Though extreme events, riots are key signals of well-being in societies. Without understanding the psychological mecha-nisms behind them, however, it is difficult to discern the social changes that need to be made in order to reduce both theriots themselves and their underlying causes. In this work we use computational models to test both relative deprivationand social identity as explanations for data from a novel experimental framework, Parklife, which provides data on howand when individuals riot in the laboratory. Our models show that whilst norm formation and distinctiveness are importantfactors in explaining the behaviour of participants in Parklife, relative deprivation is a key and necessary mechanism inthe increase in anti-social behaviour observed in disadvantaged groups. This work offers the first direct test of relativedeprivation within a specialised system, and shows the power of computational simulations in connecting theories withdata, helping us to test hypotheses.