Predicting FIFA World Cup 2018 key role and playing style features

Andrew Duncan, Jacopo Diquigiovanni, Ioannis Papastathopoulos, Konstantinos Zygalakis, Gian Campagnolo

Research output: Contribution to conferencePosterpeer-review

Abstract

Based on the analysis of f24 level Opta data from five hundred and thirty eight qualifier games, we developed a set of predictions regarding the distinctive playing styles at FIFA World Cup 2018. Positional features and player level features were combined with network science and machine learning algorithms to create a spatial model and investigate how passing network features such as closeness centrality (i.e. measuring A-B-A type of passes), betweenness centrality (that is number of A-B-C type of passes) and position of on-ball events determined playing styles.
We clustered teams based on playing styles and analysed the key role-level features that influence the cluster positions in our plot, then looking more specifically at what teams in each cluster could do to potentially close the gap with top teams. We finally analysed how role-level features change when teams from opposing clusters play each other. Our findings are presented to support decision-making when three days after this talk national team coaches will have to finalise a squad list for World Cup 2018.
Original languageEnglish
Publication statusPublished - 2018

Fingerprint Dive into the research topics of 'Predicting FIFA World Cup 2018 key role and playing style features'. Together they form a unique fingerprint.

Cite this