Peer play ability may be a protective factor against childhood mental health difficulties but there is lack of empirical evidence to support this hypothesis. We conducted longitudinal structural equation modelling study over a population cohort (N = 1676) to examine the effect of age 3 peer play ability on children’s age 7 mental health outcomes (measured by the Strengths and Difficulties Questionnaire subscales). We modelled effects for the entire population and two sub-groups at high-risk for mental health problems based on age 3 temperament. Controlling for demographic variables, temperament, maternal distress, play with parents and number of siblings, better peer play ability at age 3 years predicted lower risk of problems on all 4 SDQ subscales at age 7 years for the general population. For the low-persistence subgroup, better peer play ability at age 3 predicted lower risk of age 7 hyperactivity, emotional and peer problems, whereas better peer play ability at age 3 predicted only lower risk of age 7 hyperactivity for the high-reactivity group. Taken together our results provide evidence that supports the hypothesis that early peer play ability may be a protective factor against later mental health difficulties. We conclude that further research aimed at establishing causation is worth pursuing.