Describing Italian Football Teams' Transfer Dynamics in a Network Perspective
Roberto Rondinelli, Lucio Palazzo, Giancarlo RagoziniTransfers between football teams involve usually complicated negotiations with four parties pushing to get the best cut of the deal on their end: the selling club, the buying club, the player involved and his agent. As all trade relationships, even those of football market can be examined in a relational data framework, for this reason, classical statistical models are not well suited for analysing them.
Relational data, in their simplest formulation, are organised in two sets of elements: actors, also named vertices or nodes representing the selected units, and relations, also named edges or links, indicating ties between actors. The presence of a relation between actors is represented by a line connecting two nodes, that can be directed or undirected, weighted or not weighted.
The involved parties can be seen as nodes and their relationships (commercial exchanges) as links in the network; furthermore, based on the kind of negotiation, parties could represent attributes of relations (e.g. football agent is an attribute of the trade between two teams). In this context, according to the existing literature research, we apply a network analysis modelling to study this kind of data.
Making use of data from the Transfermarkt portal, we focus on Italian football teams trade during the 2019 summer transfer window. Specifically, from the databases contained in this portal, we extract network information and the main features (attributes) regarding the nodes and their links, such as the main features of teams, players and transaction contract terms.
Focusing on team-based and football agent-based networks, in order to study the many facets occurring during negotiations between these parties, the project is divided in different steps. Firstly, we provide an overall description of network to highlight the behavior of the entire market and the importance that parties recover in it. Secondly, we explore the community structure of the network; here the aim is threefold: we first analyse the internal anatomy of each obtained community, on the other side we investigate their connections and finally we do the same with prefixed communities. Thirdly, we apply the Exponential Random Graph Model to have a more analytical proof of what we found in the descriptive analysis. Besides the main structure organization of the network, we have evidence of the important attributes that are involved in the network formation process of Italian football teams market.