Are people prone to selecting occupations with highly skewed income distributions, irrespective of the mean and variance of their expected earnings? This paper introduces a basic theoretical framework to understand labor supply decisions in winner-take-all markets. Assembling a comprehensive longitudinal dataset of potential tennis professionals (a typical superstar market), we use objective and publicly known rankings to construct earnings projections before each player decides whether to enter that labor market. Player- and cohort-age fixed effects account for unobservable factors at the individual level and within a given cohort and year. We find prospective tennis professionals are attracted to highly skewed earnings distributions, independent of mean and variance. Hypothetically, if skewness in prize money fell to the level of the overall US labor market, males would be 7.2% and females 2.6% less likely to continue pursuing a professional tennis career, on average. Superstar labor markets may therefore systematically encourage those with modest talents to pursue long-shot careers.