Falls are the leading cause of disability, injuries or even death among older adults. Exercise programmes that include a balance component reduce the risk of falling by 40%. However, such interventions are often perceived as boring and drop-out rates are high. The characteristics of videogames may overcome this weakness and increase exercise adherence. The use of modern input devices, such as the Microsoft Kinect, enables quantification of player performance in terms of motor function while engaging with games. This capability has just started to be explored. The work presented in this paper focuses on the development of a Kinect-based system to deliver step training while simultaneously measuring parameters of stepping performance that have shown to predict falls in older people.