The number of virtual reality applications in rehabilitation and telerehabilitation medicine is increasing. Unfortunately, without telediagnostics, telerehabilitation requires frequent outpatient clinical testing. This study thus proposes a diagnostic tool for virtual-reality-supported balance training for estimating functional balance progress during telerehabilitation using objective parameters. Four weeks of physiotherapy with assessment before and after the therapy and a follow up 2 weeks after therapy were carried out with 10 sub-acute stroke subjects. Goal-based tasks for standing-frame-supported balance training were designed in a virtual environment. An accelerometer-based tilt sensor not mounted on the subjects was used. The telediagnostics apply the fast Fourier transform (FFT) to movement data and the task time and number of errors committed are analyzed. Additionally the functional progress was estimated with clinical tests (Berg Balance Scale, 10-m walk test, Timed Up & Go test) and correlations with objective data were examined. The area under the FFT curve demonstrated balance improvement and had a moderate correlation with clinical tests, considering that all parameters showed improvement for each individual patient. The proposed tool enables the remote evaluation of the effects of virtual-reality-supported balance training. The tool may decrease the number of outpatient visits and enable the continuation of the rehabilitation process at home. However, to confirm the clinical reliability and practical value of the proposed tool, a further study with a large stroke population and healthcare business models are needed.