Structural Pass Analysis in Football: Learning Pass Archetypes and Tactical Impact from Spatio-Temporal Tracking Data

arXiv:2603.28916 · cs.LG, cs.CY · Submitted 2026-03-30 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: Today's paper: "Structural Pass Analysis in Football".

Jane: The gist: Structural representations derived from tracking data can reveal interpretable tactical patterns in football by modelling how passes reshape defensive organisation, which complements existing action valuation models.

Tom: First, who's behind it and why it matters.

Title and authors: Jane: Now, let's look at what the authors suggested to make this structural analysis even better because right now the work is focused on immediate structural effects but it doesn't fully capture the long-term value of a possession sequence >

Tom: They point out that we need to integrate sequential models that can track how those structural advantages evolve over multiple actions, not just one pass >

Jane: They also propose using representation learning techniques, like graph neural networks, to learn these structural features directly from the raw tracking data instead of relying only on pre-defined metrics >

Tom: That means we could build systems that aren't just looking at one pass in isolation but seeing the whole sequence as a dynamic structure >

Jane: This moves us toward building predictive models for pass receiver outcomes using temporal graph network frameworks which is something we’ve seen discussed before on page four >

Tom: That connects this structural work to the action valuation methods we already have in place giving us a richer way to value possession >

Jane: They want to see how these structural advantages change over time which is a big step for understanding tactical flow on the field >

Tom: It suggests that the next evolution of this kind of analysis will involve modeling the entire dynamic interaction between passes and defensive geometry >

Jane: If we can do that, we could predict not just where a pass goes but how it sets up the defense for the next few actions >

Tom: It’s moving from describing a snapshot of structural impact to modeling the actual evolution of tactical setups on the field >

Jane: This opens up possibilities for creating much more sophisticated AI that understands the flow and shape of football in a continuous, time-dependent way >

Tom: The implication is that we can eventually build models that capture the full narrative of an attacking sequence not just isolated events >

Jane: So, they’re pushing toward a much deeper level of temporal understanding for this kind of structural analysis on page four >

Tom: It’s about taking those static structural insights and making them dynamic and predictive for real-time game situations >

Jane: If we can get there, it could fundamentally improve how we analyze complex, continuous sports data >

Tom: This leads us right into how the AI actually goes about learning these dynamic structures from the tracking data itself on page four >

The paper's summary: Jane: So to wrap up on "Structural Pass Analysis in Football: Learning Pass Archetypes and Tactical Impact from Spatio-Temporal Tracking Data," this research shows we can quantify how passes physically reshape a defense >

Tom: Exactly. It moves us away from just looking at whether a pass led to a goal or possession toward understanding the structural influence of the pass on the entire defensive configuration >

Jane: The key is that they use these metrics like Tactical Impact Value to identify specific patterns like line-breaking passes being more impactful than circulatory ones >

Tom: And this means we can start using AI to predict territorial progression based on the structural nature of the passing action >

Jane: It gives us a way to map out team styles based on these structural patterns which is a big step for tactical analysis >

Tom: The player-level insights show that build-up defenders are crucial drivers of this structural progression and players with high cumulative TIV are predominantly central defenders >

Jane: They also defined how much two players amplify each other structurally by looking at delta T IV p→r equals T IV p→r minus T IV p >

Tom: It really shows how much more detail we can extract from tracking data when we focus on the geometry of the game rather than just a single event >

Jane: This is a big shift because it gives us a new lens for analyzing football dynamics in spatio-temporal data >

Lu: From an AI perspective, the idea of learning these structural features directly from tracking data is fascinating because it’s like teaching an AI to see the underlying physical laws of tactical movement >

Meng: In terms of practical impact if we can model these structural interactions we could potentially build systems that suggest optimal passing strategies based on real-time defensive geometry >

Lalam: I see this as a way to improve the culture around sports analysis by giving us a deeper more physical understanding of how complex human interactions play out under pressure >

Tom: So, to sum up "Structural Pass Analysis in Football: Learning Pass Archetypes and Tactical Impact from Spatio-Temporal Tracking Data," this research shows we can quantify how passes physically reshape a defense >

Jane: It moves us away from just looking at whether a pass led to a goal or possession toward understanding the structural influence of the pass on the entire defensive configuration >

Lu: It moves the discussion toward modelling how passes reshape defensive organisation which is a powerful way to look at spatio-temporal sports data >

Meng: I think it’s the bridge between raw tracking data and actionable tactical understanding we need to build >

The paper's improvements: Tom: So we’re wrapping up on "Structural Pass Analysis in Football: Learning Pass Archetypes and Tactical Impact from Spatio-Temporal Tracking Data" It shows how we can move past just counting passes to seeing how they physically change a defense >

Jane: Exactly. We’re talking about using structural metrics like the Line Bypass Score and the Space Gain Metric to figure out exactly how much a pass messes with the defensive shape >

Tom: And they create these four archetypes—circulatory, destabilising, line-breaking, and space-expanding—using unsupervised clustering on those structural features >

Jane: So instead of just one pass we’re categorizing them by the specific mechanism they use to interact with the defense spatially >

Lu: It’s a pretty powerful way to model how passes reshape defensive organisation which complements those existing action valuation models >

Meng: I think the real utility is seeing those team-level style representations which use a two-dimensional map to categorize teams based on structural patterns like space expansion versus circulatory destabilisation >

Lalam: From an AI perspective, the idea of learning these structural features directly from tracking data is fascinating because it’s like teaching an AI to see the underlying physical laws of tactical movement >

Tom: And on the player level they found that build-up defenders are key drivers of this structural progression and players with high cumulative TIV are predominantly central defenders >

Jane: They also defined how much two players amplify each other structurally by looking at delta T IV p→r equals T IV p→r minus T IV p >

Lu: That's a concrete way to look at player synergy we can quantify how much better two players work together structurally than they would individually >

Meng: If we can model these structural interactions we could potentially build systems that suggest optimal passing strategies based on real-time defensive geometry >

Jane: It moves us away from just looking at whether a pass led to a goal or possession toward understanding the structural influence of the pass on the entire defensive configuration >

Tom: And that’s where we’re heading next looking at how we can use those structural insights to predict the next few actions in a match >

Jane: That’s where we're heading next looking at how we can use those structural insights to predict the next few actions in a match >

Lu: And I think it's important for us to think about how this framework could feed into other areas of sports science, like predictive modeling of player performance based on these structural interactions >

Meng: I agree, Lu, connecting the structure of a pass to strategy is where the real engineering value lies >

Lalam: It’s a deeper way for us to analyze football by giving us that physical understanding under pressure that we just can't get from looking at results alone >

Conclusion: Tom: So we're wrapping up on "Structural Pass Analysis in Football: Learning Pass Archetypes and Tactical Impact from Spatio-Temporal Tracking Data." We saw how they use structural metrics to quantify how passes physically reshape a defense.

Jane: Exactly. It moves us away from just counting passes to seeing the actual structural influence of that pass on the defensive configuration across the whole pitch.

Lu: It’s a pretty powerful way to model how passes reshape defensive organisation, which complements those existing action valuation models we've been using <ref:2603.28916#pg2>.

Meng: I think the real utility is seeing those team-level style representations, which use a two-dimensional map to categorize teams based on structural patterns like space expansion versus circulatory destabilisation <ref:2603.28916#pg5>.

Tom: And on the player level, they found that build-up defenders are key drivers of this structural progression, and players with high cumulative TIV are predominantly central defenders <ref:2603.28916#pg4>.

Jane: They also defined how much two players amplify each other structurally by looking at delta T IV p→r equals T IV p→r minus T IV p <ref:2603.28916#pg4>.

Lu: That's a concrete way to look at player synergy; we can quantify how much better two players work together structurally than they would individually <ref:2603.28916#pg4>.

Meng: If we can model these structural interactions, we could potentially build systems that suggest optimal passing strategies based on real-time defensive geometry <ref:2603.28916#pg4>.

Jane: It moves us away from just looking at whether a pass led to a goal or possession, toward understanding the structural influence of the pass on the entire defensive configuration <ref:2603.28916#pg1>.

Lu: It moves the discussion toward modelling how passes reshape defensive organisation, which is a powerful way to look at spatio-temporal sports data <ref:2603.28916#pg2>.

Meng: I think it’s the bridge between raw tracking data and actionable tactical understanding we need to build <ref:2603.28916#pg4>.

Lalam: I see this as a way to improve the culture around sports analysis by giving us a deeper, more physical understanding of how complex human interactions play out under pressure <ref:2603.28916#pg4>.

Tom: That's what we talked about today on structural pass analysis, using metrics like TIV to understand how passes affect defensive geometry and how teams organize themselves structurally in football.

Jane: And that’s where we’re heading next, looking at how we can use those structural insights to predict the next few actions in a match <ref:2603.28916#pg4>.

Cardiff University · Dead Ball Analytics Limited

cs.LG, cs.CY

Submitted: 2026-03-30

Updated: 2026-10-14

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 82/100

The gist: The gist: Structural representations derived from tracking data can reveal interpretable tactical patterns in football by modelling how passes reshape defensive organisation, which complements

Key concepts

Line Bypass Score (LBS): This metric measures how much a pass penetrates the opponent's defensive lines. It calculates the signed difference in remaining goal-side distance between where the passer and receiver are located, indicating if a pass successfully bypasses defenders.
Space Gain Metric (SGM): SGM quantifies how a pass changes the available attacking space at the receiver's location relative to a reference point. It compares the nearest opponent distance at the receiver against a baseline established by where the passer was, showing if the pass creates or reduces space.
Structural Disruption Index (SDI): SDI assesses how much a pass deforms defensive shape. It aggregates differences in summary statistics of defensive positions between when a pass is made and one second later, capturing the immediate structural impact of the action.

Terminology

Clustering — 1× in this paper · explained in 2 episodes
Tactical Impact Value — 1× in this paper

Summary

The gist: Structural representations derived from tracking data can reveal interpretable tactical patterns in football by modelling how passes reshape defensive organisation, which complements existing action valuation models.

Structural Framework and Metrics

The paper introduces a structural framework for analysing football passes based on their interaction with defensive structure. This framework derives three complementary structural metrics to quantify how passes alter the spatial configuration of defenders. These metrics are the Line Bypass Score (LBS), which quantifies the extent to which a pass penetrates defensive lines, the Space Gain Metric (SGM), which measures how a pass alters the available attacking space, and the Structural Disruption Index (SDI), which captures the deformation of defensive organisation induced by the pass. These metrics are combined into a composite measure termed Tactical Impact Value (TIV), which captures the structural influence of individual passes.

Pass Archetypes and Clustering

Based on these structural metrics, the study derives a taxonomy of pass archetypes through unsupervised clustering of structural features. The four interpretable pass archetypes identified are circulatory, destabilising, line-breaking, and space-expanding passes. The clustering algorithm groups passes that exhibit similar structural interactions with the defensive organisation. These archetypes correspond to conceptual mechanisms such as Circulatory passes, which maintain possession with minimal structural change and Line-breaking passes, which penetrate defensive lines and bypass defenders.

Tactical Impact Value (TIV) Analysis

The Tactical Impact Value (TIV) is defined as a weighted combination of the normalised structural metrics, where weights are set to equal values of 1/3. The TIV metric quantifies the structural impact of a pass on the defensive configuration. Empirical results show that passes with higher Tactical Impact Value are significantly more likely to lead to territorial progression, particularly entries into the final third and penalty box. The analysis of TIV distribution across archetypes reveals that Line-breaking passes exhibit the highest TIV values, while Circulatory passes are concentrated below zero, reflecting their limited influence on defensive organisation.

Empirical Findings and Team Styles

The analysis of matches from the 2022 FIFA World Cup demonstrates that structurally impactful passes are strongly associated with territorial progression. The spatial distribution of TIV shows that high-TIV passes frequently originate in deeper and wide areas of the defensive half, and terminate in advanced areas of the pitch. Furthermore, team-level analyses reveal distinct structural passing styles across teams. These styles are summarised by a two-dimensional structural style representation that categorises teams into quadrants corresponding to different tactical profiles, such as space expansion styles and circulatory destabilisation patterns.

Player-Level Insights

The analysis at the player level highlights the role of build-up defenders as key drivers of structural progression. Players with the highest cumulative Tactical Impact Value are predominantly central defenders, such as John Stones and Nicolás Otamendi. The analysis also identifies structurally influential passing partnerships, where the increase in structural impact for a passing pair is defined by ∆T IVp→r = T IV p→r − T IV p. For instance, the connection between Niklas Süle and Manuel Neuer appears among the most structurally impactful pairs, with high probabilities of both final-third and penalty-box entries.

Conclusion

The proposed framework demonstrates how structural representations derived from tracking data can reveal interpretable tactical patterns at multiple levels of the game, including pass archetypes, spatial progression dynamics, team-level attacking styles, and structurally influential player interactions. By focusing on how passes reshape defensive organisation, the framework complements existing action valuation methods and offers a new perspective for analysing tactical behaviour in spatio-temporal sports data. The results suggest that analysing passes through the lens of defensive structure provides valuable insights into attacking dynamics in football.

Limitations and Future Work

Several limitations should be acknowledged, including the restriction of empirical analysis to matches from a single tournament and the focus on immediate structural effects rather than full downstream value of possession sequences. Future work could extend the framework by integrating sequential models that capture how structural advantages evolve over multiple actions or use representation learning techniques like graph neural networks to learn structural representations directly from tracking data. The results suggest that analysing passes through the lens of defensive structure provides valuable insights into attacking dynamics in football.

References

[1] P. Lucey, A. Bialkowski, M. Monfort, P. Carr, I. Matthews, Quality vs quantity: Improved shot prediction in soccer using strategic features from spatiotemporal data, in: MIT Sloan Sports Analytics Conference, 2014

[2] T. Decroos, L. Bransen, J. Van Haaren, J. Davis, Actions speak louder than goals: Valuing player actions in soccer, in: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2019

[3] K. Singh, Introducing expected threat (xt), available at: https://karun.in/blog/expected-threat.html (2019)

[4] W. Spearman, Beyond expected goals, in: MIT Sloan Sports Analytics Conference, 2018

[5] J. Fernández, L. Bornn, Wide open spaces: A statistical technique for measuring space creation in professional soccer, in: MIT Sloan Sports Analytics Conference, 2018

[6] P. Rahimian, J. Davis, L. Toka, Temporal graph network framework for quantifying pass reception probabilities against defensive structures, Machine Learning 115 (1) (2026)

[7] O. Karakuş, H. Arkadaş, Through the gaps: Uncovering tactical line-breaking passes with clustering, in: Machine Learning and Data Mining for Sports Analytics, Springer, 2026

[8] L. Bransen, J. Van Haaren, M. van de Velden, Measuring soccer players’ contributions to chance creation by valuing their passes, Journal of Quantitative Analysis in Sports 15 (2) (2019)

[9] J. Fernández, L. Bornn, D. Cervone, A framework for the fine-grained evaluation of the instantaneous expected value of soccer possessions, Machine Learning 110 (2021)

[10] L. Szczepanski, I. McHale, Beyond completion rate: Evaluating the passing ability of footballers, Journal of the Royal Statistical Society: Series A 179 (2) (2016)

[11] G. Anzer, P. Bauer, Expected passes: Determining the difficulty of a pass in football using spatio-temporal data, Data Mining and Knowledge Discovery 36 (1) (2022)

[12] J. H. Hewitt, O. Karakuş, A machine learning approach for player and position adjusted expected goals in football (soccer), Franklin Open 4 (2023)

[13] A. Scholtes, O.

Improvements for AI systems

Improve pass evaluation by incorporating structural metrics to move beyond outcome-based metrics such as scoring probability or possession value, allowing AI systems to quantify how passes influence the defensive organisation of the opponent.
Develop a system capable of deriving an interpretable taxonomy of passing behaviour (circulatory, destabilising, line-breaking, and space-expanding passes) by applying unsupervised clustering to structural features derived from spatio-temporal tracking data.
Implement a composite metric called Tactical Impact Value (TIV) as a feature representation that summarizes the structural influence of a pass on the defensive configuration, enabling AI to predict territorial progression, particularly entries into the final third and penalty box.
Create player-level structural profiling by analyzing cumulative Tactical Impact Value to identify key roles, specifically noting that the players with the highest cumulative Tactical Impact Value are predominantly central defenders, highlighting their role as key drivers of structural progression.
Enable tactical style characterization by using team-level TIV maps to derive a two-dimensional structural style representation summarising the balance between circulatory play and structurally progressive passing, as well as the relative emphasis on destabilising versus space-expanding mechanisms.
Build predictive models for pass receiver outcomes by utilizing temporal graph network frameworks, such as those proposed by Rahimian et al., to predict pass receivers and passing outcomes from spatio-temporal match data.
Design a system that identifies structurally impactful passing partnerships by computing the increase in structural impact for a passing pair, defined as ∆T IVp→r = T IV p→r − T IV p, which reveals how receivers amplify the structural effectiveness of a passer’s actions.

Abstract

Football action-value approaches estimate scoring, possession, or other downstream outcome value; complementary descriptions can characterise how a pass relates to the opposition's spatial structure. We introduce an interpretable structural-passing framework using tracking and event information, with three descriptors: Line Bypass Score (LBS), a signed difference in opponents remaining goal-side between passer and receiver coordinates; Space Gain Metric (SGM), relative nearest-opponent distance at the receiver versus a floored passer reference; and Structural Disruption Index (SDI), aggregate differences in defensive-shape summaries between pass time and the nearest sampled state one second later. The metrics are transformed and standardised before combination in Tactical Impact Value (TIV), a fixed-weight structural composite. We apply the framework to 41,078 open-play pass events with identified passer and receiver from the 2022 FIFA World Cup. The selected K=4 partition provides an interpretable operational taxonomy of four profiles: Circulatory, Destabilising, Line-breaking, and Space-expanding. The profiles show different structural metric signatures, and the reported territorial-crossing and stored outcome indicators vary descriptively across profiles and TIV strata. Team, spatial, player, and passer--receiver analyses illustrate how the structural representation can be summarised at multiple levels. The findings position the framework as a complement to outcome-oriented action-value analysis, offering football analysts interpretable structural evidence to support tactical interpretation and guide further investigation.