Content uploaded by Rodrigo Aquino
Author content
All content in this area was uploaded by Rodrigo Aquino on Nov 20, 2019
Content may be subject to copyright.
Content uploaded by Rodrigo Aquino
Author content
All content in this area was uploaded by Rodrigo Aquino on Nov 12, 2019
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rpan20
International Journal of Performance Analysis in Sport
ISSN: 2474-8668 (Print) 1474-8185 (Online) Journal homepage: https://www.tandfonline.com/loi/rpan20
Comparisons of ball possession, match running
performance, player prominence and team
network properties according to match outcome
and playing formation during the 2018 FIFA World
Cup
Rodrigo Aquino, João Cláudio Machado, Filipe Manuel Clemente, Gibson
Moreira Praça, Luiz Guilherme C. Gonçalves, Bruno Melli-Neto, João Victor S.
Ferrari, Luiz H. Palucci Vieira, Enrico F. Puggina & Christopher Carling
To cite this article: Rodrigo Aquino, João Cláudio Machado, Filipe Manuel Clemente, Gibson
Moreira Praça, Luiz Guilherme C. Gonçalves, Bruno Melli-Neto, João Victor S. Ferrari, Luiz H.
Palucci Vieira, Enrico F. Puggina & Christopher Carling (2019) Comparisons of ball possession,
match running performance, player prominence and team network properties according to
match outcome and playing formation during the 2018 FIFA World Cup, International Journal of
Performance Analysis in Sport, 19:6, 1026-1037, DOI: 10.1080/24748668.2019.1689753
To link to this article: https://doi.org/10.1080/24748668.2019.1689753
Published online: 10 Nov 2019. Submit your article to this journal
Article views: 26 View related articles
View Crossmark data
Comparisons of ball possession, match running performance,
player prominence and team network properties according to
match outcome and playing formation during the 2018 FIFA
World Cup
Rodrigo Aquino
a
, João Cláudio Machado
b
, Filipe Manuel Clemente
c,d
,
Gibson Moreira Praça
e
, Luiz Guilherme C. Gonçalves
f
, Bruno Melli-Neto
g
,
João Victor S. Ferrari
g
, Luiz H. Palucci Vieira
h
, Enrico F. Puggina
g
and Christopher Carling
i
a
Department of Sports, Center of Physical Education and Sports (CEFD), Federal University of Espírito Santo,
Vitória, Espírito Santo, Brazil;
b
Human Performance Laboratory, Faculty of Physical Education and
Physiotherapy, Federal University of Amazonas, Manaus, Brazil;
c
Polytechnic Institute of Viana do Castelo,
School of Sport and Leisure, Melgaço, Portugal;
d
Instituto de Telecomunicações, Delegação da Covilhã,
Covilhã, Portugal;
e
Centro de Estudos em Cognição e Ação/UFMG Soccer Science Center, Departamento de
Esportes, Escola de Educação Física, Fisioterapia e Terapia Ocupacional, Universidade Federal de Minas
Gerais, Belo Horizonte, Brazil;
f
Department of Performance Analysis, Botafogo Football Club S/A, Ribeirão
Preto, Brazil;
g
School of Physical Education and Sport of Ribeirão Preto, University of São Paulo, Brazil;
h
Human Movement Research Laboratory, Post-graduate Program in Movement Sciences, São Paulo State
University, Bauru, Brazil;
i
Institute of Coaching and Performance, University of Central Lancashire,
Preston, UK
ABSTRACT
This study on the 2018 Russia FIFA World Cup examined: i) the
potential variations of ball possession, match running performance,
player prominence, team network properties according to match
outcome and playing formation; and ii) the relationships between
player prominence and total distance covered according to team
ball possession. Sixty-one matches were analysed (n = 988 player
observations). Running performance was examined using total dis-
tance covered in (TDIP) and out of possession, and that travelled in
different speed-range categories. Player prominence (micro) and
team network properties (macro) were obtained using social net-
work analysis where completed passes between teammates were
counted (n = 28,019 passes). Main findings were: i) with the excep-
tion of clustering coefficients which indicate the level of intercon-
nectivity between close teammates (win = draw > loss), match
outcome was unaffected by ball possession, running and network
measures; iii) teams employing a1‒4‒2‒3‒1 formation reported
greater values for ball possession, TDIP, and micro/macro network
measures compared to those playing 1‒4‒4‒2 and 1‒4‒3‒3 forma-
tions; iv) TDIP tended to be related to most player prominence
variables, even though the magnitude of coefficients varied con-
siderably according to network measures and playing positions.
This study has provided additional insights into elite soccer
match-play performance.
ARTICLE HISTORY
Received 30 July 2019
Accepted 9 September 2019
KEYWORDS
Association football; social
network analysis; match
analysis; contextual factors;
performance indicators;
sports sciences
CONTACT Rodrigo Aquino rodrigo.aquino@usp.br Center of Physical Education and Sports, Federal University of
Espírito Santo, Av. Fernando Ferrari, 514 - Goiabeiras, Vitória, ES 29075-910, Brazil
INTERNATIONAL JOURNAL OF PERFORMANCE ANALYSIS IN SPORT
2019, VOL. 19, NO. 6, 1026–1037
https://doi.org/10.1080/24748668.2019.1689753
© 2019 CardiffMetropolitan University
1. Introduction
Match analysis plays a key role in understanding performance in team sports (e.g.,
soccer) by evaluating factors that are associated with success (Carling, Reilly, &
Williams, 2008). Traditionally notational analysis techniques are used to collect informa-
tion regarding match performance, notably technical aspects of play. While such analyses
can provide practitioners with meaningful and actionable information, they are typically
restricted to recording counts of isolated actions on technical performance (e.g., number
and success rates in goal attempts, passes, tackles etc.). Information on the complex
cooperative interactions emerging between physical, tactical and technical elements of
play and on the effects of key contextual issues is lacking (Ribeiro, Silva, Duarte, Davids,
& Garganta, 2017).
As such, more recent research has highlighted the value of novel investigative
methods for team sports performance analysis that are related to complexity sciences
and dynamic systems (e.g., Couceiro, Dias, Araújo, & Davids, 2016;Glazier,2010;
Ribeiro et al., 2019;Vilar,Araújo,Davids,&Button,2012). Studies that re-
conceptualise soccer teams as complex social networks for example seek to uncover
patterns of behavioural interactions using social network analysis or SNA (Freeman,
2004). SNA investigates macro (e.g., team network properties) and micro (e.g., player
prominence) interaction patterns in different ways, for example, through the distri-
bution of passing between players across the same team (Praça, Lima, Bredt,
Clemente, & Andrade, 2019). Research has shown that successful teams displayed
high levels of network properties (e.g., density) during the 2014 Brazil FIFA World
Cup (Clemente, Martins, Kalamaras, Wong, & Mendes, 2015). In contrast, more
recent work has revealed that teams did not change macrostructure patterns accord-
ing to match status (i.e., winning vs. drawing vs. losing) and match outcome (i.e., win
vs. draw vs. loss) during the 2018 Russia FIFA World Cup (Clemente, 2018;Praça
et al., 2019). These discrepancies across findings reinforce the need to continue the
exploration of the potential associations of social network measures with performance
outcomes.
Over the last four decades, time‒motion analysis research has greatly improved our
knowledge of the physical demands in competition settings (Carling, 2013). However,
research has tended to quantify physical demands in isolation without match-related
context (Bradley & Noakes, 2013). Indeed, match running performance is highly depen-
dent upon a myriad of contextual factors such as match status, outcome and playing
formation (Paul, Bradley, & Nassis, 2015). For example, previous studies have shown that
professional soccer players perform less high-speed running activities when winning
compared to losing or when the scoreline is level (Bloomfield, Polman, & O’Donoghue,
2005; Castellano, Blanco-Villaseñor, & Alvarez, 2011; Lago, Casais, Dominguez, &
Sampaio, 2010). Similarly, research investigating the influence of playing formation on
running outputs during professional match-play showed that high-speed activity per-
formed in a traditionally attacking formation when the players’team was in possession
was ~30% to ~40% greater compared to more defensive formations (i.e., 1‒4‒3‒3 and 1‒
4‒4‒2vs. 1‒4‒5‒1) (Bradley et al., 2011). Arguably, a multidimensional approach, com-
bining notational analysis (e.g., ball possession strategies in different field zones), SNA
and match running performance accounting for contextual factors such as match
INTERNATIONAL JOURNAL OF PERFORMANCE ANALYSIS IN SPORT 1027
outcome (result) and playing formation would provide a more holistic understanding of
match demands at elite standards.
The aim of this study was to verify the possible variations of ball possession, match
running performance, player prominence, team network properties according to match
outcome and playing formation. Furthermore, this study investigated the relationships
between player prominence and match running performance.
2. Methods
2.1. Observational design and match sample
The observational design of this study was nomothetic (i.e., data analysed from multiple
players and teams); intra- and inter-sessional (i.e., players and teams were observed during
group and knockout phases of the 2018 Russia FIFA World Cup tournament) (Argilaga,
Villaseñor, Mendo, & López, 2011). Multidimensional analyses were performed: ball posses-
sion, running demands, player prominence and team network properties. A total of sixty-one
matches were analysed (n = 988 player observations). Three matches (i.e., Colombia vs. Japan;
Egypt vs. Uruguay; Peru vs. Denmark) were excluded from the analysis owing to FIFA reports
not providing one of the dependent variables on this study (e.g., running performance; passing
distribution). Only performance in players completing the match in its entirety were included.
Six playing positions were considered: central defenders (CD; n = 257 observations), external
defenders (ED; n = 173 observations), central midfielders (CM; n = 227 observations), external
midfielders (EM, n = 110 observations), forwards (F, n = 99 observations), and goalkeepers
(GK, n = 122 observations). Data were retrieved from the official website of 2018 Russia FIFA
World Cup (FIFA, 2018). This study was conducted in compliance with the Declaration of
Helsinki and approved by the local university research committee.
2.2. Dependent variables
2.2.1. Ball possession variables
Percentage ball possession was defined as the proportion of the playing time each team
held the ball (Collet, 2013; da Mota, Thiengo, Gimenes, & Bradley, 2016). The percentage
of total ball possession and in relation to three zones: defence, midfield, attack –using
ball possession heat map reported by FIFA was recorded. Two researchers randomly
reanalysed six matches (~10% of the total) to analyse intra- and inter-rater reliability of
the data and notably reported values for ICC = 0.94 and ICC = 0.95, respectively.
2.2.2. Match running performance
The data provider company used a real-time optical tracking system (STATS SportVU).
This system operated at 25 frames per second and each frame provided the player
coordinates in the field (xand y). A recent study reported acceptable positional accuracy
of the present tracking system (56 ± 16 cm) compared to a goal-standard system (VICON
motion capture system) (Linke, Link, & Lames, 2018). Running performance variables
measured in metres (m) included: total distance covered (TD); total distance covered in
possession –when the team was in possession of the ball (TDIP); total distance covered
when out of possession –when the opposition team had possession of the ball (TDOP);
1028 R. AQUINO ET AL.
total distance covered walking (Z1; 0–7 km.h
−1
); total distance covered in moderate-
speed running (Z2; 7.1–15 km.h
−1
); total distance covered in high-speed running (Z3;
15.1–20 km.h
−1
); total distance covered in very high-speed running (Z4; 20.1–25 km.h
−1
);
total distance covered sprinting (Z5; > 25.1 km.h
−1
). Additionally, recorded were the
frequency of sprints (i.e., number of actions > 25.1 km.h
−1
) and maximum speeds
achieved during play (km.h
−1
).
2.2.3. Player prominence and team network properties
The analysis of passing distribution was conducted to demonstrate the interpersonal
interactions between teammates through completed passes (i.e., transferring the ball
using permitted body parts from one player to another of the same team, and the player’s
team continued in ball possession; Aquino, Puggina, Alves, & Garganta, 2017).
A previous study has shown that completed passes between teammates can be considered
the most consequential form of interaction in soccer matches (Grund, 2012). Following
each match, two weighted adjacency matrixes (one for each team) were provided by the
FIFA database (n = 122 weighted adjacency matrixes). These matrixes were used to build
afinite nxnnetwork, where entries coded by number “1”, for instance, represents ways
that players interact (Ribeiro et al., 2017). As weighted matrices, passes performed more
than once were coded with the respective volume. The matrix was also bidirectional
(interaction from player A to B was different than B to A), thus allowing to determine the
digraphs within the team (Clemente, Martins, Wong, Kalamaras, & Mendes, 2015). The
players were coded according to playing position.
Two of the present researchers randomly re-analysed six matches (12 adjacency
matrixes; ~10% of the total) via the official broadcasting signal (publicly available) to
verify the results provided by the FIFA website. Intraclass correlation coefficients (ICC)
were calculated to compare the FIFA database disponible and both intra- and inter-rater
reliability (ICC = 0.90; ICC = 0.98; respectively). Player prominence (micro level) and
team properties (macro level) were calculated using the social network analysis. Micro
level analyses were obtained through centrality measures. The players’(nodes) promi-
nence in a team (i.e., graph) was calculated through four measures (Ribeiro et al., 2017): i)
in-degree and out-degree (i.e., the number of completed passes that the players received
and performed, respectively); ii) closeness centrality of a player is defined as the sum of
geodesic distances from all other teammates presented in a team (i.e., which represents
how close the player is to other teammates); iii) betweenness centrality is defined as the
number of times that a player connects two other players through their shortest paths
(i.e., the number of networks that a player “controls”, for instance, player responsible for
connecting the defensive sector within midfield area) and; iv) eigenvector indicates the
influence of a player in a team (i.e., identifies key players who play a crucial role in the
offensive phases). The general network properties of social network analysis (macro level)
includes two metrics (Clemente, Martins, & Mendes, 2016; Praça et al., 2019): i) density is
defined as the ration between the total observed vs. maximum number of possible
interactions between two players (values range from 0 –lack of cooperation –to 1 –
maximal cooperation); ii) clustering coefficients indicates the level of interconnectivity
between close teammates (values range from 0 –lack of cooperation –to 1 –maximal
cooperation). All analysis was performed using the software Gephi 0.9.2.
INTERNATIONAL JOURNAL OF PERFORMANCE ANALYSIS IN SPORT 1029
2.3. Independent variables
Two independent variables were considered: i) match outcome, and ii) playing forma-
tion. Match outcome was the final result of each match, including matches resulting in
a loss (n = 420 player observations), draw (n = 133 player observations) or win (n = 424
player observations). Playing formation was determined by two researchers and qualified
coaches by the Brazilian Soccer Confederation according to player distribution over the
entire 90-minutes of match-play (i.e., tactical line-up and actual formation reported by
FIFA). Inter-rating agreement of playing formation was evaluated by Cohens’kappa
coefficients (k= 0.93). A total of six playing formations were analysed: i) 1‒4‒2‒3‒1
(n = 458 players observations); 1‒4‒4‒2 (n = 137 players observations); 1‒4‒3‒3 (n = 104
players observations); 1‒3‒3‒2‒2 (n = 87 players observations); 1‒3‒4‒3 (n = 84 players
observations); 1‒4‒3‒2‒1 (n = 79 players observations). Two playing formations were
excluded owing to insufficient sample sizes (i.e., 1‒5‒3‒2, n = 7 players observations; 1‒5‒
2‒2‒1, n = 32 players observations).
2.4. Statistical analysis
Data normality and homogeneity of variance were checked using Kolmogorov-Smirnov
and Levene tests, respectively. Comparisons between ball possession variables, match
running performance and metrics from social network analysis according to match out-
come (loss vs. draw vs. win) and playing formation (e.g. 1‒4‒2‒3‒1vs.1‒4‒4‒2vs.1‒4‒3‒3)
were performed using general linear model (i.e., multivariate analysis of variance). Pearson
correlation was used to assess the relationships between running performance and social
network variables. Magnitudes of correlation coefficients (90% CI) were considered trivial
(r ≤0.1), small (0.1 < r ≤0.3), moderate (0.3 < r ≤0.5), large (0.5 < r ≤0.7), very large
(0.7 < r ≤0.9) and nearly perfect (0.9 < r ≤1.0) according to Hopkins (2000). When
necessary, nonparametric counterpart tests and Tukey post hoc test were used. Effect sizes
(ES) of the differences between the match outcome and playing formation were calculated
using Cohen’sd(Cohen, 1988). The dvalues were considered as follows: trivial (d≤0.1),
small (0.1 < d≤0.20), moderate (0.20 < d≤0.50), large (0.50 < d≤0.80) and very large
(d > 0.80). Statistical analysis was performed using the software IBM SPSS Statistics for
Windows, version 22.0 (IBM Corporation, Armonk, NY, USA).
3. Results
3.1. Match outcome
With the exception of clustering coefficients (p = 0.01), ball possession variables, running
outputs and SNA did not affect match outcome (p = 0.06 to 0.89) (Table 1). The
clustering coefficients presented greater values for win vs. loss matches (p = 0.02,
ES = 0.2 [moderate]) and draw vs. loss matches (p = 0.04, ES = 0.18 [small]).
3.2. Playing formation
Comparisons of ball possession variables, running performance and social network
analysis according to playing position are reported in Table 2.Sixplaying
1030 R. AQUINO ET AL.
formations were mainly used during the 2018 Russia FIFA World Cup. A 1‒4‒2‒3‒
1 formation was most frequently used by the teams (46.4%; 458 players observa-
tions), followed by a 1‒4‒4‒2 (13.9%; 137 players observations) and 1‒4‒3‒3
(10.5%; 104 players observations).
Ball possession percentages were higher for teams playing a 1‒4‒2‒3‒1com-
pared to a 1‒4‒4‒2formation(p<0.001,ES=0.50[large]).Ballpossessionin
defensive zones presented greater values in a 1‒4‒3‒3 versus a 1‒4‒2‒3‒1forma-
tion (p < 0.05, ES = 0.42 [moderate]), whereas playing a 1‒4‒2‒3‒1 resulted in
greater values of ball possession in midfield zones compared to a 1‒4‒4‒2forma-
tion (p < 0.05, ES = 0.35 [moderate]). Teams playing a 1‒4‒2‒3‒1 reported a higher
ball possession percentage in attacking zones compared to a 1‒4‒3‒3formation
(p < 0.05, ES = 0.48 [moderate]).
Regarding running performance, only TDIP was higher when playing a 1‒4‒2‒
3‒1comparedtoa1‒4‒4‒2(p < 0.01, ES = 0.59 [large]) formation. The social
network analysis showed that playing a 1‒4‒2‒3‒1 and 1‒4‒3‒3resultedingreater
values of in- and out-degree than a 1‒4‒4‒2 formation (p < 0.05, ES = 0.60‒0.72
[large]). However, teams using a 1‒4‒4‒2 formation presented higher betweenness
centrality compared to the 1‒4‒2‒3‒1 and 1‒4‒3‒3 (p < 0.05, ES = 0.31‒0.34
[moderate]). Finally, teams playing 1‒4‒2‒3‒1 and 1‒4‒3‒3 formations reported
greater values of clustering coefficients, density and completed passes compared to
a1‒4‒4‒2(p=0.01‒0.05, ES = 0.97‒1.00 [very large]).
Table 1. Means (standard deviation) for ball possession variables, match running performance and
social network analysis according to match outcome during 2018 Russia FIFA World Cup.
Variables
Loss
(n = 420)
Draw
(n = 144)
Win
(n = 424)
Ball Possession (%) 49.2 (10.0) 49.8 (13.2) 50.5 (9.9)
Ball Possession (defensive zone ‒%) 27.1 (7.5) 28.1 (9.0) 27.1 (7.1)
Ball Possession (midfield zone ‒%) 52.6 (7.9) 51.6 (6.2) 52.8 (6.8)
Ball Possession (attack zone ‒%) 20.2 (6.1) 20.3 (5.3) 20.1 (5.1)
TD (m) 9355.7 (2433.1) 9212.0 (2126.0) 9365.6 (2366.7)
TDIP (m) 3400.7 (1165.6) 3105.5 (1351.0) 3531.1 (1282.2)
TDOP (m) 3859.0 (1336.2) 3862.0 (1415.4) 3669.0 (1321.5)
Sprints (n°) 27.8 (14.3) 27.0 (14.7) 27.0 (14.6)
Top Speed (km.h
‒1
) 27.5 (4.2) 27.3 (4.8) 27.3 (4.7)
Distance Z1 (m) 3775.1 (1562.9) 3550.9 (364.8) 3788.7 (537.0)
Distance Z2 (m) 3864.5 (1428.7) 3833.8 (1326.6) 3814.0 (1427.1)
Distance Z3 (m) 1197.1 (546.8) 1204.1 (551.5) 1169.4 (725.6)
Distance Z4 (m) 468.5 (231.0) 465.4 (238.0) 454.7 (234.6)
Distance Z5 (m) 169.1 (118.7) 161.6 (117.3) 171.0 (130.2)
In degree 26.3 (18.3) 29.3 (19.1) 26.3 (16.8)
Out degree 26.2 (18.5) 29.2 (19.5) 26.4 (17.2)
Closeness Centrality 0.838 (0.115) 0.845 (0.113) 0.832 (0.112)
Betweenness Centrality 1.533 (1.437) 1.493 (1.503) 1.631 (1.483)
Eigenvector 0.823 (0.175) 0.838 (0.151) 0.823 (0.162)
Clustering Coefficients 0.778 (0.082) 0.792 (0.075)
a
0.799 (0.089)
b
Density 0.775 (0.081) 0.781 (0.078) 0.796 (0.086)
Completed Passes 222.9 (108.7) 250.9 (113.1) 230.2 (108.2)
Note: TD = Total Distance Covered; Total Distance covered In Possession = TDIP; Total Distance covered when Out of
Possession = TDOP; Distance Z1 = 0‒7 km.h
‒1
; Distance Z2 = 7.1‒15 km.h
‒1
; Distance Z3 = 15.1‒20 km.h
‒1
; Distance
Z4 = 20.1‒25 km.h
‒1
; Distance Z5 = > 25.1 km.h
‒1. a
= Draw > Loss (p < 0.05);
b
= Win > Loss (p < 0.05).
INTERNATIONAL JOURNAL OF PERFORMANCE ANALYSIS IN SPORT 1031
Table 2. Means (standard deviation) for ball possession variables, match running performance and social network analysis according to playing formation during
2018 Russia FIFA World Cup.
Variables
1‒4‒2‒3‒1
(n = 458)
1‒4‒3‒2‒1
(n = 79)
1‒3‒4‒3
(n = 84)
1‒4‒3‒3
(n = 104)
1‒4‒4‒2
(n = 137)
1‒3‒3‒2‒2
(n = 95)
Ball Possession (%) 52.0 (10.5)
a,b
46.1 (12.3) 49.5 (7.9) 50.1 (7.0) 44.9 (10.9) 53.9 (7.5)
c,d
Ball Possession (defensive zone ‒%) 26.5 (7.6) 26.8 (7.4) 29.5 (5.9)
e,f
29.7 (7.5)
g,h
28.6 (8.6)
i
24.4 (6.1)
Ball Possession (midfield zone ‒%) 52.3 (5.4)
a
55.1 (7.5)
j,k,l
56.3 (12.6)
e,m,n
51.7 (6.2) 50.1 (7.4) 56.2 (6.3)
c,d,o
Ball Possession (attack zone ‒%) 21.6 (6.4)
b,p,q,r
19 (4.8) 16.9 (4.0) 19.0 (4.1) 20.8 (4.5)
s
19.3 (4.4)
TD (m) 9409.7 (2380.4) 9101.6 (2423.8) 9241.2 (2089.9) 9233.1 (2400.7) 9103.5 (1984.3) 9531.9 (2651.3)
TDIP (m) 3695.4 (1364.2)
a,b
3134.3 (1242.1) 3448.3 (987.5) 3315 (1134.6) 2994.2 (966.0) 3765.4 (1068.6)
d,t
TDOP (m) 3675.0 (1290.8) 3902.0 (1310.7) 3692.5 (1117.0) 3813.8 (1271.0) 3887.1 (1316.3) 3570.4 (1347.4)
Sprints (repetitions) 27.8 (14.7) 27.2 (15.6) 27.4 (14.9) 28.4 (13.8) 25.5 (12.7) 27.5 (15.3)
Top Speed (km.h
‒1
) 27.3 (4.5) 27.4 (4.6) 27.8 (4.2) 27.6 (5.0) 27.2 (4.5) 27.8 (4.1)
Distance Z1 (m) 3715.1 (536.7) 4107.8 (3427.2) 3664.3 (403.7) 3685.9 (488.2) 3690.0 (416.3) 3849.5 (607.8)
Distance Z2 (m) 3877.2 (1433.4) 3523.9 (1436.3) 3823 (1268.1) 3831.5 (1359.5) 3703.4 (1278.6) 3937.8 (1475.5)
Distance Z3 (m) 1217.3 (726.2) 1132.6 (562.3) 1165.0 (488.0) 1162.8 (538.3) 1123.4 (498.2) 1200.0 (580.0)
Distance Z4 (m) 471.9 (237.8) 466.0 (248.1) 457.6 (236.6) 471.2 (219.2) 429.1 (210) 458.2 (244.8)
Distance Z5 (m) 170.5 (122.5) 177.8 (131.1) 176.8 (138.4) 176.8 (110.1) 141.8 (104.3) 179.8 (141.9)
In degree 29.3 (19.2)
a
26.4 (21.3)
l
28.7 (16.0)
u
25.3 (13.1)
v
17.7 (11.9) 30.9 (18.2)
d
Out degree 29.3 (19.5)
a
26.4 (20.8)
l
28.7 (16.3)
u
25.3 (13.6)
v
18.0 (12.6) 30.9 (19.1)
d
Closeness Centrality 0.842 (0.116)
a
0.806 (0.136) 0.857 (0.111)
u
0.838 (0.102) 0.799 (0.114) 0.865 (0.115)
d,t
Betweenness Centrality 1.491 (1.300) 1.984 (2.180)
l
1.393 (1.744) 1.506 (1.012) 1.999 (1.781)
w,i
1.264 (1.142)
Eigenvector 0.839 (0.158) 0.810 (0.185) 0.851 (0.159) 0.835 (0.155) 0.788 (0.188) 0.842 (0.174)
Clustering Coefficients 0.798 (0.072)
a
0.749 (0.112) 0.813 (0.071) 0.800 (0.030) 0.727 (0.085) 0.821 (0.065)
Density 0.790 (0.067)
a
0.738 (0.120) 0.807 (0.091) 0.794 (0.021) 0.727 (0.080) 0.811 (0.077)
Completed Passes 256.0 (119.8)
a
208.7 (146.4) 241.1 (83.8) 219.3 (72.5) 161.5 (69.0) 241.3 (82.2)
Note: TD = Total Distance Covered; Total Distance covered In Possession = TDIP; Total Distance covered when Out of Possession = TDOP; Distance Z1 = 0‒7 km.h
‒1
; Distance Z2 = 7.1‒15 km.h
‒1
;
Distance Z3 = 15.1‒20 km.h
‒1
; Distance Z4 = 20.1‒25 km.h
‒1
; Distance Z5 = > 25.1 km.h
‒1; a
=1‒4‒2‒3‒1>1‒4‒4‒2;
b
=1‒4‒2‒3‒1>1‒4‒3‒2‒1;
c
=1‒3‒3‒2‒2>1‒4‒2‒3‒1;
d
=1‒3‒3‒
2‒2>1‒4‒4‒2;
e
=1‒3‒4‒3>1‒4‒2‒3‒1;
f
=1‒3‒4‒3>1‒3‒3‒2‒2;
g
=1‒4‒3‒3>1‒4‒2‒3‒1;
h
=1‒4‒3‒3>1‒3‒3‒2‒2;
i
=1‒4‒4‒2>1‒3‒3‒2‒2;
j
=1‒4‒3‒2‒1>1‒4‒2‒3‒1;
k
=1‒4‒
3‒2‒1>1‒4‒3‒3;
l
=1‒4‒3‒2‒1>1‒4‒4‒2;
m
=1‒3‒4‒3>1‒4‒3‒3;
n
=1‒3‒4‒3>1‒4‒4‒2; ° = 1‒3‒3‒2‒2>1‒4‒3‒3;
p
=1‒4‒2‒3‒1>1‒3‒4‒3;
q
=1‒4‒2‒3‒1>1‒4‒3‒3;
r
=1‒4‒2‒
3‒1>1‒3‒3‒2‒2;
s
=1‒4‒4‒2>1‒3‒4‒3;
t
=1‒3‒3‒2‒2>1‒4‒3‒2‒1;
u
=1‒3‒4‒3>1‒4‒4‒2;
v
=1‒4‒3‒3>1‒4‒4‒2;
w
=1‒4‒4‒2>1‒4‒2‒3‒1.
1032 R. AQUINO ET AL.
3.3. Relationships between TDIP and player prominence
When all positional roles were polled together, TDIP was only related to in- and out-
degree (r = 0.52–0.69, p < 0.001) (Figure 1). Relationships between match running
demands and individual metrics from the social network analysis were more clearly
position-dependent, with large to very-large correlation coefficients for CD, ED, CM and
EM (e.g., TDIP vs. in-, out-degree: r = 0.59–0.72, p < 0.001) but small to moderate for
F and GK (e.g., TDIP vs. in-, out-degree: r = 0.24–0.38, p = 0.009 to p < 0.001).
4. Discussion
To our knowledge, this is the first study to investigate the relationship between ball
possession, running demands, team cooperation patterns and match outcome and play-
ing formation during a soccer world cup. Main findings were: i) the majority of the
variables investigated did not affect match outcome although clustering coefficients were
the exception (win = draw > loss; small‒moderate effect size); ii) a 1‒4‒2‒3‒1 formation
was most frequently used by teams (46.4%), followed by a 1‒4‒4‒2 (13.9%) and 1‒4‒3‒3
(10.5%); iii) in general, teams playing a 1‒4‒2‒3‒1 formation reported greater values for
ball possession, TDIP, and micro/macro network measures (with exception to between-
ness centrality) compared to a 1‒4‒4‒2 and 1‒4‒3‒3; iv) While the TDIP was generally
related to the majority of micro level network variables (i.e. player prominence), the
magnitude of coefficients varied considerably according to network measures and play-
ing positions.
Figure 1. Correlation coefficients (90% confidence interval) between total distance covered in
possession (TDIP) and individual metrics from social network analysis during the 2018 Russia FIFA
World Cup (n = 61 matches). Note: CD = Central Defenders; ED = External Defenders; CM = Central
Midfielders; EM = External Midfielders; F = Forwards; GK = Goalkeepers.
INTERNATIONAL JOURNAL OF PERFORMANCE ANALYSIS IN SPORT 1033
Here, greater values of clustering coefficients were observed when winning teams were
compared to losing ones. It is noteworthy that during the 2018 Russia FIFA World Cup,
France the world champion presented a mean of 0.746 arbitrary units (au) of clustering
coefficients, while Panama (one of the lowest-ranked teams) showed a mean of 0.644 au.
These results suggest that the French players presented higher levels of interconnectivity
between close teammates (greater cooperation). Clemente et al. (2015) also reported that
teams who attained the final stages of the 2014 World Cup presented high values of
clustering coefficients, which can lead to a high level of offensive efficacy. A study of
283,529 passes recorded in 760 English Premier League soccer matches demonstrated
that goals scored were associated with density and centralisation metrics; i.e. high levels
of coordination –density –leading to increased team performance whereas a centralised
interaction was associated with a decrease in team performance (Grund, 2012). In
contrast, previous studies focusing upon offensive phases and goals scored suggest that
reduced density may help attain better team performance (Mclean, Salmon, Gorman,
Stevens, & Solomon, 2018; Pina, Paulo, & Araújo, 2017). Therefore, the associations
observed between network measures and performance outcomes should be interpreted
with caution. These discrepancies across findings have been attributed, at least in part, to
differences across team playing styles (Clemente, 2018). Collectively, these results suggest
that SNA metrics show more potential for providing practical information about how
teams organise their offensive process (e.g. adopting a more direct playing style or a more
collective style through a positional attack or possession game) than information relating
to what discriminates between successful or unsuccessful teams. Successful teams can
adopt distinct playing styles according to their players at their disposal and their coach’s
philosophy. Therefore, the variety of successful game models can result in high variability
of micro/macro network measures.
Running demands reported here were not associated to match outcome during the
2018 Russia FIFA World Cup. For example, World Champion (France) covered a mean
of ~165 m in Z5 (distance > 25 km.h
‒1
) while Panama covered a mean of ~162 m in Z5.
However, the present study did not consider score-line and its effects on overall
running output during match-play [i.e., winning vs. drawing vs. losing status) as elite
players tend to perform more high-speed activity when losing vs. winning (e.g.,
Bloomfield et al., 2005; Castellano et al., 2011;Lagoetal.,2010)]. In this sense, the
differences in running performance and network metrics were more evident when team
formation was taken into consideration. Teams that adopted a 1–4–4–2 formation
during the 2018 Russia FIFA World Cup reported lower values for lower TD and
completed passes, and adopted a less collective behaviour compared to 1–4–2–3–1
formation. In addition, here the percentage of time spent in possession was superior in
a1–4–2–3–1(~52%)vs.a1–4–4–2 formation (~45%). Therefore, these findings tend to
confirm that playing formations govern player tactical-technical abilities and physical
efforts during elite soccer matches.
Furthermore, large to very-large correlations were observed between the TDIP and in
and out-degree individual networks metrics. These results suggest that a greater TDIP
might be related to the need for players to work harder to adopt useful positions on the
field thereby facilitating passing exchanges –in other words receiving and completing
passes. A previous study also reported similar relationships in matches from the Spanish
first division (Castellano & Echeazarra, 2019). In contrast, when the present CM were
1034 R. AQUINO ET AL.
well positioned in the field to connect with other two teammates (i.e., high values of
betweenness centrality), reduced running demands were observed (i.e., lower TDIP). CM
play an important role in constructing offensive phases requiring good positioning in the
field for inter-sectoral connectivity. In addition, when defensive players (CD, ED) and
EM were closer to other teammates and played a crucial role in the offensive phases (i.e.,
high values closeness centrality and eigenvector, respectively), greater physical require-
ments were reported for these players (i.e., high values for TDIP). Therefore, these
players need to adopt correct positions on the field (facilitating passing exchanges) so
they can maintain ball possession and create scoring situations, thus increasing physical
demands.
A limitation of the present study was that as only a single world cup tournament was
investigated, the relevance of these findings for future world cups may be limited. In addition,
we did not consider the influence of opposition playing formation on the present national
teams’performance. Finally, further studies should compare ball possession strategies,
running performance and SNA variables in different combinations of oppositions playing
formation (e.g., 1–4–2–3–1vs.1–4–4–2; 1–4–2–3–1vs.1–4–3–3; 1–4–4–2vs.1–4–3–3).
5. Practical application
In our opinion, the present results regarding team performance in the 2018 FIFA Russia
World Cup provide practical information to sports scientists and soccer practitioners in
three different ways. First, ball possession, running performance, player prominence and
team network properties were affected by playing formation. In contrast, these factors
were not linked to match outcome. Therefore, playing formation should be considered
when evaluating performance in elite competitions such as the FIFA World Cup. Second,
out of all the formations studied, a 1‒4‒2‒3‒1 was most frequently used by teams (46.4%),
resulting in high values of ball possession, TDIP, and micro/macro network measures
(the exception was betweenness centrality). Third, the relationships between player
prominence and TDIP were position-dependent. The ability to exchange passes, being
close to other teammates and play a crucial role in offensive phases required greater
physical demands, particularly in CD, ED, CM and EM. Nevertheless, the ability of CM
to connect two other teammates (e.g., connecting the ED within EM) required reduced
TDIP. This result suggests, at least in part, that good player positioning may result in
lower physical demands.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Rodrigo Aquino http://orcid.org/0000-0002-4885-7316
João Cláudio Machado http://orcid.org/0000-0001-9827-5296
Filipe Manuel Clemente http://orcid.org/0000-0001-9813-2842
Gibson Moreira Praça http://orcid.org/0000-0001-9971-7308
Luiz H. Palucci Vieira http://orcid.org/0000-0001-6981-756X
INTERNATIONAL JOURNAL OF PERFORMANCE ANALYSIS IN SPORT 1035
Enrico F. Puggina http://orcid.org/0000-0002-8379-2247
Christopher Carling http://orcid.org/0000-0002-7456-3493
References
Aquino, R., Puggina, E. F., Alves, I. S., & Garganta, J. (2017). Skill-related performance in soccer:
A systematic review. Human Movement,18(5), 3–24.
Argilaga, M. T. A., Villaseñor, A. B., Mendo, A. H., & López, J. L. L. (2011). Diseños observacio-
nales: Ajuste y aplicación en psicología del deporte. Cuadernos De Psicología Del Deporte (In
Spanish: English abstract),11(2), 63–76.
Bloomfield, J., Polman, R., & O’Donoghue, P. (2005). Effects of score-line on intensity of play in
midfield and forward players in the FA Premier League. Journal of Sports Sciences,23(2),
191–192.
Bradley, P. S., Carling, C., Archer, D., Roberts, J., Dodds, A., Di Mascio, M., . . . Krustrup, P. (2011).
The effect of playing formation on high-intensity running and technical profiles in English FA
Premier League soccer matches. Journal of Sports Sciences,29(8), 821–830.
Bradley, P. S., & Noakes, T. D. (2013). Match running performance fluctuations in elite soccer:
Indicative of fatigue, pacing or situational influences? Journal of Sports Sciences,31(15),
1627–1638.
Carling, C. (2013). Interpreting physical performance in professional soccer match-play: Should
we be more pragmatic in our approach? Sports Medicine,43(8), 655–663.
Carling, C., Reilly, T., & Williams, A. M. (2008). Performance assessment for field sports. Abingdon:
Routledge.
Castellano, J., Blanco-Villaseñor, A., & Alvarez, D. (2011). Contextual variables and time-motion
analysis in soccer. International Journal of Sports Medicine,32(6), 415–421.
Castellano, J., & Echeazarra, I. (2019). Network-based centrality measures and physical demands
in football regarding player position: Is there a connection? A preliminary study. Journal of
Sports Sciences. Advance online publication. doi: 10.1080/02640414.2019.1589919.
Clemente, F. M. (2018). Performance outcomes and their associations with network measures
during FIFA World Cup 2018. International Journal of Performance Analysis in Sport,18(6),
1010–1023.
Clemente, F. M., Martins, F. M. L., Kalamaras, D., Wong, P. D., & Mendes, R. S. (2015). General
network analysis of national soccer teams in FIFA World Cup 2014. International Journal of
Performance Analysis in Sport,15(1), 80–96.
Clemente, F. M., Martins, F. M. L., & Mendes, R. S. (2016). Social network analysis applied to team
sports analysis. New York: Springer.
Clemente, F. M., Martins, F. M. L., Wong, P. D., Kalamaras, D., & Mendes, R. S. (2015). Midfielder
as the prominent participant in the building attack: A network analysis of national teams in
FIFA World Cup 2014. International Journal of Performance Analysis in Sport,15(2), 704–722.
Cohen, J. (1988). Statistical Power Analysis for the Behavioural Sciences (2nd ed.). New York:
Routledge.
Collet, C. (2013). The possession game? A comparative analysis of ball retention and team success
in European and international football, 2007–2010. Journal of Sports Sciences,31(2), 123–136.
Couceiro, M. S., Dias, G., Araújo, D., & Davids, K. (2016). The ARCANE project: How an
ecological dynamics framework can enhance performance assessment and prediction in
football. Sports Medicine,46(12), 1781–1786.
da Mota, G. R., Thiengo, C. R., Gimenes, S. V., & Bradley, P. S. (2016). The effects of ball
possession status on physical and technical indicators during the 2014 FIFA World Cup
Finals. Journal of Sports Sciences,34(6), 493–500.
FIFA. (2018). FIFA World Cup Russia 2018. Available at: https://www.fifa.com/worldcup/
(accessed August 20, 2018).
Freeman, L. (2004). The development of social network analysis: A study in the Sociology of Science.
Vancouver: Empirical Press.
1036 R. AQUINO ET AL.
Glazier, P. S. (2010). Game, set and match? Substantive issues and future directions in performance
analysis. Sports Medicine,40(8), 625–634.
Grund, T. U. (2012). Network structure and team performance: The case of English Premier
League soccer teams. Social Networks,34(4), 682–690.
Hopkins, W. G. (2000). Measures of reliability in sports medicine and science. Sports Medicine,30
(1), 1–15.
Lago, C., Casais, L., Dominguez, E., & Sampaio, J. (2010). The effects of situational variables on
distance covered at various speeds in elite soccer. European Journal of Sport Science,10(2),
103–109.
Linke, D., Link, D., & Lames, M. (2018). Validation of electronic performance and tracking
systems EPTS under field conditions. PloS One,13(7), e0199519.
Mclean, S., Salmon, P. M., Gorman, A. D., Stevens, N. J., & Solomon, C. (2018). A social network
analysis of the goal scoring passing networks of the 2016 European Football Championships.
Human Movement Science,57, 400–408.
Paul, D. J., Bradley, P. S., & Nassis, G. P. (2015). Factors affecting match running performance of
elite soccer players: Shedding some light on the complexity. International Journal of Sports
Physiology and Performance,10(4), 516–519.
Pina, T. J., Paulo, A., & Araújo, D. (2017). Network characteristics of successful performance in
association football. A study on the UEFA Champions League. Frontiers in Psychology,8, 1173.
Praça, G. M., Lima, B. B., Bredt, S. G. T., Clemente, F. M., & Andrade, A. G. (2019). Influence of
match status on players’prominence and teams’network properties during 2018 FIFA World
Cup. Frontiers in Psychology,10, 695.
Ribeiro, J., Davids, K., Araújo, D., Silva, P., Ramos, J., Lopes, R., & Garganta, J. (2019). The role of
hypernetworks as a multilevel methodology for modelling and understanding dynamics of team
sports performance. Sports Medicine. Advance online publication. doi: 10.1007/s40279-019-
01104-x.
Ribeiro, J., Silva, P., Duarte, R., Davids, K., & Garganta, J. (2017). Team sports performance
analysed through the lens of social network theory: Implications for research and practice.
Sports Medicine,47(9), 1689–1696.
Vilar, L., Araújo, D., Davids, K., & Button, C. (2012). The role of ecological dynamics in analysing
performance in team sports. Sports Medicine,42(1), 1–10.
INTERNATIONAL JOURNAL OF PERFORMANCE ANALYSIS IN SPORT 1037