See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320260631

A Methodology for the Analysis of Soccer Matches Based on PageRank

Centrality

Chapter · October 2017

DOI: 10.1007/978-3-319-63907-9_16

CITATIONS READS

4 2,346

4 authors, including:

Julio Rojas-Mora Nicolas Valdebenito

Temuco Catholic University Universidad Austral de Chile
74 PUBLICATIONS 300 CITATIONS 8 PUBLICATIONS 415 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

EnviroInfo 2017 (http://www.enviroinfo2017.org) View project

DID S-2014-27. Improvement of the mechanical behavior of bituminous mixtures with recycled asphalt (RAP) through the inclusion of copper slag View project

All content following this page was uploaded by Julio Rojas-Mora on 23 September 2018.

The user has requested enhancement of the downloaded file.

Chapter 16
A Methodology for the Analysis of Soccer
Matches Based on PageRank Centrality

Julio Rojas-Mora, Felipe Chávez-Bustamante, Julio del Río-Andrade,

and Nicolás Medina-Valdebenito

Abstract Data analysis in sports has adopted many different approaches given its
usefulness in quantitative and objective management. Several advances have been
made considering the researches and technologies that have been developed up until
now. It is possible to find many complex methodologies of sport performance analy-
sis in order to have as much as information as possible to achieve success. Therefore,
a wide variety of options are available for sport managers, coaches or anyone inter-
ested, including advances on information systems, data mining, machine learning
and motion analysis. However, the cost of these powerful methodologies induces
the search of cheaper techniques based on basic but proper notation methodology.
The aim of this chapter is to provide an observational methodology for soccer match
analysis. When paired with PageRank as the main indicator of performance, it
allows for a deep analysis of the data and better decision-making and performance
analysis in soccer. To show some insights about the proposed model, real data from
past matches are presented and discussed. Results show graph visualization that
sum up the whole match in terms of the flows of a network modelled with passes
and recoveries from the players as weights of its edges. One implication of our
research is to be a first approach in generalizing the PageRank algorithm to soccer
team’s management, which could be extrapolated to other disciplines. It also points
to the feasibility of making a quantitative analysis for sport managers with a reasonable

J. Rojas-Mora (*)
School of Informatics, Universidad Católica de Temuco,
Rudecindo Ortega n°, 2950 Temuco, Chile
e-mail: [email protected]
F. Chávez-Bustamante
Faculty of Business and Economics, Universidad del Desarrollo,
Ainavillo 456, Concepción, Chile
e-mail: [email protected]
J. del Río-Andrade • N. Medina-Valdebenito
Business Administration Institute, Universidad Austral de Chile,
Calle Viel s/n, Valdivia, Chile
e-mail: [email protected]; [email protected]

© Springer International Publishing 2017 257

M. Peris-Ortiz et al. (eds.), Sports Management as an Emerging Economic
Activity, DOI 10.1007/978-3-319-63907-9_16
258 J. Rojas-Mora et al.

cost-benefit ratio. This analysis opens the paths to further analysis that could include
spatiotemporal variables.

Keywords PageRank • Graph theory • Social network • Observational methodology

• Centrality

16.1 Introduction

Knowing how to collect, access, retrieve and integrate information is critical to

effective performance analysis and decision-making processes (Vincent et al.
2009). The development and research of data in sports has taken different perspec-
tives, like data mining (Ofoghi et al. 2013; Li 2014; Leung and Joseph 2014;
Haghighat et al. 2013), information systems (Shao et al. 2014; Qi and Wang 2014;
Xie and Cai 2014; Luo and Deng 2014), event detection in videos (Li and Sezan
2002; Taki et al. 1996; Tong et al. 2005; Taki and Hasegawa 2000), behavioural
models (Menéndez et al. 2013; Cheng et al. 2002; Hernandez Mendo and Anguera
2002), social network analysis (Lusher et al. 2010; Vaz de Melo et al. 2012;
Pardalos and Zamaraev 2014; Passos et al. 2011) and also outcome prediction by
different statistical applications (Baker and Scarf 2006; Groll et al. 2015;
Stekler et al. 2010; Leitner et al. 2010). A characteristic feature of all of these
approaches is the amount of data obtained by any method chosen or perspective
taken. As Wang and Wang (2015) stated, sports data feature strong timeliness,
multiple types, many specifications, large quantities and very complex storage.
Rapid development of information technology allows users access to all types of
information, including high-quality footage of live sport events, and has led to an
expansion of market size for sport, not seen before in the history of the industry
(Westerbeek 2013). Given how big sport business has become, many efforts have
been made to handle as much information as possible, and even more, these advances
in information technologies have allowed researchers and managers to be able to
advance towards sport-specific metrics (Gyarmati and Hefeeda 2016). With these
mechanisms, managers from professional to college or amateur teams are now
capable to develop deeper analysis of a given match in order to take some opportune
corrective measures in their team performance.
In soccer, match analysis is a fully accepted detection vehicle for any serious-­
minded managerial and coaching staff (Carling et al. 2005). These authors define
match analysis as the objective recording and examination of behavioural events
occurring during competition. It could provide objective information about the
underlying causes of a determined problem, e.g. a poor delivery in the penalty area
or difficulties in taking the ball off the goalkeeper’s zone. The final score, in contrast
to detailed information as the unique performance indicator, is insufficient to prop-
erly asses a team. As a consequence, measurement tools play an important role in
this particular field because human observation and memory are not reliable enough
to provide accurate and objective information from athletes in high-performance
competitions (Henriques Abreu et al. 2012). Furthermore, in most team sports, the
16 A Methodology for the Analysis of Soccer Matches Based on PageRank Centrality 259

observer is unable to assimilate the entire action taking place on the field, due to its
attention to the game critical areas; hence, most of the peripheral play action gets
usually lost (Hughes et al. 2001).
Given this context, the aim of this chapter is to provide an observational method-
ology for soccer match analysis, armed with just some basic information in order to
develop a non-expensive tool based on network analysis. This is achieved by using
the PageRank centrality measure created by Google (Brin and Page 1998) and some
insights from graph theory and social networks. Our contribution has to do with a
simple and economic application of the PageRank algorithm to the performance
analysis of soccer players in understanding their performance, as centres in the flow
of the ball, in any given match.
A key claim of our paper has to do with the fact that network analysis, when
applied to soccer, allows for the representation of teamwork, which leads to a better
understanding of the team as a whole, in contrast to the analysis of individuals and
their personal contributions. This possibility of reinterpreting individual statistical
data based on the comprehension of the group dynamic is an example that the con-
ciliation of the common performance indexes with the novel approaches permits to
cover every single level of analysis (Maya Jariego and Bohórquez 2013).
Some of the practical applications that our model features are as follows:
• Identification of the most relevant players
• Team dominance throughout the game
• Direction of the flows
• Relevance of the substitutes (compared to the time played)
• Ball recoveries or interceptions by player
• Identification of tactics (i.e. defensive or offensive game, long passes, area where
the ball most circulated, etc.)
• Analysis of the rival team
The previous features are explained and discussed in this article on the following
sections. After this introduction, the next section outlines the theoretical and con-
ceptual framework that embodies our proposition with some important background
to understand the core of our methodology. Section 16.4 describes the methodology
required for the match analysis and the indicators used. Section 16.5 presents the
results of different matches from the group phase of the past Copa America 2015
used as example. Finally, this chapter ends with a brief conclusions section, present-
ing the guidelines for future researches.

16.2 Conceptual and Theoretical Framework

16.2.1 Observational Methodology

Observational methods have been explained thoroughly by Bakeman and Gottman

(1987). This type of methodology is characterized because the level of participation
is “nonparticipative observation”, given that the observer does not interact with the
260 J. Rojas-Mora et al.

observed players and the degree of perception is complete, direct observation

(Lapresa et al. 2013).
Furthermore, in the field of sports, observation is important from both the proce-
dural and substantive points of view. In terms of procedure, it is the only scientific
approach that is capable of gathering data directly from participants (athletes,
coaches, trainers, etc.) in both the training and competitive contexts without
­eliciting a response from them (Anguera and Hernandez-Mendo 2013). In this last
article, a thorough revision of the existent literature about observational methodol-
ogy is applied, containing examples from basketball, handball, soccer, judo and
swimming, among other several disciplines from sports.
In our specific case, we observe the gameplays from video matches that are
recorded from a specific match, but any video recorded would work, considering
how easy is to replay a whole match in television and on the Internet. The benefits
from using video replays (Carling et al. 2005) are that video:
• Provides a permanent record of performance which can be watched as many
times as desired
• Provides valuable information that may have been missed or forgotten by coaches
or players during the match
• Helps to concentrate in a specific aspect or a specific player’s performance
• Allows for an action sequence to be repeated as often as necessary to ensure that
players have absorbed and understood the required information
• Can be used in real time for immediate analysis and evaluation or post-match for
a deeper insight
• Is a familiar mean of presenting and discussing performance
The video analysis process is described as follows:
1. Match is observed/recorded.
2. Analysis is made under digital video editing and/or data coding.
3. Four core elements are identified: player, action, time and position.
4. A database is generated containing two-dimensional match reconstruction,
edited match video, tables, graphs and/or spatial data.
There are some important aspects that make observation a proper and objec-
tive tool to overcome some of the weaknesses of the common analysis, e.g. the
coach gets a lot of information and is not capable to exploit that data, and the
observer gets baffled by the number of actions taking place simultaneously or in
rapid sequence which cannot be immediately processed given that his or her
attention is directed to the most critical areas of the game. Emotional factors of
the observer play also an important role (Frank and Miller 1991; Carling et al.
2005; Hughes et al. 2001).
Thus, a wide variety of research has been conducted using perspective of obser-
vational studies. These have contributed to facilitate the systematic observation of
sports. Some of them can be found in the following articles: Castellano et al. (2008),
Leitão and Campaniço (2009), Sarmento et al. (2009), Lapresa et al. (2016), Jonsson
et al. (2006) and Santos et al. (2014). To sum up, working under an observational
16 A Methodology for the Analysis of Soccer Matches Based on PageRank Centrality 261

methodology ensures to be under the most suitable methodology used in sport studies
when the objective is to analyse matches in their natural context and dynamics
(Anguera and Hernandez-Mendo 2014).

16.2.2 Graph Theory

It is said that Euler in 1741 founded both topology and graph theory by solving the
Konigsberg bridges problem. It consisted in visiting the four land masses of the
entire city, starting and finishing in the same one, while completely crossing once
over each of the seven bridges. Like this one, many situations can be described by
means of a diagram consisting of a set of points connected with lines. This is the
basic principle of graph theory; points are called “nodes” and the lines that connect
them are named “edges”. A graph G is an ordered triple V(G), E(G), φG consisting
of a non-empty set V(G) of vertices; a set E(G), disjoint from V(G), of edges; and an
incidence function φG that associates with each edge of G an unordered pair of ver-
tices of G. If e is an edge and u and v are vertices such that φG(e) = uv, then e is said
to join u and v; the vertices u and v are called the ends of e (Bondy and Murty 1976).

16.2.3 Network and Centrality Measures

The scientific study of networks, including computer networks, social networks and
biological networks, has received an enormous amount of interest in the last few
years. Much of this interest can be attributed to the appeal of social network analysis
on relationships among social entities and on the patterns and implications of these
relationships. That is, relations defined by linkages among units are a fundamental
component of network theories (Wasserman and Faust 1994). The structure of net-
works has been of interest of many branches of science: methods for analysing
network data, including methods developed in physics, statistics and sociology; the
fundamentals of graph theory, computer algorithms and spectral methods; mathe-
matical models of networks, including random graph models and generative mod-
els; and theories of dynamical processes taking place on networks (Newman 2010).
Centrality has been widely studied in the context of social network analysis
(Clemente et al. 2016; Lusher et al. 2010). Thus, several measures have been devel-
oped, like “betweenness” (Freeman 1979), “eigenvector centrality” (Bonacich
1972) and “closeness” (Freeman 1979), among others. Even though many measures
and different approaches about the concept of centrality in a network exist, Freeman
(1979) offers three intuitive conceptions:
(a) The most intuitive conception is that point centrality is some function of the
degree of a point. The degree of a point pi is the count of number of other points
(nodes) pj(j ≠ i) that are adjacent 1 to it.

1
Two incident vertices with a common edge are adjacent.
262 J. Rojas-Mora et al.

(b) The second view is based upon the frequency with which a point falls between
pairs of other points on the geodesic paths connecting them.
(c) The third conception is based upon the degree to which a point is close to all
other points in the graph.
The main idea of these different approaches of centrality is to define a measure
that determines the relative importance of a node within a graph. The discussion of
these different researches focuses on what the most appropriate measurement should
be. A complete summary and revision of the concept of centrality in networks and
the different existing measures and interpretations of the concepts can be found in
Borgatti (2005). The key claim of that paper is that centrality measures can be
regarded as generating expected values for certain kinds of node outcomes (such as
speed and frequency of reception) given implicit models of how traffic flows.
Regarding network analysis and sports, intra-group relationships are important
for sport teams and include aspects such as cohesiveness and hierarchies among
players (Lusher et al. 2010). Social network analysis (SNA) methods allow for the
exploration of “social2” relations between team members and their individual-level
qualities simultaneously. Its usefulness has to do with addressing the issue of inter-
dependencies in the data inherent in team structures. The most basic concept of
relationship in network analysis is defined by the existence of a link between two
players (e.g. i and j). It is binarily defined, where ei, j = 1 models the existence of a
relation, and ei, j = 0 represents its absence. More complex networks might consider
valued edges, depending on the importance or strength of the bond.

16.2.3.1 PageRank Centrality

PageRank, a registered trademark of Google, is an algorithm introduced by Brin and

Page (1998) and used to determine the relative importance of a node, i.e. its central-
ity in the graph. Its most intuitive definition is that a node has high rank if the sum
of the ranks of its backlinks3 is high.
The definition of a simple ranking is as follows: Let u be a node, Fu be the set of
nodes u points to, Bu be the set of nodes that point to u, Nu = |Fu| be the number of
edges to/from u and c be the factor of normalization; in order to keep constant the
total rank of all the nodes, then the definition of R, a simplified version of PageRank,
would be:

R (v)
R (u ) = c å (16.1)
vÎBu N (v)

Even though the equation is recursive, it may be computed by starting with any set
of ranks and iterating the computation until it converges. To overcome the problem

2
“Social”, in this case, refers to how frequently a player passes the ball to another.
3
For a given node in a graph, its “backlinks” are the nodes linking to it.
16 A Methodology for the Analysis of Soccer Matches Based on PageRank Centrality 263

that, during the iteration, the loop will not distribute rank because two nodes are
linked only between each other and, therefore, there are no out edges, Page et al.
(1999) introduced the following rank source.
Let E(u) be some vector over the nodes that corresponds to a source of rank, then
the PageRank of a set of nodes is an assignment R′, to the nodes which satisfies:

R¢ ( v )
R¢ ( u ) = c å + cE ( u ) (16.2)
vÎBu Nv

Such that cc is maximized and ‖R‖1 = 1, where ‖R‖1 denotes the L1 norm of R′.
As presented from the creators of the method, the PageRank algorithm may be com-
puted as:

R0 ¬ S
loop:
Ri +1 ¬ ARi
d ¬ Ri 1 - Ri +1 1 (16.3)
Ri +1 ¬ Ri +1 + dE
d ¬ Ri +1 - Ri 1
while d > e

The complete axiomatization of this page ranking algorithm can be found in

Altman and Tennenholtz (2005). Some known applications of the algorithm pre-
sented are citation networks (Ding et al. 2009; Ma et al. 2008), and like ours, many
other applications have been made in chemistry, biology, bioinformatics, neurosci-
ence and engineered systems, among others (Gleich 2014). Several approxima-
tions of the PageRank algorithms have been developed, and quantitative analyses
have been provided to illustrate the effectiveness of the PageRank computation
(Chung 2014).
In simple words, our approach generalizes the PageRank algorithm, commonly
used to rank the importance of websites, to value each node given the frequency of
the flows in the network, which in this case refers to the passing of the ball. This, in
order, allows for a complete visualization of the game that shows important infor-
mation. This representation of the match in one map of relationships between the
players as a team allows for the identification of game patterns (Lago and Anguera
2003). Given that goals are infrequent events, it is necessary to have a different
independant variable, i.e. passes, in order to better understand the performance of a
player in a match. Supporting this, Lago and Martín (2007) found that the variance
of the goals scored in soccer matches is not large enough to identify statistically
significant determinants.
264 J. Rojas-Mora et al.

16.3 Methodology

16.3.1 Data Collection

To provide a structure for the application of the methodology, we propose a simple

graph where players are nodes and the weights of the edges represent the passes
between any two players. In order to do this, there are some considerations in the
notation to bear in mind.
Notational analysis is defined as a means of recording events so that there is
an accurate and objective record of what actually took place. It provides a factual
record that does not lie (Carling et al. 2005). The first part of our proposal is
pretty straightforward. It begins with an almost standard notational procedure,
like those presented in Sarmento et al. (2009), Anguera and Hernandez-Mendo
(2013) and Lapresa et al. (2013). It consists on dividing the area of the field into
12 same-sized zones: three rows (left, centre and right) and four columns (e.g. I,
II, III and IV).
With this starting point, the aim is to collect information of each play of the game
considering the following information:
• Time of the play
• Sender of the ball
• Zone of the field that the ball is sent from
• Defending player (if any)
• Receiver of the ball
• Zone of the field where the ball is being received
• Defending player of the receiver player (if any)
It is important to bear in mind that the analysis is going to be made under the
basic assumption that the relative importance of a player is given by the importance
of the flows received, which we model as passes. Therefore, the whole match is
going to be understood as a network where flows are given by the ball passing
through the players. The collection of all of these data can be made with a simple
spreadsheet. With the information of each play obtained in this simple way, it is
possible to model an adjacency matrix of 28 columns and 28 rows (11 starting play-
ers plus the 3 substitute players for each team).
In the second part, the collected data is processed through an R script (a GNU
software4) using the “igraph” package (Csardi and Nepusz 2006), which contributes
with routines for graph and network visualization and analysis. This package
includes an implementation of the PageRank algorithm that we used for our
analysis.

4
For further inquiries about GNU R (R Core Team 2016), check the website of the R Foundation
for Statistical Computing, URL https://www.R-project.org/
16 A Methodology for the Analysis of Soccer Matches Based on PageRank Centrality 265

16.4 Results

In order to visualize the results of our work, the methodology was applied to three
soccer matches from the group phase of the Copa America 2015. In specific, those
matches analysed were Chile-Ecuador, Chile-Mexico and Chile-Bolivia. They were
broadcasted by a public television channel, so the recording was made with digital
video tools.5
The nodes represent the players, the notation “C” is for the Chilean players,
while “E” is used for the Ecuadorian players, and the number next to each letter
corresponds to the shirt number worn by any given player in that match. The size of
the node is proportional to its PageRank centrality measure, and the boldness of the
edges is proportional to the number of passes from one player to another. Nodes that
seem to be outliers on the bottom of the figures are the substitute players. The same
goes for the next figures, where “M” denotes Mexican players and “B” denotes
Bolivian players. We present a brief discussion of the graphs showing some basic
examples of the types of analysis that can be made with this tool.
Figure 16.1 shows the Chilean dominance on the match against the Ecuadorians
with a higher volume of passes and variety of options to deliver; it meant a higher
possession of the ball through the entire match. Boldness of the Chilean arrows
indicates an offensive tendency towards the middle field to get to the centre forward
C7, Alexis Sánchez. Another conclusion has to do with the principal nodes of each
team. For the Chilean case, the bigger node corresponds to C8, Arturo Vidal, the
most relevant player given the flows that passed through him and the importance of
the players who passed the ball to him, i.e. his PageRank value. Therefore, Vidal
became the most important player for the Chilean team which won for 2 goals
against 0, with the first one scored by Vidal.
For sport managers, it is useful to analyse other aspects too, like the flows
between two nodes of different teams. For the case of E13, Enner Valencia, Fig. 16.1
shows that from the several flows that are directed to him, many correspond to
Chilean nodes. This means that he interfered with many passes intended for Chilean
players and, accordingly, recovered the ball many times, playing a defensive role
even though his position was that of a winger.
In the Chile-Mexico match, Fig. 16.2 shows that the ball circulated significantly
through every player, with a higher density on the Chilean side of the field. In time,
Chilean nodes are bigger, have a higher centrality or are more important, with the
most relevant players being C8, Arturo Vidal, and C10, Jorge Valdivia, with Vidal
scoring 2 goals throughout the match.
For the Mexican team, the most relevant players are their three forwarders, show-
ing a considerable amount of passes received from the central and lateral defenders,
which means that the Mexican game was characterized by long passes (e.g. M3 to
M7 edge). The most relevant players for Mexico were M9, Raul Jimenez, and M19,
Matías Vuoso, both managed to score goals for their team.

5
The dataset can be requested to the corresponding author.
266 J. Rojas-Mora et al.

Fig. 16.1 Network of the Chile-Ecuador match in the group phase of the Copa America 2015

Finally, the Chile-Bolivia game ended with a lopsided score of 5-0. In this match,
Chile played with two more “open” forwarders than the previous games, which
allowed them to dominate more positions of the field with a considerable flow of the
ball going to C11, Eduardo Vargas. There is clear relation between the volume of the
flows for the Chilean team against the Bolivian team and the final score of the
match, i.e. Chile owned the ball and, thus, the opportunities to score. In this aspect
it is important to consider: in the discipline, one of the most important findings is the
correlation between the ability to retain possession of the ball for prolonged periods
of time and success (Bate 1988; Gómez and Álvaro 2002; James et al. 2004; In:
Lago and Martín 2007).
Another analysis to be made is that in this match, the substitute players had a big-
ger importance in the development of the plays for both of the teams, which could not
be seen, e.g. in the Ecuadorian team on Fig. 16.1. This adds another type of analysis
16 A Methodology for the Analysis of Soccer Matches Based on PageRank Centrality 267

Fig. 16.2 Network of the Chile-Mexico match in the group phase of the Copa America 2015

to our approach: relevance of the substitute players during the time they played. In
Fig. 16.3, the difference of the relevance of the substitute nodes for both teams is
remarkable. Some are even similar to the nodes of players that started the game from
the minute one. Hence, sport managers can take decisions and evaluate performance,
given the relevance that substitutes manage to acquire during a shorter period than the
regular first team players, and maybe evaluate if they are making the most of their time
on field. For example, a coach might decide to turn a substitute into a regular starter
because of the relevance of his node and flows directed from and to him.
A last thing to keep in mind is that the analysis can also have another focus. Sport
managers can use this methodology not only to analyse their team’s weaknesses and
strengths but also to analyse the adversaries through the data collection of other
previous matches in order to set proper tactics and, in this way, develop that
­competitive advantage that is so much important in such a competitive environment
(i.e. see which are the opponents that do not have so much relevance and then focus
the direction of the ball through the less explored areas of those nodes).
268 J. Rojas-Mora et al.

Fig. 16.3 Network of the Chile-Bolivia match in the group phase of the Copa America 2015

16.5 Conclusion

Graph-based algorithms have been proved to be relevant to a wide variety of appli-

cations. Even though, there is no such thing as a “perfect” algorithm to rank in
sports, there is strong evidence to believe that the Google PageRank algorithm pro-
vides reliable insights (London et al. 2014; Govan et al. 2008). In our case, it aids to
provide answers to the “who, when and where” of the plays of the match, gathering
the data from point to point and transforming it into a single graph that becomes a
powerful visualization, in order to aid coaches or managers to make deeper analysis.
This information gets often lost at the time being played, given the huge amount of
interactions that occur between the players, which disables the observers to take in-­
game decisions based on this data.
Our approach is within context of social networks and graph theory, by using the
PageRank centrality measure as a key element to rank the nodes. That algorithm has
demonstrated a wide variety of applications and a resourceful way to measure the
relevance of the vertices of a given graph in many different situations. It is also
16 A Methodology for the Analysis of Soccer Matches Based on PageRank Centrality 269

embodied within the observational methodologies and notational analysis, which

ensures to be under the most suitable design methodology for sport analysis as
stated before, given the importance of studying sports with its natural dynamics and
that contributes to keep an objective record. Other benefits of this type of method-
ologies are related to overcoming the lack of attention that is paid during a match
due to the amount of information for the observers and the emotional or personal
factors of the observer that could potentially affect the analysis carried out.
We applied the methodology proposed to three real matches for the Copa America
2015 (Chile-Mexico, Chile-Bolivia and Chile-Ecuador), which illustrated some of
the ideas that can be concluded with this observational methodology. Some of the
insights that this method could provide for a match analysis are most relevant play-
ers, direction of the ball through the entire game, ball recoveries, effective passes,
zone of the field that was less covered by seeing the participation of the node in its
assigned section of the field, relevance of the substitutes and analysis of the rival
team, among others.
The application of social network analysis (SNA) methodologies to sports has
been of wide interest (Lusher et al. 2010). Its application to soccer allows model-
ling a team as a micro-system, whose components are linked by stable and ordered
interactions that represent the collective work (Lago and Anguera 2003). At the
same time, it allows to analyse the role that different nodes in the graph play, hav-
ing a wider perspective on how this so-called system works (Maya Jariego and
Bohórquez 2013).
Implications for sport managers are significant. Considering the low cost of this
methodology, it could generate many benefits for the right performance analysis and
decision-making process through a simple but powerful visualization. Although
many big teams are using different methods specially developed for sport analysis
(e.g. video detection analysis, specialized software or appropriate databases for
large amounts of data storage) which have a considerable cost, it is important to bear
in mind the budget that smaller teams might have. Our proposal contributes by add-
ing a simple, but scientific, methodology to analyse the data, with a widely used
algorithm as PageRank, which will contribute with more angles than just the final
score as a performance index. The only costly side of our proposal has to do with
the time needed to write down all the plays of the match. However, with just a basic
spreadsheet and some simple automation, this burden can be somewhat eased.
For future work, we intend to study the indirect flow of passes given our initial
adjacency matrix, to see which player of a team can indirectly allow for the flow of
the ball between two almost unlinked players. We would also like to analyse the
network flows from a spatiotemporal perspective, to obtain information like the
position of the field which is more exploited, or the dynamics in the flow of passes
of a given game. This would lead to the need of adding more detailed information at
the data collection stage, and therefore, a larger database should be required.
However, it would not change the core of the proposed methodology. The context
for this future research has to do with the main idea of being able to find that “little
bit extra” which could potentially make the difference between success and failure
in sport team management (Carling et al. 2005).
270 J. Rojas-Mora et al.

References

Altman, A., & Tennenholtz, M. (2005). Ranking systems: The PageRank Axioms. In: EC ‘05
Proceedings of the 6th ACM conference on electronic commerce, ACM, 1–8.
Anguera, M. T., & Hernandez-Mendo, A. (2013). Observational methodology in sport sciences.
Revista de Ciencias del Deporte, 9(3), 135–160.
Anguera, M. T., & Hernandez-Mendo, A. (2014). Técnicas de análisis en estudios observacionales
en ciencias del deporte. Cuadernos de Psicología del Deporte, 15(1), 13–30.
Bakeman, R., & Gottman, J. M. (1987). Applying observational methods: A systematic view. In
J. D. Osofsky (Ed.), Handbook of infant development (pp. 818–853). New York: Wiley.
Baker, R., & Scarf, P. (2006). Predicting the outcomes of annual sporting contests. Journal of the
Royal Statistical Society Series C (Applied Statistics), 55(2), 225–239.
Bate, R. (1988). Football chance: Tactics and strategy. In T. Reilly, A. Lees, K. Davids, &
W. Murphy (Eds.), Science and football (pp. 293–301). London: E & FN Spon.
Bonacich, P. (1972). Factoring and weighting approaches to status scores and clique identification.
The Journal of Mathematical Sociology, 2(1), 113–120.
Bondy, J. A., & Murty, U. S. R. (1976). Graph theory with applications. Great Britain: The
Macrmillan Press Ltd.
Borgatti, S. P. (2005). Centrality and network flow. Social Networks, 27, 55–71.
Brin, S., & Page, L. (1998). The anatomy of a large-scale hypertextual web search engine.
Computer networks and ISDN systems, 30(1), 107–117.
Carling, C., Williams, A. M., & Reilly, T. (2005). Handbook of soccer match analysis. New York:
Taylor & Francis Group.
Castellano, J., Perea, A., Alday, L., & Hernandez-Mendo, A. (2008). The measuring and observa-
tion tool in sports. Behavior Research Methods, 40(3), 898–905.
Cheng, F., Christmas, W. J., & Kittler, J. (2002). Recognising human running behaviour in sports
video sequences. Proceedings - International Conference on Pattern Recognition, 16(2),
1017–1020.
Chung, F. (2014). A brief survey of PageRank algorithms. IEEE Transactions on Network Science
and Engineering, 1, 38–42.
Clemente, F. M., Lourenço, F. M., & Sousa, R. (2016). Social network analysis applied to team
sports analysis. Cham: Springer International Publishing.
Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research.
Interjournal, Complex Systems, 1695(5), 1–9.
Ding, Y., Yan, E., Frazho, A., & Caverlee, J. (2009). PageRank for ranking authors in co-citation
networks. Journal of the American Society for Information Science and Technology, 60(11),
2229–2243.
Frank, I. M., & Miller, G. (1991). Training coaches to observe and remember. Journal of Sports
Sciences, 9, 285–297.
Freeman, L. C. (1979). Centrality in social networks conceptual clarification. Social Networks, 1,
215–239.
Gleich, D. F. (2014). PageRank beyond the web. SIAM Review, 57(3), 321–363.
Gómez, M., & Álvaro, J. (2002). El tiempo de posesión como variable no determinante del resul-
tado en los partidos de fútbol. El Entrenador Español, 97, 39–47.
Govan, A. Y., Meyer, C. D., & Albright, R. (2008). Generalizing Google’s PageRank to National
Football League Teams. In: Proceedings of the SAS Global Forum, San Antonio, paper 151.
Groll, A., Schauberger, G., & Tutz, G. (2015). Prediction of major international soccer tourna-
ments based on team-specific regularized Poisson regression: An application to the FIFA world
cup 2014. Journal of Quantitative Analysis in Sports, 11(2), 97–115.
Gyarmati, L., & Hefeeda, M. (2016). Analyzing in-game movements of soccer players at scale.
MIT Sloan Sports Analytics Conference 2016.
Haghighat, M., Rastegari, H., & Nourafza, N. (2013). A review of data mining techniques for result
prediction in sports. Advances in Computer Science: an International Journal, 2(5), 7–12.
16 A Methodology for the Analysis of Soccer Matches Based on PageRank Centrality 271

Henriques Abreu, P., Moura, J., Castro-Silva, D., Reis, L. P., & Garganta, J. (2012). Performance
analysis in soccer: A Cartesian coordinates based approach using RoboCup data. Soft
Computing, 16(1), 47–61.
Hernandez Mendo, A., & Anguera, M. T. (2002). Behavioral structure in sociomotor sports:
Roller-hockey. Quality and Quantity, 36(4), 347–378.
Hughes, M., Evans, S., & Wells, J. (2001). Establishing normative profiles in performance analy-
sis. Journal of Performance Analysis in Sports, 1, 4–27.
James, N., Jones, P. D., & Mellalieu, S. D. (2004). Possession as a performance indicator in soccer.
International Journal of Performance Analysis in Sport, 4, 98–102.
Jonsson, G. K., Anguera, M. T., Blanco-Villaseñor, A., Losada, J. L., Hernández-Mendo, A., Ardá,
T., Camerino, O., & Castellano, J. (2006). Hidden patterns of play interaction in soccer using
SOF-CODER. Behavior Research Methods, 38(3), 372–381.
Lago, C., & Anguera, M. T. (2003). Utilización del análisis secuencial en el estudio de las interac-
ciones entre jugadores en el fútbol de rendimiento. Revista de Psicología del Deporte, 12(1),
27–37.
Lago, C., & Martín, R. (2007). Determinants of possession of the ball in soccer. Journal of Sport
Sciences, 25(9), 969–974.
Lapresa, D., Álvarez, L., Arana, J., Garzón, B., & Caballero, V. (2013). Observational analysis
of the offensive sequences that ended in a shot by the winning team of the 2010 UEFA futsal
championship. Journal of Sports Sciences, 31(15), 1731–1739.
Lapresa, D., Arana, J., Anguera, M. T., Pérez-Castellano, J. I., & Amatria, M. (2016). Application
of logistic regression models in observational methodology: Game formats in grassroots foot-
ball in initiation into football. Anales de Psicología, 32(1), 288–294.
Leitão, J. C., & Campaniço, J. (2009). Research methods support in observation sports laboratory.
Motricidade, 5(3), 84.
Leitner, C., Zeileis, A., & Hornik, K. (2010). Forecasting sports tournaments by ratings of (prob)
abilities: A comparison for the EURO 2008. International Journal of Forecasting, 26(3),
471–481.
Leung, C., & Joseph, K. (2014). Sports data mining: Predicting results for the college football
games. Procedia Computer Science, 35, 710–719.
Li, B., & Sezan, M. (2002). Event detection and summarization in American football broadcast
video. Proceedings of SPIE, 4676.
Li, H. (2014). Strategy and analysis of sport events based on data mining technology. Applied
Mechanics and Materials, 687-691, 1137–1140.
London, A., Németh, J., & Németh, T. (2014). Time-dependent network algorithm for ranking in
sports. Acta Cybernet., 21(3), 495–506.
Luo, S., & Deng, G. (2014). Research on the development of sports information system based on the
sports information management theory. Advanced Materials Research, 998-999, 1327–1330.
Lusher, D., Robins, G., & Kremer, P. (2010). The application of social network analysis to team
sports. Measurement in Physical Education and Exercise Science, 14(4), 211–224.
Ma, N., Guan, J., & Zhao, Y. (2008). Bringing PageRank to the citation analysis. Information
Processing & Management, 44(2), 800–810.
Maya Jariego, I., & Bohórquez, M. R. (2013). Análisis de las redes de distribución de balón en fút-
bol: pases de juego y pases de adaptación. Revista Hispana para el Análisis de Redes Sociales,
24(2), 135–155.
Menéndez, H., Bello-Orgaz, G., & Camacho, D. (2013). Extracting behavioral models from 2010
FIFA world cup. Journal of Systems Science and Complexity, 26(1), 43–61.
Newman, M. (2010). Networks. Oxford: Oxford University Press.
Ofoghi, B., Zeleznikow, J., MacMahon, C., & Raab, M. (2013). Data Mining in Elite Sports: A
review and a framework. Measurement in Physical Education and Exercise Science, 17(3),
171–186.
Page, L., Brin, S., Motwani, R., & Winograd, T. (1999). The pagerank citation ranking: Bringing
order to the web. Technical Report, Stanford University, Stanford.
 272 J. Rojas-Mora et al.

Pardalos, P., & Zamaraev, V. (2014). The impact of social networks on sports. In P. Pardalos &
V. Zamaraev (Eds.), Social networks and the economics of sports (pp. 1–8). Cham: Springer.
Passos, P., Davids, K., Araújo, D., Paz, N., Minguéns, J., & Mendes, J. (2011). Networks as a novel
tool for studying team ball sports as complex social systems. Journal of Science and Medicine
in Sport, 14(2), 170–176.
Qi, N., & Wang, L. (2014). Constructing of sports information management theories and devel-
oping research of application system. In: Proceedings of the IEEE Workshop on Advanced
Research and Technology in Industry Applications (WARTIA), 2014, Ottawa.
R Core Team. (2016). R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL https://www.R-project.org
Santos, S., Sarmento, H., Alves, J., & Campaniço, J. (2014). Construcción de un instrumento para
la observación y el análisis de las interacciones en el waterpolo. Revista de Psicología del
Deporte, 23(1), 191–200.
Sarmento, H., Leitao, J., Anguera, T., & Campaniço, J. (2009). Observational methodology in
football: Development of an instrument to study the offensive game in football. Motricidade,
5(3), 19–24.
Shao, L., Luo, L., & Qi, H. (2014). Theoretical construction of sports information management
and research on application system development. Advanced Materials Research, 926-930,
2759–2762.
Stekler, H., Sendor, D., & Verlander, R. (2010). Issues in sports forecasting. International Journal
of Forecasting, 26(3), 606–621.
Taki, T., Hasegawa, J., & Fukumura, T. (1996). Development of motion analysis system for quan-
titative evaluation of teamwork in soccer games. In: Proceedings of 3rd IEEE International
Conference on Image Processing, 1996, Laussane.
Taki, T., & Hasegawa, J. (2000). Quantitative measurement of teamwork in ball games using domi-
nant region. International Archives of Photogrammetry and Remote Sensing, XXXIII(Suplement
B5), 125–131.
Tong, X., Duan, L., Lu, H., Xu, C., Tian, Q., & Jin, J. (2005). A mid-level visual concept genera-
tion framework for sports analysis. In: Proceedings of the IEEE International Conference on
Multimedia and Expo, 2005, Amsterdam.
Vaz de Melo, P., Almeida, V., Loureiro, A., & Faloutsos, C. (2012). Forecasting in the NBA and
other team sports. ACM Transactions on Knowledge Discovery from Data, 6(3), 1–27.
Vincent, J., Stergiou, P., & Katz, L. (2009). The role of databases in sport science: Current practice
and future potential. International Journal of Computer Science in Sport, 8(2), 50–66.
Wang, Q., & Wang, Y. (2015). Standarized storage of sports data based on XML. The Open
Cybernetics & Systemics Journal, 9, 2312–2316.
Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications.
Cambridge, UK/New York: Cambridge University Press.
Westerbeek, H. (2013). Global sport business: Community impacts of commercial sport. London:
Routledge.
Xie, P., & Cai, X. (2014). Research on sports information integrated management application
development. Advanced Materials Research, 926-930, 4182–4185.

View publication stats