Using passing network measures to determine the performance difference between foreign and domestic outfielder players in Chinese Football Super League

Abstract

The study was aimed to identify the passing performances related to social network between domestic (Chinese) and foreign (non-Chinese) players that contributed to the offensive play according to playing positions in Chinese Football Super League. Data were collected from all 16 teams, with 63,446 passes from 8885 player observations of 346 unique players (258 domestic and 88 foreign) being filtered. Nine network metrics were adopted to assess the passing performance between two player groups. The results revealed that foreign players, especially midfielders and forwards, had increased values than domestic counterparts in passes in, indegree, stress, partner and betweenness centrality (standardised Cohen’s d range: 0.18–0.62; inference: possibly to most likely). Moreover, foreign midfielders also demonstrated higher values in passes out, outdegree and closeness centrality (0.30–0.56; likely to most likely). The analysis of passing performance allows for a better understanding of team’s attacking properties and highlights the prominent role foreign players act within the build of attack. It is possible that recruiting high-level overseas players contribute to Chinese teams’ goal scoring performance, but coaching staffs and team managers should take cautions in terms of the degree of reliance on them, as the development of domestic midfielders and forwards would be compromised.

Keywords

Association football foreign players labour migration performance analysis soccer sport analytics

Introduction

Computer-based approaches have been applied to collect and analyse technical, tactical and physical aspects in team sport, especially in football, and progressively increased the information and dimensions related to the study of match performance behaviours.^1–3 Recently, the general structure of football teams and connections among players during match-play were investigated using social network analysis (SNA).^4–7 Contrast to the traditional notational analysis, the SNA considers and models passing distribution of teams and is proved to be useful since its concepts and methods could not only illuminate player’s or team’s dynamics, but also improve the understanding of complex relational interactions within the team.^8,9 Specifically, this promising body of research helped to understand the most prominent players’ positions that best contributed to the build of attack during football matches,¹⁰ analysed the difference in network density between two halves in the match¹¹ and investigated the number of passes per minute and the clustering coefficient of teams during the FIFA World Cup.¹²

Generally, network analysis is based on the accumulation of passes in which happen numerously in every match irrespective of the quality of the teams.¹³ The passing network of a football team consists of the players as vertices and the passes between the players as the edges. Prior studies adopting this approach went beyond simplistic frequency counts of passes made and provide insights into teams’ passing flow, player interaction and prominent player during offensive play by using metrics such as degree centrality, betweenness centrality (BC) and closeness centrality (CC).^14,15 In particular, Grund⁶ found that high network intensity and low centralisation were related to better team performance in Premier League. Clemente et al.¹⁶ analysed 128 passing adjacency matrices from 2014 World Cup and concluded that midfielders played the most important role in building the attack. A more recent study analysing the last 2018 World Cup showed that teams adapted their passing microstructures according to different match status, and a more direct play style was preferred when teams were losing.⁹ However, among available literature, it appears that relative analysis of foreign transferred players in football developing countries, such as in Asia (e.g. China), was seldom addressed.¹⁷

The phenomenon of labour migration is a salient feature of modern sports that has been investigated by scholars for over a decade covering a wide range of sports. For instance, the subject of football has also been relatively well researched,^18–25 and it is positively established that migration in sport is a complex and multidimensional process and a part of global sport systems.^23,26 The societal multi-dimensionality of migratory issues, however, has led to the proliferation of academic interest with often divergent views on the subject.^22,27 This migration is influenced by foreign player policies with differences noted between some competitions. For example, more restrictive policies to sign foreign players are experienced in Asia compared to other competitions.²⁸ There is a great emphasis on including domestic players and limited foreign players per team within the Korean, Japanese and Chinese competitions.²⁸ As the first league of a football developing country, the Chinese Super League (CSL) is a representative of such phenomenon and recruits many top foreign players each year to improve team quality and obtain success.²⁹ Under such circumstances, the Chinese Football Association had to establish a foreign player policy since 2017 CSL, stating that each team may have a maximum of four foreign outfield players, and during the match, a team should have no more than three foreign players on the field at any given minute. Given this specific policy, one that is more restrictive than most other domestic elite competitions,³⁰ it is expected that teams within the CSL may exhibit distinctive performance profiles with differences potentially notable between domestic and foreign players according to playing positions, and determining the prominence of foreign and domestic players could better inform team’s tactical training and improve inter-player dynamics.

Therefore, based on the above rationale, this study tried to provide the knowledge about the match performance of foreign players in Asian professional football league and look into their roles within the team’s network. The specific aim was to investigate the difference in match performances between foreign and domestic outfield players within team’s passing network from CSL, regarding different playing positions. It was hypothesised that foreign players of all playing positions outperformed domestic players in all network-related metrics.

Materials and methods

Sample and variables

By undertaking an observational design, this study included all 240 matches from the CSL 2017 season. The inter-player passing statistics of all matches were tracked and provided by the Shanghai Champion Information Technology Co., Ltd. through a validated data collection system (Champdas Master System, http://www.champdas.com) operated by trained collectors, and they have at least two years of experience in operating the system.³ The starting players and substitutes were all included. However, as the study only considered the passing performance of outfield players when teams were in possession of the ball, after organising the data, any inter-player passing count less than five passes was ruled out, and goalkeeper-involved passes were also excluded during the filtering process. Finally, 63,446 passes were filtered, and they belonged to 8885 player observations of 346 unique players (258 domestic and 88 foreign) from all 16 CSL teams.

The players were further divided into three different groups according to their positions based on the tactical line-up of each CSL team: defender (2442 domestic player observations from 103 unique players and 379 foreign player observations from 19 unique players), midfielder (3257 from 120 unique players and 1407 from 28 unique players, respectively) and forward (481 from 35 unique players and 919 from 41 unique players, respectively). All players’ positions were determined in the match statistics offered by the data provider, so that positions such as wingers, forwards and strikers were assigned into the category of forward; attacking midfielders, centre midfielders, left and right midfielders were classified as midfielders; while left and right fullbacks, centre backs and sweepers were as defenders. The personal characteristics for both domestic and foreign players of these three categories are shown in Table 1. Based on the number of matches contested, a total of 480 adjacency matrices were established, and the corresponded network graphs were generated. This matrix represents the connections between a player and an adjacency teammate.¹⁶

Table 1.

Personal characteristics for CSL domestic and foreign players.

SNA variables	Defender		Midfielder		Forward
	Domestic (n = 1026–1272)	Foreign (n = 160–176)	Domestic (n = 1272–1357)	Foreign (n = 397–411)	Domestic (n = 295–300)	Foreign (n = 520–555)
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Height (cm)	185.3 (4.1	188.1 (4.7)	179.3 (5.1)	181.3 (5.1)	182.7 (4.8)	180.4 (6.8)
Weight (kg)	77.1 (5.1)	82.1 (6.3)	71.8 (5.7)	74.8 (5.4)	72.5 (5.2)	74.2 (6.0)
Age (years)	26.9 (3.8)	29.3 (3.1)	26.9 (3.8)	29.1 (2.5)	26.7 (4.4)	27.8 (3.7)
Time played (%)	96.6 (11.1)	97.9 (8.2)	86.4 (19.3)	97.9 (8.2)	82.0 (20.3)	97.1 (9.2)

The following two passing-related variables and seven SNA metrics were used to evaluate players’ passing performance and their importance within team’s passing network.

Passing performance variables:

Passes out: total number of passes that the player realised effectively to his teammates during the match.

Passes in: total number of passes that the player received effectively from teammates during the match.

Generally, SNA contains various centrality metrics to evaluate the importance of certain node (player) inside a given social network. But considering the practical application of SNA to football analysis, the study just applied the following variables that are suitable for the interpretation of players’ importance within a passing network:^14,15,31,32

Neighbourhood connectivity (NC): in SNA, the connectivity of a node is determined by the number of its neighbours. Therefore, the NC of a player in football is defined as the average connectivity of all teammates of that player.

In-degree centrality (IDC): the count of inbound edges (arcs) from all neighbouring nodes (teammates) to a certain node (player);

Out-degree centrality (ODC): the count of outbound edges (arcs) from a certain node (player) to all nodes (players) that are connected to it (neighbours);

Stress centrality (SC): the metric is the summa of the number of shortest paths passing through n. A player has a high stress if it is traversed by a high number of shortest paths

Stress (A) = \sum_{\begin{array}{l} i, j \in V \\ i \neq j \neq A \end{array}} g_{i j} (A)

where g_ij(A) is the number of shortest paths between player i and player j that pass through player A.

5. Partner (partner of multi-edged node pairs): the value indicates if certain node (player) is a partner of node pairs with multiple edges.

6. Betweenness centrality (BC): it tries to measure the extent of the control that each node (player) holds over the network by considering the shortest paths between all pairs of nodes (players). In other words, this metric indicates the amount of network that a particular player ‘controls’ and is one of the most meaningful measure among other metrics because it successfully quantifies how often each player lies between other players of the passing network, acting as a mediator or ‘bridge’ for them

Betweenness (A) = \sum_{i \neq A \neq j} (\frac{σ_{i j} (A)}{σ_{i j}})

where i and j are players in the passing network different from player A, σ_ij denotes the number of shortest paths from player i to player j and σ_ij (A) is the number of shortest paths from i to j that A lies on.

7. Closeness centrality (CC): This index attempts to quantify actor importance in terms of their total graph theoretic distance in the social network. It indicates how easy it is for a player to be connected with teammates (by passing relation); therefore, that player is requested by the team as a target to pass the ball. Thus, it provides a direct measurement on how easy it is to reach a particular player within a team. A high closeness score corresponds to a small average distance, indicating a well-connected player within the team

Closeness (A) = \frac{1}{\sum_{i \in V} L (A, i)}

where L(A,i) is the length of the shortest path between two players A and i, the formula then calculates the summa of player A and other teammates and the reciprocal of the value is obtained. Players with low closeness score have little proximity to others.

Statistical analysis

After determining the normal distribution of the data (via Kolmogorov–Smirnov test), through nine passing network-related variables that measure player’s prominence inside a team’s network were calculated and were compared between domestic and foreign players via standardised (Cohen’s d) mean differences, computed with pooled variance and respective 90% confidence intervals (CI). Uncertainty in the true differences of the comparisons was assessed using the non-clinical magnitude-based decisions,³³ and the magnitudes of clear differences were assessed as follows: <0.20, trivial; 0.20–0.60, small; 0.61–1.20, moderate; 1.21–2.0, large and >2.0, very large.³⁴ Differences were deemed clear if the CI for the difference in the means did not include positive and negative values (±0.2 times the standardisation estimated from between-subject standard deviation) simultaneously.³⁵

Results

Table 2 represents the descriptive statistics of all passing network metrics for domestic and foreign players of different playing positions. Foreign midfielders and forwards presented higher values than domestic players in passes in, indegree, SC, partner and BC (standardised Cohen’s d range: 0.18–0.62; inference: possibly to most likely). Moreover, foreign midfielders also demonstrated higher values in passes out, outdegree and closeness centrality (0.30–0.56; likely to most likely). On the contrary, while both domestic defenders and midfielders had higher values of NC than foreign counterparts (–0.29 to –0.18; possibly to likely), domestic defenders were also shown to possess higher passes in, indegree and partner (–0.25 to –0.19; possibly to likely) (see Figure 1).

Table 2.

Descriptive statistics of passing and SNA related variables for CSL domestic and foreign players.

SNA variables	Defender		Midfielder		Forward
	Domestic (n = 1026–1272)	Foreign (n = 160–176)	Domestic (n = 1272–1357)	Foreign (n = 397–411)	Domestic (n = 295–300)	Foreign (n = 520–555)
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Passes out	16.00 (10.82)	15.72 (11.74)	17.50 (12.50)	26.05 (17.51)	12.38 (9.86)	12.90 (9.09)
Passes in	14.88 (10.01)	13.11 (8.96)	17.32 (12.34)	24.2 (16.22)	13.19 (10.46)	16.01 (10.94)
Neighborhood	3.97 (1.61)	3.72 (1.51)	3.93 (1.59)	3.49 (1.48)	4.10 (1.89)	3.86 (1.69)
Out-degree	1.96 (2.56)	1.96 (1.49)	1.94 (2.49)	3.14 (5.14)	1.45 (1.52)	1.54 (1.54)
In-degree	1.69 (2.11)	1.27 (1.18)	2.11 (2.28)	3.03 (3.69)	1.54 (1.66)	1.86 (1.80)
Stress	6.00 (9.41)	5.69 (9.46)	7.36 (11.43)	13.71 (15.73)	2.91 (5.29)	4.08 (7.93)
Partner	0.88 (1.21)	0.66 (0.96)	1.05 (1.37)	1.61 (1.71)	0.55 (0.87)	0.71 (1.05)
Betweenness	0.06 (0.09)	0.06 (0.11)	0.08 (0.12)	0.18 (0.17)	0.04 (0.07)	0.06 (0.10)
Closenness	0.48 (0.27)	0.51 (0.26)	0.45 (0.31)	0.57 (0.30)	0.45 (0.34)	0.47 (0.34)

Figure 1.

Effect sizes [90% CI] of differences in the mean counts between domestic and foreign players, when bars of one variable crossed the negative and positive smallest worthwhile change threshold in the meantime, the effect was unclear. Asterisks indicate the likelihood for the magnitude of the true differences in mean as follows: *possible; **likely; ***very likely; ****most likely. Asterisks located in the trivial area denote for trivial differences.

Discussion

The aim of this study identified the most prominent players that build the attack of a team within the CSL according to playing positions (defender, midfielder and forward) and nationality (foreign and Chinese). The results of this study showed that foreign players played a prominent role in the offensive play of CSL teams and contributed to the build of attack, especially on midfielders. Moreover, coach must take into consideration the individual contribution of each player for the overall connection of the team can be important indicator that may increase the possibility of optimising the tactical performance of football players. To the best of our knowledge, this is the first study to analyse the collective behaviour of team via SNA according to playing position in the CSL.

For defender, the results showed that all SNA-related variables were similar between domestic and foreign players except for IDC, which is an important variable that should be considered in match analysis. In particular, this player’s position receives most of the passes from their teammates. The domestic players showed higher values in IDC than their counterparts, which is in line with the previous study of Clemente et al.¹⁶ who reported that the high value related to IDC was found in the central defender based on 1-3-5-2 playing formation. At the meantime, it is highly probable that most CSL teams prefer to contract more foreign players that could instantly improve attacking quality and make them the priority in the attacking process, given the current league’s policy that limits the number of on-pitch foreign players in the match and the outnumbered foreign defender observations throughout the whole 2017 CSL season. Therefore, domestic players might have played more important role in the defending process as central defender mainly showed high volume of passes in the first phase of building the attack for increasing the connection among teammates in the defensive zone.¹⁶

For midfielders, the current results are in line with previous studies,^11,36 where the midfielders demonstrated greater values for most of SNA-related variables such as IDC, ODC, CC and BC, compared to the others playing positions (e.g. defender and forward). The highest volume of passes between such players’ positions, mainly in the offensive phase, increases the connection among teammates in the middle sectors. Likewise, our finding also showed that all SNA-related variables significantly differentiated domestic and foreign players. More specifically, foreign players showed high values than their domestic counterparts in most variables except for Neighbourhood, revealing that foreign midfielders not only promoted connectivity between all teammates, but also played the prominent role in the offensive phase, fostering a global cooperation among the team. Although domestic players demonstrated comparatively higher value in Neighbourhood, it is likely that rather than making key passes to opponent’s defensive zones, they are generally tending to make more “safe” passes that are less tackled, to defending players, as most of them are still refining their passing effectiveness, tactical creativity and decision-making efficiency.¹⁷ However, the current study contrasts with the previous study of Gai et al.,¹⁷ where the differences of playing styles could indicate a more collective style of play for foreign players in their home countries while domestic players reflected a minor tactical focus. Despite the disparities of playing styles during formative years between foreign and domestic players,³⁰ domestic players still try to cover the limitations of their less technical and tactical skills giving the teammates frequent cooperation with each other. Moreover, as team managers of CSL league are prone to design most offensive tactics emphasising foreign players’ major participation, domestic midfielders would not get as many chances as their foreign peers to attempt more penetrative passes to forwards or receive more passes from defenders. This phenomenon is mirrored by that greater values of foreign players in CC and BC, which are more meaningful indexes than the other degree centrality metrics, such as IDC and ODC.³⁷

For forwards, the results showed that all SNA-related variables were similar between domestic and foreign players except for pass in. The current study is in line with the previous study,³⁶ which indicated that forwards can always be identified as those players having the lowest closeness and betweenness, as they are isolated players waiting to receive passes as well as players who get replaced more often. Particularly, the current study confirmed that forwards received more passes from their teammates, especially for foreign forwards, because of the limits of the number of foreign player’s policy in the CSL, the teams likely recruit these players to key positions related to goal scoring (forward), decision-making and passing skills (central midfielder) and defensive actions (central defender). Additionally, the imbalanced recruitment of teams from the CSL is focused on those playing positions (e.g. central defender, central midfielder and forward) associated with team’s success.²⁸

The current study has some limitations that should be taken into consideration. First, this study only considered three playing positions (defender, midfielder and forward) provided by data provider, which somehow limits the value of study’s finding when more practical applications are expected from the analysis of specific positions. Further research should include a wider range of playing positions (e.g. central defender and defensive midfielder), because those playing positions may be the prominent players for the attacking process.³⁷ Second, the current study fails to identify the key (top) player who has different levels of abilities, which could be alternatively represented by their values on transfer market. And this may increase or decrease the individual contribution of a player to the team.⁷ Last, the situational variables (e.g. team quality and match status) should be explored to identify their influence on the network properties in future analyses.

Conclusion

Using social network-related metrics is helpful to identify how players connect with each other and compare the strength of the connection between domestic and foreign players. Foreign players played a prominent role in the offensive play of CSL teams and contributed to the attack organisation. However, cautions should be taken for stake-holders of the league and clubs, given that an over-relying on foreign players would probably hinder the development of domestic players. This study provided useful information for coaching staffs in training and preparing the appropriate strategies during the matches.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by National Key R&D Program of P.R. China (2018YFC2000600). The corresponding author was supported by the China Postdoctoral Science Foundation and the Fundamental Research Funds for the Central Universities (2019QD033).

ORCID iD

Yixiong Cui

References

Grünz

Haas

Soukup

, et al. Structural features and bioavailability of four flavonoids and their implications for lifespan-extending and antioxidant actions in C. elegans. Mech Ageing Dev 2012; 133: 1–10.

Memmert

Lemmink

Sampaio

Current approaches to tactical performance analyses in soccer using position data.

Sport Med 2017; 47: 1–10.

Gong

Cui

Gai

, et al. The validity and reliability of live football match statistics from Champdas master match analysis system. Front Psychol 2019; 10: 1339.

Bourbousson

Poizat

Saury

, et al. Team coordination in basketball: Description of the cognitive connections among teammates. J Appl Sport Psychol 2010; 22: 150–166.

Duch

Waitzman

Amaral

LAN.

Quantifying the performance of individual players in a team activity.

PloS One 2010; 5: e10937.

Grund

TU.

Network structure and team performance: the case of English Premier League soccer teams. Social Netw 2012; 34: 682–690.

Clemente

Couceiro

Martins

FML

, et al. Using network metrics to investigate football team players’ connections: a pilot study. Motriz Revis Educ Física 2014; 20: 262–271.

Ribeiro

Davids

Araújo

, et al. The role of hypernetworks as a multilevel methodology for modelling and understanding dynamics of team sports performance. Sports Med 2019; 49: 1337–1344.

Praça

Lima

Bredt

SdGT

, et al. Influence of match status on players’ prominence and teams’ network properties during 2018 FIFA World Cup. Front Psychol 2019; 10: 695.

10.

Clemente

Martins

FML

Kalamaras

, et al. General network analysis of national soccer teams in FIFA World Cup 2014. Int J Perform Anal Sport 2015; 15: 80–96.

11.

Clemente

Couceiro

Martins

FML

, et al. Using network metrics in soccer: a macro-analysis. J Hum Kinet 2015; 45: 123–134.

12.

Cotta

Hewett

, et al. Bounds on dark matter interactions with electroweak gauge bosons. Phys Rev D 2013; 88: 116009.

13.

Castellano

Echeazarra

Network-based centrality measures and physical demands in football regarding player position: is there a connection? A preliminary study.

J Sports Sci 2019; 37: 2631–2638.

14.

Ribeiro

Silva

Duarte

, et al. Team sports performance analysed through the lens of social network theory: implications for research and practice. Sports Med 2017; 47: 1689–1696.

15.

Gonçalves

Coutinho

Santos

, et al. Exploring team passing networks and player movement dynamics in youth association football. PLOS One 2017; 12: e0171156.

16.

Clemente

Martins

FML

Wong

, et al. Midfielder as the prominent participant in the building attack: a network analysis of national teams in FIFA World Cup 2014. Int J Perform Anal Sport 2015; 15: 704–722.

17.

Gai

Leicht

Lago

, et al. Physical and technical differences between domestic and foreign soccer players according to playing positions in the China Super League. Res Sport Med 2019; 27: 314–325.

18.

Lanfranchi P. The migration of footballers -- the case of France, 1932-1982. In: Bale J and Maguire J (eds) The global sports arena: athletic talent migration in an interdependent world. London, UK: Frank Cass, 1994, pp.63--77.

19.

Maguire

Pearton

The impact of elite labour migration on the identification, selection and development of European soccer players.

J Sport Sci 2000; 18: 759–769.

20.

Stead

Maguire

“Rite De Passage” or passage to riches? The motivation and objectives of Nordic/Scandinavian players in English league soccer. J Sport Social Issue 2000; 24: 36–60.

21.

Lanfranchi

Taylor

Moving with the ball: the migration of professional footballers. Oxford, UK: Berg, 2001.

22.

Magee

Sugden

“The world at their feet” professional football and international labor migration. J Sport Social Issue 2002; 26: 421–437.

23.

Maguire

Sport worlds: a sociological perspective. Champaign, IL: Human Kinetics, 2002.

24.

McGovern

Globalization or internationalization? Foreign footballers in the English League, 1946-95. Sociology 2016; 36: 23–42.

25.

Bromberger

Foreign footballers, cultural dreams and community identity in some north-western Mediterranean cities. In: J

Bale

Maguire

(eds) The global sport arena: athletic talent migration in an independent world. London, UK: Frank Cass, 2013, pp.171–183.

26.

Adcroft

Teckman

Theories, concepts and the Rugby World Cup: using management to understand sport. Manag Decis 2008; 46: 600–625.

27.

Taylor

Global players? Football, migration and globalization, c. 1930-2000. Histor Social Res/Historische Sozialforsch 2006; 31: 7–30.

28.

Velema

A game of snakes and ladders: player migratory trajectories in the global football labor market. Int Rev Sociol Sport 2018; 53: 706–725.

29.

Zhou

Zhang

Calvo

, et al. Chinese soccer association super league, 2012–2017: key performance indicators in balance games. Int J Perform Anal Sport 2018; 18: 645–656.

30.

Vamplew

Creating the English Premier Football League: a brief economic history with Some possible lessons for Asian Soccer. Int J History Sport 2017; 34: 1807–1818.

31.

Wäsche

Dickson

Woll

, et al. Social network analysis in sport research: an emerging paradigm. Eur J Sport Soc 2017; 14: 138–165.

32.

Clemente

FM.

Performance outcomes and their associations with network measures during FIFA World Cup 2018. Int J Perform Anal Sport 2018; 18: 1010–1023.

33.

Hopkins

Magnitude-based decisions. Sportscience, http://sportsci.org/2019/inbrief.htm#decisions (accessed 19 June 2019).

34.

Batterham

Hopkins

WG.

Making meaningful inferences about magnitudes. Int J Sport Physiol Perform 2006; 1: 50–57.

35.

Hopkins

Marshall

Batterham

, et al. Progressive statistics for studies in sports medicine and exercise science. Med Sci Sport Exer 2009; 41: 3–12.

36.

Pena

Touchette

A network theory analysis of football strategies. arXiv Prepr 2012; arXiv: 1206.6904.

37.

Freeman

LC.

Centrality in social networks conceptual clarification. Social Netw 1978; 1: 215–239.