The game beyond the field: on football players’ performance through social media, sentiment and topic analysis

Ortu, Marco

;Mola, Francesco

2024-01-01

---

Abstract

This study investigates the complex relationship between social media sentiment and football players’ performance in the English Premier League (EPL). We adapt the TOpic modeling Based Index Assessment through Sentiment (TOBIAS) framework, originally developed for educational settings, to the domain of sports analytics. This novel application faces difculties in handling the volume and variability of social media data, as well as in accurately linking pre-match sentiments to post-match performance metrics. Our methodology integrates advanced Natural Language Processing (NLP) techniques, including sentiment analysis and topic modeling, with Partial Least Squares Path Modeling (PLS-PM). We analyze a dataset of 167,841 tweets related to 512 English Premier League (EPL) players, collected from May 2022 to May 2023. The study is conducted in two phases: pre-match analysis to assess public expectations, and post-match analysis to evaluate reactions to player performances. Experimental analysis reveals signifcant correlations between pre-match sentiments and subsequent player performance, with negative sentiments showing a stronger predictive power than positive ones. Post-match, we observe a shift in the relationship between sentiments and performance metrics, indicating the public’s responsiveness to match outcomes. Our fndings contribute to the broader understanding of social media’s role in sports performance and ofer insights for potential applications in regulating online behaviors in sports contexts.