This paper uses data from expert chess games to study the underperformance and underrepresentation of women in competitive environments. Chess is an ideal testbed because it is a sport where women and men compete together and where an objective measure of ability exists – the Elo rating. We show that, everything equal, women perform worse than men. We then measure quality of play by computing the difference between actual play and the best possible movement. We find that women players’ quality of play decreases when playing against the opposite sex. This effect is stronger in weaker female players and absent in male players. We interpret these results to be in line with the phenomenon of stereotype threat.