Assessing and Visualizing Principles of Play in Soccer

uben Filipe Rocha

1 a

and Rui J. Lopes

1,2 b

ISCTE-Instituto Universit

ario de Lisboa, Lisbon, Portugal

Instituto de Telecomunicac¸

oes, Lisbon, Portugal

Keywords:

Performance Analysis, Visualisation Tool, Soccer, Principles of Play, StatsBomb Data.

Abstract:

Recent advances in soccer analytics have signiﬁcantly expanded the use of spatial and temporal data to un-

derstand tactical behaviour. However, many visualisation tools remain limited in their ability to contextualize

these behaviours according to established principles of play. This paper presents a context-aware visualisation

framework that leverages StatsBomb Open Data to identify, illustrate, and interpret tactical patterns in pro-

fessional soccer. The tool enables analysts to explore match phases through key tactical dimensions such as

compactness, pressing, width, support, and penetration, using event-level data enriched with positional data

(360 freeze frames). Unlike generic dashboards or statistical summaries, the proposed system integrates spa-

tial relationships and collective movements, offering a more accurate representation of team behaviour across

different match contexts. By combining visual methods with tactical theory, the tool supports coaches and an-

alysts in identifying the match principles of play, thereby facilitating a deeper understanding of performance

dynamics.

1 INTRODUCTION

In today’s soccer, vast amounts of data are collected

during matches, including player tracking data and

event records, like passes, dribbles and shots. How-

ever, raw data must be transformed into actionable

insights that accurately reﬂect the match context and

can be interpreted by the coaching staff to inform their

tactical decisions (Liu, 2022).

Data visualisation techniques convert complex nu-

merical and spatial data into visual representations

that highlight features or patterns relevant to tactical

analysis. These methods assist analysts and coaches

in identifying recurring behaviours, comparing player

performance in speciﬁc contexts, and improving com-

munication within the technical team (Perin et al.,

2018; Liu, 2022). For example, visual tools can il-

lustrate how a team constructs attacking sequences,

the zones where players exert spatial dominance, or

the structural variations in formation across differ-

ent match phases (Krishnamurthy and Nanda, 2021;

Bauer et al., 2023).

Over time, visualisation approaches have pro-

gressed from static or aggregated outputs like

heatmaps and radar charts to more advanced, inter-

https://orcid.org/0009-0001-0129-5977

https://orcid.org/0000-0002-8943-0415

active systems. These include animated movement

trajectories, dynamic passing networks, and forma-

tion models that evolve throughout the match (Sacha

et al., 2017; Krishnamurthy and Nanda, 2021; Janet-

zko et al., 2014).

The core function of these tools is to extract and

interpret features and tactical structures from the ﬂow

of the match. This includes recognizing repeated

movement patterns, coordinated passing sequences,

and adaptive behaviours (for example, responses to

changes in possession status). Some systems can

even detect which structural breakdowns contributed

to match outcomes by contrasting successful and un-

successful performance proﬁles (Krishnamurthy and

Nanda, 2021).

Nonetheless, several challenges remain. Notably,

there is no standardization in how visualisations rep-

resent time and space, which are fundamental to un-

derstanding the dynamics of soccer (Sotudeh, 2025;

Bauer et al., 2023). Consequently, many tools fail to

incorporate core tactical concepts such as the princi-

ples of play, offering visual summaries that lack the

contextual depth required for high-level analysis.

In this context, this paper examines existing visu-

alisation systems and evaluates how effectively they

support the identiﬁcation and interpretation of tacti-

cal patterns within professional soccer matches. Fur-

Rocha, R. F. and Lopes, R. J.

Assessing and Visualizing Principles of Play in Soccer.

DOI: 10.5220/0013716200003988

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 13th International Conference on Sport Sciences Research and Technology Support (icSPORTS 2025), pages 90-99

ISBN: 978-989-758-771-9; ISSN: 2184-3201

thermore, this paper presents a solution to detect tac-

tical principles of play in soccer through the combi-

nation of event-level data and 360 freeze frames from

the StatsBomb Open Data model. The term ”princi-

ples of play” refers to a set of fundamental tactical

guidelines that govern how teams organize and be-

have collectively during a soccer match such as press-

ing, compactness, width, support, and penetration,

which emerge throughout different phases of play.

2 RELATED WORK

Over the past decade, several researchers have pro-

posed advanced methods to enhance tactical insights.

Among them are approaches relying on static or

aggregated representations such as heatmaps, radar

charts, and passing network diagrams. While effec-

tive for summarizing performance data, these visual-

isations fall short in capturing the full tactical com-

plexity of the match, particularly the spatial and tem-

poral dynamics that underpin team behaviour (Janet-

zko et al., 2014; Krishnamurthy and Nanda, 2021).

Perin et al. (2018) presented a comprehensive re-

view of sports data visualisation, classifying existing

work by its goals, target users, and data complex-

ity. They also emphasize that working closely with

coaches and analysts during the design process is es-

sential to create tools that are not only functional, but

also practical and relevant in real-world settings. In

terms of goals, the authors give examples such as

design studies with sports commentators, user need

analysis, algorithm and model development, integra-

tion into real-world environments, and new visual or

interaction techniques recommendations. With re-

gards to target users, the work distinguishes between

analysts and coaches and athletes (in practical eval-

uation tasks, usually), fans and the general public

(in turn accessed via infographics), and journalists

or data specialists. Finally, the authors deﬁne three

categories of sports data: box score data (event data

per individual match), tracking data (spatio-temporal

movement data), and meta-data (contextual data like

player proﬁles or stadium features).

The later classiﬁcation criteria taps into the core of

our proposal, namely in the fact that the vast majority

of visualization tools fall into only one of these cat-

egories, hindering their capability to unveil features

or patterns actionable into principles of play related

tasks.

In addition to this classiﬁcation, Table 1 groups

seminal studies on soccer analytics research based on

the following criteria: visualisation type and tactical

analysis level. This overview enables us to distinguish

between static and dynamic approaches, the level of

user engagement, and the tactically investigated gran-

ularity each approach focuses on (individual or team

level).

Most visualization tools use positional data. Liu

(2022) explored geo-visualisation techniques for the

analysis of player positioning data. This study uses

tools such as Python, R, and GIS platforms to visual-

ize spatial and spatio-temporal patterns, with empha-

sis on position-level tactical dynamics. Liu (2022) in-

vestigated how player tracking data can be animated

and visualised using GIS tools and web technologies

such as Python, HTML5, and Tableau. These tools

add spatial and spatio-temporal context to soccer an-

alytics, which is essential for understanding tactical

positioning. As illustrated in Figure 1, simple player

positional data can be combined with other more

complex positional features such as convex hulls and

Voronoi diagrams.

Figure 1: Interactive visualisation of player positions and

convex hull selection interface. Adapted from Liu (2022).

In other research, Sacha et al. (2017) introduced

a dynamic visual abstraction method that enables

the analysis of collective player movement patterns.

Their approach emphasizes the importance of move-

ment over time, capturing dynamic tactical structures

such as defensive lines and team pressing. Sacha et al.

(2017) also propose visual abstraction techniques that

simplify player movement patterns (see Figure 2),

making it easier for analysts to observe and interpret

tactical behaviour over time. In this application, in-

stead of a single snapshot of the positional status of

the match, the aggregated positions along a certain

time interval are presented.

Other tools present values associated usually with

player or team performance. These values can be su-

perimposed or embedded in pitch diagrams (notable

examples are pass networks and positional heat maps)

or use other visual references (notable examples are

radar charts). One example of the later is illustrated

in Figure 3 from Liu (2022) where these charts are

Assessing and Visualizing Principles of Play in Soccer

Table 1: Comparison of visualisation approaches in soccer analytics literature.

Author (Year) Visualisation Type Level of Tactical Analysis

Liu (2022) Geo-visualisation Position Level

Sacha et al. (2017) Dynamic Team Level

Janetzko et al. (2014) Static + Exploratory Team Level

Krishnamurthy and Nanda (2021) Interactive Dashboard Team Level

Bauer et al. (2023) Static Analysis Team Level

Delibas et al. (2019) Visual Analytics Tool Team Level

Sotudeh (2025) Conceptual/Survey Both

Figure 2: Visual representation of player movements

through dynamic abstraction techniques. Adapted from

Sacha et al. (2017).

Figure 3: Radar chart illustrating statistical metrics for

player performance. Adapted from Liu (2022).

used to represent different performance metrics of a

speciﬁc player.

Several other studies have adopted varied visual-

ization approaches, such as an interactive dashboards

developed to visualize performance metrics at both

player and team levels, using radar charts, heatmaps,

and passing networks to analyse individual contri-

bution and collective strategy (Krishnamurthy and

Nanda, 2021). In addition, Janetzko et al. (2014)

proposed a feature-oriented visual analytics frame-

work bringing together static and exploratory visual-

isations, combining team measures enabling the dis-

covery of patterns by using interactive ﬁltering and

brushing capabilities. Furthermore, an interactive

exploratory analysis tool was introduced by Delibas

et al. (2019) to investigate soccer event data through

customizable parameters, supporting team-level tacti-

cal assessments by linking event-based data with vi-

sual analytics features. Taken together, these con-

tributions demonstrate that different forms of visu-

alisation, from dashboards to exploratory analysis

frameworks, play an important role in helping ana-

lysts identify relevant patterns, understanding collec-

tive behaviour, and supporting decision-making pro-

cesses in football analytics.

In recent years, more dynamic and interactive vi-

sualisation systems have emerged. One notable ex-

ample is SoccerMetrics, an open-source framework

that identiﬁes key players, analyses strategies against

different opponents, and highlights tactical changes

associated with unsuccessful match outcomes (Krish-

namurthy and Nanda, 2021).

One important area in development is recognis-

ing how formations evolve across different phases of

the match. Sotudeh (2025) presented a comprehen-

sive review on tactical formation identiﬁcation. Their

analysis goes from position-level and team-level per-

spectives, discussing data preprocessing, clustering,

and template-based methods, while also highlighting

methodological limitations and future directions.

Studies focused on tactical formations have also

gained attention in recent years. Bauer et al. (2023)

utilized static visual analysis to contextualize team

formations, providing a conceptual framework to un-

derstand the variability of tactical structures under

different match conditions. In addition, they em-

ployed a combination of convolutional neural net-

works and tracking data to segment matches into

phases and detect formation changes beyond static la-

bels like “4-4-2”. Similarly, Sotudeh (2025) reviewed

principles for identifying formations and argued that

many current methods oversimplify them by relying

on spatial averages and clustering, often ignoring the

temporal context of the match and the emergence of

dynamic behaviours. Together, these studies empha-

icSPORTS 2025 - 13th International Conference on Sport Sciences Research and Technology Support

sise the importance of moving beyond static represen-

tations to approaches that capture the ﬂuid and dy-

namic nature of formations throughout a match.

Even with recent progress, many tools still rely

heavily on aggregated data or static views, falling

short in representing the ﬂuid and complex dynamics

of team play (Sotudeh, 2025; Bauer et al., 2023). Tra-

ditional heatmaps and static pass maps are often in-

sufﬁcient to represent interaction between players or

the effect of context like match state or opponent strat-

egy (Janetzko et al., 2014). Moreover, several tools

rely heavily on event data, which lacks full spatial

context, and under-use the rich potential of tracking

data when not combined with tactical insights (Perin

et al., 2018).

Other options are commercial platforms, such as

Stats Edge Analysis, that provide analysts with the ca-

pability to examine tactical behaviour through inter-

active visual interfaces. These systems also provide

synchronized positional and event information access,

allowing one to ﬁlter periods of play (e.g., attack, de-

fence, transition), select speciﬁc match situations, and

project team arrangements onto a tactical board. One

key component of this technology is the ability to au-

tomatically measure spatial statistics—such as team

length, width, and compactness—that are directly ap-

plied to pitch maps for simpler interpretation.

In Stats Edge Analysis, for instance, people can

distinguish average positioning of a player across se-

lected sequences of matches and assess coordinated

collective behaviours like coordination of pressing,

line compactness, and block height. The system also

supports custom queries by ball position and player

role and offers dynamic frame-by-frame playback to

put tactical trends in context. This spatial-temporal

data-driven technology enables pre-match and post-

match analysis, opponent team scouting, and detec-

tion of structural weaknesses or strengths.

Although existing visualisation tools offer useful

summaries of performance data, they often fail to cap-

ture the underlying tactical complexity of soccer. Tac-

tical principles are deﬁned as spatial and temporal

patterns that recur across matches. They help de-

termine how teams act throughout different phases

of the match. However, current tools rarely account

for these principles, limiting their ability to support

training design, match analysis, and tactical decision-

making (Table 2). Consequently, there is a need for

visualisation systems explicitly guided by principles

of play, which can more effectively inform coaches

and analysts.

3 StatsBomb DATA

StatsBomb Open Data is one of the most widely

used open datasets in soccer analytics research. It

provides detailed match-level data in JSON format,

freely accessible for academic and non-commercial

use. The data covers various professional competi-

tions and includes structured information about com-

petitions, matches, player line-ups, match events, and

spatial context via 360 freeze frames. Each data type

is organized in separate ﬁles, allowing modular and

layered analysis of soccer performance.

3.1 Competitions and Matches

The competitions.json ﬁle lists all available com-

petitions and seasons, including metadata such as

country, sex, and competition IDs (StatsBomb,

2019a). Each competition is associated with one or

more seasons, and matches are grouped under these

seasons using the matches.json ﬁles (StatsBomb,

2019d). Each match entry contains information such

as team names, match date, score, stadium, referee,

and a unique match id used to access further data

ﬁles.

3.2 Line-Ups

The lineups.json ﬁles describe the starting eleven

and substitutes for each team in a given match. For

each player, the dataset includes their ID, full name,

nickname, jersey number, and nationality (Stats-

Bomb, 2019c). This data is crucial for linking actions

and roles to speciﬁc individuals during the analysis.

3.3 Events

The event data is the analytical core of the dataset.

Stored in events.json ﬁles, it includes all the on-

the-ball actions performed during a match. Each

event is timestamped and linked to a player, team,

and spatial location on the pitch (StatsBomb, 2019b).

The event types cover actions such as passes, drib-

bles, shots, fouls, interceptions, pressures, and

more. Additional context is provided through tags

like under_pressure, carry, counterpress, and

play_pattern, which support more detailed tactical

analysis.

StatsBomb describe in their open-data documen-

tation that every on-the-ball event is tagged manually

by analysts watching match videos. They follow a

protocol that deﬁnes the type of event, who was in-

volved and its time and pitch location.

Assessing and Visualizing Principles of Play in Soccer

Table 2: Core tactical principles, their deﬁnitions and application context.

Type Principle Description

Team

Without Ball

Pressing Immediate pressure on the ball carrier to force a mistake or regain possession.

Compactness Players stay close together, reducing space and blocking passing options.

Coverage and Balance Maintain defensive structure by covering space left by the pressing player.

Containment Delay the opponent’s advance without trying to win the ball immediately.

Control Control tempo and dangerous zones with structured positioning.

Team

With Ball

Penetration Break defensive lines using passes, dribbles, or forward runs.

Space Creation Stretch the opponent’s defence through width, movement, or decoys.

Player Movement Off-ball actions that unbalance opponents and open new options.

Support Stay close to the ball carrier to provide safe passing options.

Creativity Unpredictable actions like dribbles or disguised passes to destabilise defences.

3.4 Positional Data: 360 Freeze Frames

Positional data is provided with more detail, i.e., in-

volving more players, in the 360 freeze frame data

layer of the dataset (StatsBomb, 2021). Each freeze

frame captures a snapshot of the pitch at the moment

of a key event (e.g., a pass or shot). It includes the

locations of nearby players in the visible area of the

pitch based on a broadcast camera. Each player in the

frame is tagged with their position on the ﬁeld and

whether they are a teammate or opponent.

3.5 Limitations

Although StatsBomb Open Data is highly detailed it

presents some limitations. Notably, the 360 freeze

frame positional data is not available for all competi-

tions and, when available, it does not provide contin-

uous tracking data for all 22 players, but rather static

snapshots around events (i.e., there is no data between

events) and event focus players.

Due to the limited capture scope of the broadcast

cameras, positional data is not provided for all players

in the pitch but only for those in the visible are of

the camera (as illustrated in Figures 4 to 8). When

players are out of the broadcast camera scope, they

are temporarily not included in the freeze frame data

until they reappear in subsequent frames. Moreover,

each 360 freeze frame data only identiﬁes the player

performing the event (i.e., the actor) other players are

tagged simply as from the same or opposing team to

the actor, therefore limiting player-speciﬁc tracking

analysis.

However, the data always captures the ball, the

ball carrier, and the surrounding context of the ac-

tion, which ensures that event-based metrics such as

passes, shots, dribbles, pressures and duels are not af-

fected. Although hampering metrics that depend on

the full pitch coverage and identiﬁcation of all players

(e.g., team width, compactness, and spatial control),

Statsbomb positional data still allows for the study of

general spatial behaviour and team structure.

4 VISUALISATION TOOL

The developed visualisation tool is supported by neu-

ral networks and enables the exploration of tactical

patterns in soccer with a focus on supporting coaches

and analysts in understanding key principles of play.

Built on top of StatsBomb Open Data, the tool allows

the user to ﬁlter, extract, and visualise sequences of

play where tactical behaviours emerge—such as com-

pactness, width, support, and pressing.

As discussed earlier, tactical principles (see Table

2) are recurring patterns of collective behaviour that

guide how teams act during different phases of the

match. They help structure training, inform match

analysis, and support decision-making by coaches

and analysts. Tactical principles are typically divided

into two categories: with-ball principles, which gov-

ern attacking behaviour, and without-ball principles,

which relate to defensive organisation.

4.1 Tactical Principles Without Ball

Pressing

Pressing involves applying immediate pressure on the

opponent with the ball, especially during transitions,

to force mistakes or rushed decisions. It aims to dis-

rupt the opponent’s rhythm and recover possession

quickly, preferably near the goal (Rein and Memmert,

2016; Forcher et al., 2024).

Pressure effectiveness can be measured by the op-

ponent’s response time:

= t

pressure

−t

decision

(1)

Where t

pressure

and t

decision

correspond to the time

when pressure is applied and the time the opponent

reacts. These two timestamps are relevant not only

for computing the interval, t

, but also as time bound-

aries for visualization representation of this principle.

Additionally, the number of passes forced or inter-

icSPORTS 2025 - 13th International Conference on Sport Sciences Research and Technology Support

cepted under pressure can be evaluated as:

∑

i=1

, P

(

1 if forced or intercepted

0 otherwise

(2)

Here, P

and P

correspond to the total pressured

passes and the number of individual pass outcome.

Figure 4 from the developed visualization tool, il-

lustrates pressure on a pass action.

Figure 4: Pressure: Player under pressure, highlighted with

a red circle. The white line indicates the destination of the

next pass. The red polygon represents the visible area of the

broadcast camera.

Compactness

Compactness refers to the team’s ability to remain

close together, minimising spaces between players

and lines. This limits the opponent’s passing op-

tions and enhances the collective defensive response

(Plakias et al., 2024; Meerhoff et al., 2019).

A useful metric is the average distance between

teammates:

d =

∑

i, j=1,i̸= j

− x

)

+ (y

− y

)

(3)

where x

, y

and x

, y

are player i and j spatial coor-

dinates; and

d is the average pairwise distance.

We can also calculate how much space the team

occupies on the ﬁeld:

ﬁeld

team

(4)

where A

ﬁeld

corresponds to the total ﬁeld area and

team

the area covered by a particular team.

And assess their compactness via the Stretch In-

dex:

max

i, j

)

d(i, j)

(5)

here max(d

i, j

) is the maximum distance between

player pairs and

d(i, j) its mean value.

This principle of play is illustrated in Figure 5

from the visualization tool. Here the possible pairs

from each team are represented using the Delaunay

triangulation.

Figure 5: Compactness: Visual representation of team com-

pactness using Delaunay triangulation.

Coverage and Balance

This principle ensures that the team maintains struc-

ture when pressing by repositioning nearby players to

cover the space left by the presser. It provides de-

fensive balance and avoids opening exploitable gaps

(Plakias et al., 2024; Herold et al., 2019).

The proximity between pressing and covering

players is essential and given by the Euclidean dis-

tance between pressing and covering players:

d =

− x

)

+ (y

− y

)

(6)

where x

, y

and x

, y

are the coordinates of pressing

and covering players.

We can quantify the fraction, B of players main-

taining their defensive roles:

B =

def

total

(7)

where N

def

is the number of players in defensive roles

and N

total

the total number of players.

And evaluate team balance using its centroid:

∑

i=1

, C

∑

i=1

(8)

where C

, C

is the team centroid coordinates; and

, y

the coordinates of each player of the focus team.

Containment

Containment slows down the opponent’s offensive

momentum without direct ball recovery attempts. It

forces lateral or backward play and provides time

Assessing and Visualizing Principles of Play in Soccer

for team reorganization (Rein and Memmert, 2016;

Forcher et al., 2024).

The defender’s reaction time, t

is given by:

= t

decision

−t

pressure

(9)

where t

decision

and t

pressure

correspond to the oppo-

nent’s action defender’s approach timestamps.

Passes blocked or intercepted during containment

are measured as:

I =

∑

i=1

, I

(

1 if pass i is intercepted

0 otherwise

(10)

where I is the total interceptions and I

= binary indi-

cator of interception (or not) for pass i.

Control

Control reﬂects the ability to dictate space and rhythm

defensively, usually by positioning to block danger-

ous zones and slow the match’s tempo Nouraie et al.

(2023).

The controlled area, A

, can be estimated by:

= d × P

(11)

here d stands for the defender’s proximity and P

the

number of blocked passes.

Possession time while maintaining defensive con-

trol, t

is:

= t

possession

−t

interception

(12)

where t

possession

and t

interception

correspond to the start

of controlled possession and interception timestamps.

4.2 Tactical Principles with Ball

Penetration

Penetration is the ability to move the ball past defen-

sive lines using passes, dribbles, or forward runs and

is key to disrupting defensive structures and creating

chances (Bauer and Anzer, 2021; Rein and Memmert,

2016).

The forward progress achieved, d, via pass or ball

carry, is given by:

d =

− x

)

+ (y

− y

)

(13)

where x

, y

and x

, y

are the initial and ﬁnal posi-

tions.

In addition to the pitch progress we can also count

the number of successful penetrative passes, P

as:

∑

i=1

, P

(

1 if line-breaking pass is completed

0 otherwise

(14)

where P

is an indicative binary function associated to

progressive pass completion. Figure 6 illustrates the

representation of a progressive pass situation.

Figure 6: Example of a penetrative pass breaking the de-

fensive line.A red dashed line with an arrow represents the

penetrating pass.The faded point indicates the player who

will receive the ball in the next frame.

Space Creation

Space creation stretches the opposing defence using

width, movement, and decoy runs. It helps open

passing lanes and create favourable match-ups (Stival

et al., 2023; Caldeira, 2023; Meerhoff et al., 2019).

The available space generated, A

is:

= A

total

− A

occupied

(15)

where A

total

is full ﬁeld area and A

occupied

the used

area.

The Stretch Index, S

also applies here:

max

i, j

)

d(i, j)

(16)

where max(d

i, j

) is the maximum inter-player distance

and

d(i, j) is the average inter-player distance.

Player Movement

This principle involves intelligent off-ball movement

to unbalance defences, either by dragging defenders

away or creating new passing lanes (Bauer and Anzer,

2021; Rein and Memmert, 2016; Plakias et al., 2024).

We can measure the frequency of such move-

ments, M

as:

∑

i=1

(

1 movement creates passing option

0 otherwise

(17)

here M

is a binary indicator function for meaningful-

ness of movement i.

Support

Support ensures that the ball carrier has immediate

and safe passing options nearby, enabling ﬂuid and

continuous attack (Muacho et al., 2022; Bauer and

Anzer, 2021; Rein and Memmert, 2016).

icSPORTS 2025 - 13th International Conference on Sport Sciences Research and Technology Support

The total number of supported passes, P

is:

∑

i=1

, P

(

1 if pass is received

0 otherwise

(18)

here P

is a binary indicator function if pass i had sup-

port (or not).

In addition to the number of supported passes, the

distance, d

between the ball carrier and a support

player can also be computed as:

player

− x

pass

)

+ (y

player

− y

pass

)

(19)

where x

player

, y

player

and x

pass

, y

pass

are the positional

coordinates for the player with ball and its support

player. Figure 7 illustrates this concept.

Figure 7: Support, contextual positioning of players, high-

lighting available passing lanes and supporting options.

Creativity

Creative actions introduce unpredictability through

dribbles, disguised passes, or unusual movements.

These actions break rigid defensive lines and sur-

prise opponents (Rein and Memmert, 2016; Bauer

and Anzer, 2021; Stival et al., 2023).

We track the number of successful creative ac-

tions, J

by:

∑

i=1

, J

(

1 if creative action succeeds

0 otherwise

(20)

where J

is a binary indicator function accounting if

creative action i succeeded (or not).

The ground covered in these creative moments can

be assessed by distance d

− x

)

+ (y

− y

)

(21)

where x

, y

and x

, y

correspond to the start and end

coordinates for the creative action.

As illustrated in the previous ﬁgures, rather than

focusing purely on statistical metrics, the tool is de-

signed to represent how and where these tactical prin-

ciples occur on the pitch. By combining event data

with 360 freeze frames, it becomes possible to explore

the spatial context of each action, including the posi-

tioning of teammates, opponents, and the visible area

perceived by the player at the time of the event.

Notably, the use of timestamped notational data

enables the clear deﬁnition of time boundaries for the

representation of each principle of play. In Figure

8 different metrics are represented at the two time

boundaries of the selected event (a pass).

The aim is to go beyond isolated data points and

enable the analysis of movement, positioning, and in-

teraction patterns within the collective behaviour of

the team. The system helps identify when certain

principles are respected or broken, and under what

match conditions these patterns tend to emerge, offer-

ing valuable support for tactical evaluation and coach-

ing decision-making.

5 CONCLUSIONS

Building upon the review of existing tools, this pa-

per examined the evolution of soccer data visualisa-

tion, pointing out the transition from static statistical

outputs to more sophisticated, context-aware systems.

While the ﬁeld has advanced considerably, many ex-

isting tools still fall short in representing the tactical

nuances of the match—particularly those related to

the principles of play that guide collective team be-

haviour.

To address these limitations, we developed a visu-

alisation prototype based on StatsBomb Open Data,

designed to capture and interpret tactical patterns such

as compactness, pressing, width, support, and pene-

tration. The tool integrates spatial and temporal con-

text using event data and 360 freeze frames, allowing

analysts to explore not just what happens on the pitch,

but how and why those behaviours emerge in different

phases of play.

Rather than relying on isolated metrics or pre-

deﬁned statistics, the system visualizes the position-

ing, interactions, and movements that underpin team

dynamics. This enables analysts and coaches to eval-

uate whether tactical principles are being respected,

where they break down, and under what match condi-

tions such patterns tend to appear.

Match-related factors such as opponent, score,

referee and lineups are integrated as contextual infor-

mation and can be used to ﬁlter or group the visualisa-

tions. This allows analyses across multiple matches,

Assessing and Visualizing Principles of Play in Soccer

Figure 8: Two consecutive moments.

for example by aggregating positional data to build

heatmaps or computing average metrics for speciﬁc

opponents, score-lines or lineups.

The tool has been developed in close collabora-

tion with an elite-level coach, who provides continu-

ous expert feedback as a form of practical validation.

Ultimately, the approach bridges the gap between

raw match data and applied tactical insight, offering a

practical resource for those seeking to enhance perfor-

mance evaluation, coaching feedback, and decision-

making in professional soccer. Although using Stats-

Bomb open data the principles and tools proposed are

of general validity and can be used, with proper adap-

tation and caution, to other data sets.

ACKNOWLEDGEMENTS

We would like to thank Nelson Caldeira for shar-

ing valuable insights into the key aspects that soccer

coaches prioritise when analysing team performance.

Rui J. Lopes was partly supported by FCT - Fundac¸

para a Ci

encia e Tecnologia, I.P. by project refer-

ence UIDB/50008/2020, and DOI identiﬁer https:

//doi.org/10.54499/UIDB/50008/2020 awarded

to Instituto de Telecomunicac¸

oes.

REFERENCES

Bauer, P. and Anzer, G. (2021). Data-driven detection of

counterpressing in professional football. Data Mining

and Knowledge Discovery, 35:2009–2049.

Bauer, P., Anzer, G., and Shaw, L. (2023). Putting team for-

mations in association football into context. Journal

of Sports Analytics, 9(1):39–59.

Caldeira, N. C. A. (2023). Functional space-time properties

of team synergies in high-performance football. Sen-

sors, 23(5):1284.

Delibas, E., Uzun, A., Inan, M. F., Guzey, O., and Cakmak,

A. (2019). Interactive exploratory soccer data analyt-

ics. INFOR, 57(2):141–164.

Forcher, L., Beckmann, T., Wohak, O., Romeike, C., Graf,

F., and Altmann, S. (2024). Prediction of defensive

success in elite soccer using machine learning – tac-

tical analysis of defensive play using tracking data

and explainable ai. Science and Medicine in Football,

8(4):317–332.

Herold, M., Goes, F., Nopp, S., Bauer, P., Thompson, C.,

and Meyer, T. (2019). Machine learning in men’s pro-

fessional football: Current applications and future di-

rections for improving attacking play. International

Journal of Sports Science and Coaching, 14(6):798–

817.

Janetzko, H., Sacha, D., Stein, M., Schreck, T., Keim, D. A.,

and Deussen, O. (2014). Feature-driven visual analyt-

ics of soccer data. In 2014 IEEE Conference on Visual

Analytics Science and Technology (VAST), pages 13–

22.

Krishnamurthy, P. and Nanda, N. (2021). Dynamic data

icSPORTS 2025 - 13th International Conference on Sport Sciences Research and Technology Support

visualization for soccer performance metrics. In

2021 6th International Conference for Convergence in

Technology (I2CT), pages 1–6.

Liu, N. (2022). Geovisualisation of football players move-

ment. Master’s thesis, Palack

y University Olomouc

and Paris Lodron University Salzburg. Erasmus

Mundus Joint Master in Digital Earth.

Meerhoff, L. A., Goes, F. R., Leeuw, A.-W. D., and Knobbe,

A. (2019). How dots behave in two different pitch

sizes? analysis of tactical behavior based on posi-

tion data in two soccer ﬁeld sizes. Communications in

Computer and Information Science, 1168:235–246.

Muacho, H., Ribeiro, R., and Lopes, R. (2022). The elusive

features of success in soccer passes: A machine learn-

ing perspective. In 10th International Conference on

Sport Sciences Research and Technology Support (ic-

SPORTS 2022), pages 110–116.

Nouraie, M., Eslahchi, C., and Baca, A. (2023). Intelligent

team formation and player selection: A data-driven

approach for football coaches. Applied Intelligence,

53:30250–30265.

Perin, C., Vuillemot, R., Stolper, C. D., Stasko, J. T., Wood,

J., and Carpendale, S. (2018). State of the art of

sports data visualization. Computer Graphics Forum,

37(3):663–686.

Plakias, S., Kokkotis, C., and Tsaopoulos, D. E. (2024).

Can artiﬁcial intelligence revolutionize soccer tactical

analysis? Trends in Sport Sciences, 31(3).

Rein, R. and Memmert, D. (2016). Big data and tactical

analysis in elite soccer: Future challenges and oppor-

tunities for sports science. SpringerPlus, 5:1410.

Sacha, D., Al-Masoudi, F., Stein, M., Schreck, T., Keim,

D. A., Andrienko, G., and Janetzko, H. (2017). Dy-

namic visual abstraction of soccer movement. Com-

puter Graphics Forum, 36(3):305–315.

Sotudeh, H. (2025). The principles of tactical formation

identiﬁcation in association football (soccer) – a sur-

vey. Frontiers in Sports and Active Living, 6:1512386.

StatsBomb (2019a). Statsbomb open data speciﬁca-

tion: Competitions v2.0.0. https://github.com/

statsbomb/open-data. Accessed May 2025.

StatsBomb (2019b). Statsbomb open data speciﬁcation:

Events v4.0.0. https://github.com/statsbomb/

open-data. Accessed May 2025.

StatsBomb (2019c). Statsbomb open data speciﬁcation:

Lineups v2.0.0. https://github.com/statsbomb/

open-data. Accessed May 2025.

StatsBomb (2019d). Statsbomb open data speciﬁcation:

Matches v3.0.0. https://github.com/statsbomb/

open-data. Accessed May 2025.

StatsBomb (2021). Statsbomb open data speciﬁcation: 360

frames v1.0.0. https://github.com/statsbomb/

open-data. Accessed May 2025.

Stival, L., Pinto, A., dos Santos Pinto de Andrade, F., San-

tiago, P. R. P., Biermann, H., da Silva Torres, R., and

Dias, U. (2023). Using machine learning pipeline to

predict entry into the attack zone in football. PLoS

ONE, 18(1):e0265372.

Assessing and Visualizing Principles of Play in Soccer