dominant region in team games and its application to teamwork analysis. InProceedings computer graphics international 2000, pages 227–235. IEEE, 2000

2000
[31]

Is space-time attention all you need for video understanding? InProceedings of the International Conference on Machine Learning, 2021

Gedas Bertasius, Heng Wang, and Lorenzo Torresani. Is space-time attention all you need for video understanding? InProceedings of the International Conference on Machine Learning, 2021

2021
[32]

Decoupled weight decay regularization

Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. InProceedings of the International Conference on Learning Representations, 2019

2019
[33]

Adam: A method for stochastic optimization.(No Title), 2014

Kingma Diederik. Adam: A method for stochastic optimization.(No Title), 2014. |20

2014
[34]

Gentac: Generative modeling and forecasting of soccer tactics1.https://github.com/jyrao/GenTac, 2026

Jiayuan Rao, Tianlin Gui, Haoning Wu, and et al. Gentac: Generative modeling and forecasting of soccer tactics1.https://github.com/jyrao/GenTac, 2026

2026
[35]

ArXiv abs/2508.19182(2025), https://api.semanticscholar.org/CorpusID:280870241

Silvio Giancola, Anthony Cioppa, et al. Soccernet 2025 challenges results.arXiv preprint arXiv:2508.19182, 2025

work page arXiv 2025
[36]

1 35 90

Laurie Shaw and Friends of Tracking. Laurieontracking: Expected possession value (epv) grid.https: //github.com/Friends-of-Tracking-Data-FoTD/LaurieOnTracking, 2020. Accessed: 2026-02-07. |21 9 Acknowledgments Weidi would like to acknowledge the funding from Scientific Research Innovation Capability Support Project for Young Faculty (ZY-GXQNJSKYCXNLZCXM-I...

2020
[37]

Detailed Metadata （Time,Frame） 3.Task & Description
[38]

Segment Editing Panel (Annotation List & Buttons)
[39]

1";Step 2:Identify target sequences based on the specific task and description “3

Consistency Verfication (Check if logic is incorrect) Supplementary Figure 1|Manual data annotation interface.The original Chinese layout is overlaid with red English labels to highlight core components. Volunteers execute the annotation process through the following steps:Step 1:Watch tactical plays in the video playback view “1";Step 2:Identify target s...
[40]

• Build: The team progresses from the defensive third via patient, structured passing sequences while maintaining secure possession

Build Definition: Phases focused on controlled possession and structural advancement originating from the defensive zones. • Build: The team progresses from the defensive third via patient, structured passing sequences while maintaining secure possession. Criteria: Filtered viaPASS originating in the defensive third requiring forward progression< 0.05and ...
[41]

Transition Definition: Phases characterized by dynamic shifts in possession and rapid vertical advancements across the pitch. • Ball Win: Midfield interceptions and recoveries aimed at regaining possession from the opponent.Criteria: MidfieldCHALLENGE (Metrica); start_type∈{pass_interception,recovery}with separation< 3.0indicating high pressure (Open Data...
[42]

Threat Definition: Final-stage actions involving direct scoring attempts, critical box defenses, or high-danger area interventions. • Goal: Successful scoring events where the ball legally crosses the goal line.Criteria:SHOT ON TARGET-GOAL or WOODWORK-GOAL (Metrica);game_interruption_before∈{goal_for,goal_against}(Open Data);SuccessfulShot(Bundesliga). • ...
[43]

Set Piece Definition: Restarts of play from dead-ball situations governed by specific game rules. • Corner: A restart of play awarded to the attacking team when the ball completely crosses the goal line after last touching a defending player.Criteria: SET PIECE+CORNER KICK (Metrica); start_type∈{corner_reception,corner_interception}(Open Data);CornerKick(...
[44]

shot off target

Interruption Definition: Non-active play periods caused by rule infractions, ball out of bounds, or administrative stoppages. • Interruption: Temporary game stoppages primarily due to rule infractions, substitutions, or the ball going out of bounds. Criteria: CARD/FAULT RECEIVED (Metrica); end_type∈{foul_suffered,foul_committed}(Open Data); Foul/Substitut...