Animal displays (i.e. movement-based signals) often involve extreme behaviours that seem to push signallers to the limits of their abilities. If motor constraints limit display performance, signal evolution will be constrained, and displays can function as honest signals of quality. Existing approaches for measuring constraint, however, require multiple kinds of behavioural data. A method that requires only one kind could open up new research directions. We propose a conceptual model of performance under constraint, which predicts that the distribution of constrained performance will skew away from the constraint. We tested this prediction with sports data, because we know a priori that athletic performance is constrained and that athletes attempt to maximize performance. Performance consistently skewed in the predicted direction in a variety of sports. We then used statistical models based on the skew normal distribution to estimate the constraints on athletes and displaying animals while controlling for potential confounds and clustered data. We concluded that motor constraints tend to generate skewed behaviour and that skew normal models are useful tools to estimate constraints from a single axis of behavioural data. This study expands the toolkit for identifying, characterizing, and comparing performance constraints for applications in animal behaviour, physiology and sports.