Correlation Between NBA Free-throw Shooting Accuracy And Twitter Sentiment Scores

A body of research has tested if private “self-talk” impacts sport performance, but whether athletes’ public communication relates to sport performance is less researched. Two prior studies found small relationships between overall NBA game performances across a season and positive sentiment expressed by text and emojis in pre-game tweets (Xu et al., 2015a & b). A key limitation of these studies was the potential bias in the performance metrics (e.g. , positive sentiment may be more frequent prior to playing easier opponents). PURPOSE: To quantify relationships between pre-game tweet sentiment expressed in NBA players’ twitter accounts and their free-throw shooting accuracy, which is a closed motor skill performance uninfluenced by a defender or the quality of the opposing team. METHODS: To avoid an imprecise measure of performance, players attempting ≥ 6 free-throws in one or more game during the 2021/22 regular season were identified; from this initial group only those who also sent one tweet 0.75- to 26-hours prior to the target game were included in the analysis. For those meeting the criteria above in more than one game or those who had multiple pre-game tweets, one game and one associated tweet were randomly selected. Game free-throw accuracy (G-FT%) was obtained from StatMuse and player efficiency rating, one index of season performance, was obtained from basketballreference.com and used as a covariate. Tweet text sentiment scores were measured using the Valence Aware Dictionary and sEntiment Reasoner (VADER). Positive scores reflect positive sentiment in tweets while negative scores reflect negative sentiment. Emoji sentiment scores were generated independently by two researchers; scores were 0 (no emoji), 1 (negative), 2 (slightly negative), 3 (neutral), 4 (slightly positive), and (5) very positive. Criterion total sentiment scores were the sum of VADER and emoji scores. RESULTS: G-FT% was significantly correlated with total sentiment scores (rho = -0.27; 95% CI: -0.460, -0.014; P=0.033; n = 88). This relationship was unchanged after controlling for player efficiency rating (rho = -0.29). CONCLUSIONS: NBA free-throw shooting accuracy tended to be higher when pre-game tweets sent from players’ accounts contained more negative sentiment based on a measure that combined sentiment from both text and emojis.