Technological and medical advances are leading to great improvements in overall quality of life and life expectancy. However, these positive developments are accompanied by considerable challenges. The modern sedentary lifestyle and common afflictions that become more prevalent with age are contributing to considerable burdens on health care systems and on a great number of individuals. In addition to specific primary treatments, physical activity plays a major role both in prevention and in the treatment of such afflictions, for example though the application area of physiotherapy. Games for health (GFH) in general and motion-based games for health (MGH) in particular are being discussed in research and industry for their ability to play a supportive role in health, by offering (a) motivation to engage in treatments, (b) objective insights on the status and development of individuals or groups based on data collection and analysis, and © guidance regarding treatment activities, which is especially promising when health professionals are not available in person. However, applications in health need to be tailored to the individual needs and abilities of patients in order to facilitate the best possible outcomes. While most games can be adjusted to a general level of player abilities, this is typically achieved with a single difficulty setting with a limited number of discrete tiers, such as “easy”, “medium”, and “hard”. For most serious application use cases in health, more fine-grained and far-reaching adjustments are required. This can quickly lead to a need for applying adjustments on complex sets of parameters, which can be overwhelming for patient-players and even trained profession-als. Automatic adaptivity and efficient manual adaptability are thus major concerns for the design and development of GFH and MGH. Despite a growing amount of research on specific methods for adaptivity, general considerations on human-computer interaction with adaptable and adaptive MGH are rare and scattered across reports from specific developments. Based on a thorough consideration of the existing background and related work, this thesis therefore focuses on establishing and augmenting theory for adaptability and adaptivity in human-computer interaction in the context of MGH. The considerations are supported by a series of studies and practical developments. Working with older adults and people with Parkinson’s disease as frequent target groups that can arguably benefit from tailored activities, explorations and comparative studies that investigate the design, acceptance, and effectiveness of MGH are presented. The outcomes encourage the application of adaptivity for MGH following iterative human-centered design that considers the respective interests of the complex collage of involved parties and stakeholders, provided that the users receive adequate information and are empowered to exert control over the automated system when desired or required, and if adaptivity is embedded in such a way that it does not interfere with the users’ sense of competence or autonomy.