Figure 2 - uploaded by Filipe Manuel Clemente
Content may be subject to copyright.
Source publication
The aim of this case study was to apply a set of network metrics in order to characterize the teammates' cooperation in a football team. Metrics were applied in three levels of analysis: i) micro (individual analysis); ii) meso (players' contribution for the team); and iii) macro (global inter-action of the team). One-single case study match was ob...
Contexts in source publication
Context 1
... (0.808). In the case of the outcomes from clustering coe cient was possible to observe the higher values on the goalkeeper (0.631), left mid elder (0.614), forward (0.604) and striker (0.619). After to consider the outcomes from meso analysis (the contribution of each player for the teammates’ cooperation), it is now possible to characterize the individual contribution of each player during the attacking plays. e ‘micro’ analysis can allow for an understanding of the most prominent players that generate or participate in the team’s o ensive play. ese prominent players will be called centroid players. In this case it is necessary to remember that the matrices were built based only on the frequency that one player participated in the o ensive plays. e results about the centroid players can be seen in the gure 5. In this case study the right defender (1.000), central defender (0.826), right mid elder (1.000) and the forward (0.783) can be considered the centroid players. ese centroid players were the ones that participated the most in the o ensive plays and can be considered the most prominent players for the building of the o ensive plays. e main performance analysis methods have been based on statistical data and discrete actions performed by players and teams (Hughes & Franks, 2005). Nevertheless, the notational analysis ( i.e. , based on discrete actions) cannot identify the main reasons and processes that justify the outcomes (Vilar, Araújo, Davids, & Bar-Yam, 2013). erefore, some alternatives have been suggested and discussed seeking for a whole understanding about the product and mainly the process (Clemente, Couceiro, Martins, Mendes, & Figueiredo, 2013). One of them is the network performed by the team players (Passos et al., 2011). Despite their importance as a graphical viewpoint for coaches and their sta , there are important metrics that can identify in an easier way some properties of the network. erefore, the aim of this study was to apply in the football context some network metrics used in Social Sciences in order to identify the characteristics of the teammates’ interactions during attacking plays. e rst outcome from network analysis was the graph developed based on the weighted matrix (Figure 2). Taken in account such graph it was possible to observe that the connectivity between teammates is not equal. e player 4 (central defender from left side) seems to be one of the most recruited defensive players. is can suggest that the o ensive plays can start many times with this player. erefore, this can be considered to be an important information for the opponent team. Using this interpretation, the opponent team’s strategy can be to avoid starting the o ensive play with player 4, forcing them to build the o ensive play with player 3 (central defender from right side) who could be the player with less possibility for this (considering his connections with the other players), increasing the possibilities to fail. Observing the entire network it is possible to identify that the teams have a tendency to build their o ensive plays using the defensive and mid eld players. e high connections are between the players 2, 3, 4, 5, 6, 7 and 8. e striker (player 11) has a reduced interaction, meaning that this team opts for an indirect o ensive play. Moreover, gure 2 allows to be identi- ed that the large connection is performed by players 2 and 8 (right defender and right mid elder). Also, it can be observed that player 10 has a strong connection with players 2, 6 and 8. us, for the opponent team’s analysis this can be used for a global strategic adjustment. Nevertheless, the network could not only be used for the opponent analysis, but also for the own team’s analysis. e coach can use this information to improve the higher connections between the teammates or simply to organize the training sessions based on this. us, the coach can predict that the opponent team will press player 4 not to start the o ensive plays or will ‘block’ the connection between player 2 and 8. Using these elements, it is possible to training alternative o ensive strategies, such as increasing the lower connections or simply changing the style of o ensive play. Also, these connections can be used to identify some ‘negative’ tendencies. Sometimes, players would keep a pattern. Nevertheless, this pattern cannot be pro table. us, using this information enables the possibility of undertaking some changes during the training sessions in order to increase the variability and reduce the ‘negative’ patterns. Besides the analysis performed based on the graph it was also analysed the network metrics computed. e density values are closer to 0.5, thus meaning that a partial ambiguous relationship in the players’ interaction exists. By Using this metric for a technological online approach enables the coaches or their sta to analyse the evolution of the players’ interaction. Values closer to 1 suggest a high value of ambigu- ity, thus suggesting that the team does not work as a whole. us, if all players do not have a similar participation in the team, this can be due to a speci c strategy or it could ...
Context 2
... cooperation level of a given player. In this case, the highest values (tends to 1) suggest that player participate with most of the other players from the team. e opposite (tend to 0) suggest the player had speci c preferences to participate with some players within the team. e clustering coe cient of a given player measure the degree of interconnectivity in the neighbourhood of the player, i.e., reveals if the player promotes a connectivity between their remaining teammates. Highest values of clustering coe cient suggest that the teammates of a given player cooperates with frequency between each other. Lowest values reveals that the teammates of a given player do not cooperates much each other. e centroid player can be de ned as the most recruited and connected node in the network, thus suggesting the centrally located node (Horvath, 2011). In this case, highest values reveals the higher connected node and the lowest values suggest the player recruited in a smallest way. After to develop the weighted matrix of the teammates’ interactions during the attacking plays was possible to gra- phically compute the following graph (Figure 2). Besides to the graph, were computed the network metrics based on the weighted matrix developed during the attacking plays. Such descriptive values will be showed throughout this section. Based on gure 2 it is possible to observe the connectivity between teammates is not equal or homogeneous. ere is some heterogeneity in the connections that can be explored. e goalkeeper (player 1) seems to interact more with the central defender (player 4). Moreover, player 4 seems to be a highly connected to players 2 (right defender), 3 (central defender), 5 (left defender) and 6 (defensive mid elder). us, player 4 showed to be one of the most recruited defensive players. Despite the network importance as a single graph, deeper interpretations can be performed with higher e ciency. us, a macro analysis was performed using three metrics. e results from such metrics can be following observed (Figure 3). From the macro analysis performed was possible to identify a density value of 0.512, a heterogeneity value of 0.263 and a centralization value of 0.177. Such results comes from the analysis performed to the weighted matrix of the one-single case study match. e macro classi cation only depends from the all level of connections within the graph not identifying the individual contribution of players. us, a meso analysis was performed in order to characterize how each player contributes to the teammates’ cooperation. e ‘meso’ analysis was undertaken using the scaled connectivity of each player and the clustering coe cient. e gure 4 shows the values for each player. Figure 4 shows that in this particular case the players with higher scaled connectivity values are the right defender (1.000), central defender from the left side (0.993), defensive mid elder (0.980), right mid elder (0.887) and the ...
Citations
... One of the notable features of the available literature is a wide array of research methods. In this regard, some papers have very small samples of as little as one match or team, [52][53][54] which substantially limits generalisation and application of the findings, whilst others include datasets as large as 1941 matches. 31 Additionally, there is substantial variety in the methods of including players in the formation of networks. ...
Team invasion games are sports in which individual team members interact and exchange information to coordinate their behaviours and actions in pursuit of the common goal of winning matches. Researchers have used social network analysis to quantify the cooperative behaviours of sports teams (cooperative network analysis), yet this research exists across an array of disciplines and uses various methods. Therefore, accessibility for practitioners and researchers interested in using it to quantify team cooperation in team invasion games may be limited. This systematic mapping review aimed to identify, report and discuss research in this emerging research area. Articles were systematically searched in electronic databases and reference list scans resulting in 112 papers included. Football was the most studied sport ( n = 91), and passing was the most observed interaction between players within a sports team ( n = 107). This review further revealed a lack of consistency in reporting between the included studies with respect to nomenclature and network measures. A comprehensive map of the current literature on the use of cooperative network analysis in team invasion games is provided which can be used by practitioners and researchers tasked with or interested in analysing team performance.
... This research concept of compatibility is pragmatic to the working of social groups as well. Despite individual characteristics, players in team sports need to work in coordination [5,6] so that the aggregate performance improves. With team sports like football, where formation transition, counter-attacking press and different decisions are taken within a fraction of seconds, good coordination between players is very cardinal. ...
The popularity of football among fans to analyze the game has been immense with the advent of internet. The concept of making a dream team in football has become a new fashion for the football lovers. The paper focuses in helping achieving this prediction of a football dream team. The aim of this research is to assess the dynamics of a complex topological structure when prompted with random entities whose attributes are known to us. Using graph theory and vectorial distances, the dream team is evaluated on the basis of individual abilities and interplayer synergy. Instead of focusing on discrete events in a match, this framework proposes an idea in which a dream team is quantified on the basis of their positional attributes. Each player is rated in accordance to the position he is playing, which eventually helps in finding the overall team rating. The second part of this research uses graph theory to evaluate structural and topological properties of interpersonal interactions of teammates. Teammates are treated as nodes of a graph, where each edge exemplifies the strength of their interpersonal interaction. The strength of the bond depends on on-field interactions via ball passing, ball receiving and communication which depend on experience of playing together, Nationality and Club. The methodology adopted in this paper can be a formidable basis for similarly situated larger setups involving much larger intricacies. Using this framework, we can see the behavior of a hypothetical topological structure whose node attributes are known to us, thus projecting its performance as a team and individual entities.
... Analysis based on passing networks mainly deals with the data on the relationship between players, emphasizing the sports social network structure, driven by relational quantification occurring among them. e representative measure is centrality in the passing networks [32][33][34][35][36]. e analysis based on passing networks has undergone major shifts, from degree centrality [33] to flow centrality [37], from unweighed measures [38] to weighted measures [37,39], and from homogeneous passing networks [40] to heterogeneous passing networks [32]. ...
... Analysis based on passing networks mainly deals with the data on the relationship between players, emphasizing the sports social network structure, driven by relational quantification occurring among them. e representative measure is centrality in the passing networks [32][33][34][35][36]. e analysis based on passing networks has undergone major shifts, from degree centrality [33] to flow centrality [37], from unweighed measures [38] to weighted measures [37,39], and from homogeneous passing networks [40] to heterogeneous passing networks [32]. ...
... In the past decade, sports social network analysis has undergone major shifts, from degree centrality [36] to eigenvector centrality [40,41], from unweighed centrality measures [12,42] to weighted centrality measures [37,39]. In team sports, centrality is mainly used in the following aspects: (1) Identifying prominent players [33,43]. (2) Determining the key pitch zones for matches [29]. ...
With the rapid growth of information technology and sports, a large amount of sports social network data has emerged. Sports social network data contains rich entity information about athletes, coaches, sports teams, football, basketball, and other sports. Understanding the interaction among these entities is meaningful and challenging. To this end, we first introduce the background of sports social networks. Secondly, we review and categorize the recent research efforts in sports social networks and sports social network analysis based on passing networks, from the centrality and its variants to entropy, and several other metrics. Thirdly, we present and compare different sports social network models that have been used for sports social network analysis, modeling, and prediction. Finally, we present promising research directions in the rapidly growing field, including mining the genes of sports team success with multiview learning, evaluating the impact of sports team collaboration with motif-based graph networks, finding the best collaborative partners in a sports team with attention-aware graph networks, and finding the rising star for a sports team with attribute-based convolutional neural networks. This paper aims to provide the researchers with a broader understanding of the sports social networks, especially valuable as a concise introduction for budding researchers interested in this field.
... Passos et al. (2011) revealed that the number of network interactions between team members should differ between successful and unsuccessful performance outcomes. Clemente et al. (2014) applied a set of network metrics to characterize the co-operation between teammates in a football team. ...
In basketball, measures of individual player performance provide critical guidance for a broad spectrum of decisions related to training and game strategy. However, most studies on this topic focus on performance level measurement, neglecting other important factors, such as performance variability. Here we model shooting performance variability by using Markov switching models, assuming the existence of two alternating performance regimes related to the positive or negative synergies that specific combinations of players may create on the court. The main goal of this analysis is to investigate the relationships between each player's performance variability and team line‐up composition by assuming shot‐varying transition probabilities between regimes. Relationships between pairs of players are then visualized in a network graph, highlighting positive and negative interactions between teammates. On the basis of these interactions, we build a score for the line‐ups, which we show correlates with the line‐up's shooting performance. This confirms that interactions between teammates detected by the Markov switching model directly affect team performance, which is information that would be enormously useful to coaches when deciding which players should play together.
Read-only version available at
https://onlinelibrary.wiley.com/share/author/NH3VF5WSADCBHWZX6TMN?target=10.1111/rssc.12442
... Among the methods proposed to deal with teamwork assessment, a noteworthy approach is given by networks methods. Warner et al. (2012) and Lusher et al. (2010) used social network analysis to investigate how team cohesion and individual relationships impact team dynamics, Passos et al. (2011) revealed that the number of network interactions between team members should be able to differentiate between successful and unsuccessful performance outcomes, Clemente et al. (2014) applied a set of network metrics in order to characterize the teammates cooperation in a football team by considering individual analysis, players contribution for the team and global interaction of the team. ...
Basketball players' performance measurement is of critical importance for a broad spectrum of decisions related to training and game strategy. Despite this recognized central role, the main part of the studies on this topic focus on performance level measurement, neglecting other important characteristics, such as variability. In this paper, shooting performance variability is modeled with a Markov Switching dynamic, assuming the existence of two alternating performance regimes. Then, the relationships between each player's variability and the lineup composition is modeled as an ARIMA process with covariates and described with network analysis tools, in order to extrapolate positive and negative interactions between teammates
... Combining network-based approaches with spatial information has been identified as the next step forward in performance analysis, and is crucial in considering the interactions between players during a match (Clemente, Martins, Couceiro, Mendes, & Figueiredo, 2014;Fewell, Armbruster, Ingraham, Petersen, & Waters, 2012;Gyarmati & Anguera, 2015). The Apriori algorithm is a classical data mining technique used to find co-occurring items in large datasets, and can also be used to detect spatiotemporal trends (Agrawal & Srikant, 1994;Morgan, 2011;Spencer, Morgan, Zeleznikow, & Robertson, 2016). ...
In netball, analysis of the movement of players and the ball across different court locations can provide information about trends otherwise hidden. This study aimed to develop a method to discover latent passing patterns in women’s netball. Data for both pass location and playing position were collected from centre passes during selected games in the 2016 Trans-Tasman Netball Championship season and 2017 Australian National Netball League. A motif analysis was used to characterise passing-sequence observations. This revealed that the most frequent, sequential passing style from a centre pass was the “ABCD” motif in an alphabetical system, or in a positional system “Centre–Goal Attack–Wing Attack–Goal Shooter” and rarely was the ball passed back to the player it was received from. An association rule mining was used to identify frequent ball movement sequences from a centre pass play. The most confident rule flowed down the right-hand side of the court, however seven of the ten most confident rules demonstrated a preference for ball movement down the left-hand side of the court. These results can offer objective insight into passing sequences, and potentially inform team strategy and tactics. This method can also be generalised to other invasion sports.
... Sequences of passes between players can be represented as a network with players as the vertices and weighted edges for the frequency of passes between pairs of players and so to quantify the passing performance (Clemente, Martins, Wong, Kalamaras, & Mendes, 2015c;Grund, 2012). The network process can quantify different levels of analysis, from player-team (micro) such as degree centrality, degree prestige, betweenness centrality and closeness centrality (Clemente et al., 2016b(Clemente et al., , 2016aClemente, Silva, Martins, Kalamaras, & Mendes, 2016c); dependence between players (meso) such as clustering coefficient or scaled connectivity (Clemente et al., 2015b;Gama, Couceiro, Dias, & Vas, 2015;Peña & Touchette, 2012); and, the general properties of a team (macro) such as density, heterogeneity or network diameter (Clemente, Couceiro, Martins, & Mendes, 2015a;Clemente, Martins, Couceiro, Mendes, & Figueiredo, 2014), allowing to characterize the teammates' interaction during the offensive phase, providing specific information about passes connections that can be useful for match analysis . ...
This study’s main objective is to analyse the relationship between network-based centrality measures and physical demands in elite football players. Thirty-six matches from La Liga, the Spanish league, were analysed in the 2017/18 season. The analysis of networks formed by team players passing the ball included: degree-prestige (DP), degree-centrality (DC), betweenness-centrality (BC), page-rank (PRP) and closeness-centrality (IRCC). A video-based system was used for analysing total distance (TDpos) and distance run >21Km/h (TD21pos) when the team was in possession of the ball. A magnitude-based inference and correlation analysis were applied. There were different styles of play, team-A was characterized by greater ball circulation (e.g. higher values of DP, DC, BC and IRCC) while team-B used a more direct game (lower values in centrality-metrics except with PRP). Furthermore, TDpos was higher in team-A than in team-B, but those differences disappeared for TD21pos between teams with the exception of the forwards. Finally, the correlation among centrality measures and physical performance were higher in team-B. Coaches could identify the key opponents and players who are linked to them, allowing to adjust performance strategies. Furthermore, interaction patterns between teammates can be used to identify preferential paths of cooperation and to take decisions regarding these relations in order to optimize team performance.
... The high creativity of the last two previous actions to the shots at goal seems to be the fundamental factor that allows us to explain the effectiveness against the opponent's goal. As a general rule, in a well-organized team, it is the central midfielders (especially the playmakers, or creative midfielders) that manage to establish mutually activating relationships, that is to say in both planes, or in the prospective plane with the wingers and forwards (Robles et al., 2014;Clemente et al, 2014). In our case, this is also reflected in the analysis of the interactions (type and quantity) in which the centre forward participates. ...
The interactions of a Spanish football team of the Second A (Liga 123) (10 official games) are analyzed, evaluating possible behavioral patterns that appear in a regular way in high level football. Observational methodology was used, by Polar Coordinates Analysis, to discover and evaluate the relationships generated between a considered focal behavior and the different conditioned categories, describing behavioral masses among the players. The matches were observed and recorded with an ad hoc observation instrument. The relations of dual character between the players taken as (focal behaviors) right midfielder and forward and the other players (conditioned conducts) were analyzed. The results showed differences in the relationship established based on the outcome of the match.
... The high creativity of the last two previous actions to the shots at goal seems to be the fundamental factor that allows us to explain the effectiveness against the opponent's goal. As a general rule, in a well-organized team, it is the central midfielders (especially the playmakers, or creative midfielders) that manage to establish mutually activating relationships, that is to say in both planes, or in the prospective plane with the wingers and forwards (Robles et al., 2014;Clemente et al, 2014). In our case, this is also reflected in the analysis of the interactions (type and quantity) in which the centre forward participates. ...
... From the brief review of the literature, we make the following conclusions: a) Most studies manually create the passing matrices and analyze a limited amount of games. For instance, [3,4] analyze a single match, while [5] analyze four and [6] five matches. ...
... Density: This measure gives an impression about the level of affection between teammates [3]. Higher values mean that all players interact with each other, while lower values mean a more ambiguous relationship [3]. ...
... Density: This measure gives an impression about the level of affection between teammates [3]. Higher values mean that all players interact with each other, while lower values mean a more ambiguous relationship [3]. Although the values are not too far apart, Portugal has a higher score by approximately point one. ...
Social networks have been applied in football, or football match analysis to analyze the passing distributions between teams. However, analysis has been mostly done on a manually collected data by considering the widely adopted network metrics such as betweenness and closeness centrality. In this paper, we use professional tracking event data provided by OPTA Sports and analyze the final game of the Euro2016 between Portugal and France. We use Gephi and the NetworkX Python library and apply dynamic network analysis by integrating the timestamps of the passes. We further look into traditional performance metrics from both teams and make an attempt to connect those to the network results and the outcome of the game.
