Objectives: To determine whether 12-week home-based exergame step training can improve stepping performance, gait and complementary physical and neuropsychological measures associated with falls in Parkinson’s disease. Design: A single-blinded randomised controlled trial. Setting: Community (experimental intervention), university laboratory (outcome measures). Subjects: Sixty community-dwelling people with Parkinson’s disease. Interventions: Home-based step training using videogame technology. Main measures: The primary outcomes were the choice stepping reaction time test and Functional Gait Assessment. Secondary outcomes included physical and neuropsychological measures associated with falls in Parkinson’s disease, number of falls over six months and self-reported mobility and balance. Results: Post intervention, there were no differences between the intervention (n = 28) and control (n = 25) groups in the primary or secondary outcomes except for the Timed Up and Go test, where there was a significant difference in favour of the control group (P = 0.02). Intervention participants reported mobility improvement, whereas control participants reported mobility deterioration—between-group difference on an 11-point scale = 0.9 (95% confidence interval: −1.8 to −0.1, P = 0.03). Interaction effects between intervention and disease severity on physical function measures were observed (P = 0.01 to P = 0.08) with seemingly positive effects for the low-severity group and potentially negative effects for the high-severity group. Conclusion: Overall, home-based exergame step training was not effective in improving the outcomes assessed. However, the improved physical function in the lower disease severity intervention participants as well as the self-reported improved mobility in the intervention group suggest home-based exergame step training may have benefits for some people with Parkinson’s disease.