Using Network Science to Analyze Football (Soccer) Passing Networks
Link: https://www.frontiersin.org/articles/10.3389/fpsyg.2018.01900/full
This opinion article dives into the intricacies of how networks are used to evaluate individual and team performance in soccer. As the article states, the organization of a team can be considered as “the result of the interaction of its players creating passing networks. The detail these passing networks can have is quite remarkable, as multiple factors are considered when analyzing these networks. These include if they are directed (links between players go in one direction, the weight of the links in the network(based on the number of passes between players), how the networks are spatially embedded (the position of the ball and players is extremely relevant for the performance of networks), among multiple other factors.
An interesting thing the article touches upon is how networks can be divided into topological scales. The microscale is the importance each player has in the network. The microscale for passing networks has numerous concepts that are related to chapters 2 and 3. Related concepts in the article are degree, closeness, and betweenness of nodes. In the case of passing networks, degree is the number of passes a player makes. Closeness measures the minimum number of steps (nodes/players) the ball has to go from one player to reach any other player on the team. Betweenness refers to how many times a given player is necessary for compelting the routes connecting any other two players on the team. The higher the betweenness of a player, the more often the ball goes through their feet and the more important they are to the team’s performance.
Other topological scales include the mesoscale and the macroscale. The mesoscale indicates patterns describing the interaction between three or four players that intercat frequently with each other. The frequency of certain passes between groups of players can be directly correlated to the success of a team and the identification of players with the highest betweenness. The macroscale is the scale that considers the network as a whole. The position of the center of the network for a whole team can be related to team performance in which the farther forward the centroid is, the better. However, the centroid has shown to move backwards when teams play as visitors, interestingly. Duch et al. (2010) have also developed a performance metric that focuses on the betweenness of players and how it correlates with their probability of winning a match. Other aspects of the macroscale include how a teams average degree (average number of passes) and a players individual degree have been used to evaluate performance.
It is important to note that passing networks are very dynamic. The complete identification and quantification of how certain variables used to describe these networks fully affect the game is still a problem to be dealt with. Because soccer is such a complex game, stochastic forces act on the game all the time. Data analysis has gotten to a point where, although most things can be measured statistically, these same things can’t be precisely predicted due to the random nature of the game. However, the level of randomness in a specific network can be identified and precisely measured. A step further that can be taken with the analysis of passing networks is how a soccer match, in essence, is the result of a competition between two teams. In other words, it is the result of the competition between two passing networks. To achieve a precise interpretation of the game using networks, a passing network for a team needs to be analyzed in conjunction with the passing network of the opponent it is facing. Using this, it is possible to draw informed conclusions about how a team adapts in response to what their opponent does as well as what topological organization leads to the best results.