Purpose This study examines emotional expression and engagement on social media during a collegiate football rivalry game. We ask (1) how discrete emotions change across the event when discourse is separated by team stance and (2) how discrete emotions relate to different engagement modes on X. Design/methodology/approach Using the X Search API, we collected posts from the 2023 UF vs. Florida State University football game (FSU 24, UF 15) with balanced team search terms (n = 21,913). We inferred stance using cue-based rules that capture explicit support and opponent-directed attacks (Pro-UF, Pro-FSU, Neutral/Unclear). Emotions were measured with a RoBERTa-based classifier trained on the GoEmotions dataset. We aggregated probabilities into 30-min bins by stance and linked dominant emotions to engagement mode. Findings Emotion trends differed by stance, showing that overall discourse reflects a mix of affiliation groups. Each stance group exhibited distinct emotion trajectories across the game window, with distinct profiles in both magnitude and timing as game events unfolded. Emotions within the same valence diverged, and engagement patterns were mode-specific. Admiration and excitement drew likes and retweets; curiosity, confusion and disapproval drew replies; and disappointment appeared broadly across replies, quotes, and views. Originality/value The study advances sport social media research by integrating stance detection with multi-emotion classification and by linking these measures to mode-specific engagement outcomes. This approach demonstrates that binary measures and stance-blind analyses can mask meaningful complexity in fan behavior, while offering clearer guidance for sport social media research.
Han et al. (Wed,) studied this question.