On the relationship between passing-related features and the result of offensive actions in football

Alongside the usual match statistics, the information content of passing net- works is increasingly used to analyse performance and to model and/or predict football results. Unfortunately, the indicators produced by such networks, ranging from more conventional ones (e.g., the number of passes made and received) to slightly more complex ones (such as measures based on centrality or density), suffer from a high level of aggregation and a lack of spatial or temporal details. A passing network can usually be generated considering the time intervals in a match: the whole match, halves, or smaller time segments. In this short paper, passing networks are defined by football actions, which represent the statistical units. As a result, an entire match (or fractions of it) can be seen as the collection of passing networks. We collect network-based features generated by each offensive action to investigate the possible relationship between network metrics and the football outcome, which is defined as positive if the action leads to a shot. This approach is illustrated through a single case study based on a football match that occurred in the Italian first division (2015-2016 season). It can easily be extended to a wide range of matches considering national and international competitions.