While virtual reality has been widely used for gaming and entertainment purposes for decades, its potential for rehabilitation has only recently been recognized. Virtual reality (VR)-based rehabilitation has many advantages for patients (high motivation) and clinicians (customized and adaptive therapy with transparent real-time data collection), especially as the virtual reality technology is available today at diminished cost. Adoption of this therapy in the clinical setting, however, has been relatively slow due to lack of research on its best practices. Thus, we aim to design 3D VR games and practice schedules based on fundamental principles in motor control and learning from our research. The long-term goal is to establish a paradigm for efficient and effective VR-based therapy. The first therapeutic game is designed building on our recent finding that augmentation of perceived visual error improves control and retention of a virtual throwing task in healthy subjects. In order to test the efficacy of this and future therapeutic games, we are developing a VR-based physical therapy using the Microsoft Kinect for motion analysis and Mathworks MATLAB software for data collection, virtual environment generation, and statistical analysis. We will develop the system software to be customizable such that it can be used in a variety of settings. Easy-to-use software will enable therapists and researchers with limited programming knowledge to implement their own 3D VR therapeutic games.