Time-Aware Ranking in Sport Social Networks

Abstract

Sport social networks concern different types of relationship among athletes or teams in specific sports. Such networks have recently been used to address problems related to prediction of results of matches or championships and rankings of athletes or teams. In many cases, such analyses consider a complete and static view of the network that does not take into account the temporal nature of sports events. In this article, we present a time-aware ranking method for sport social networks that explicitly considers these temporal factors. In particular, we propose modeling such networks with edge weights that decay over time, in order to represent the relative importance of past interactions. We apply the proposed method to two sports, namely Mixed Martial Arts (MMA) and tennis. Our results show that our rankings are more accurate than a baseline ranking that ignores temporal factors, when both are compared to gold standards derived from well known rankings.