The impact of a stroke and the individual neuronal regeneration capacity varies highly from patient to patient. In this context, a consistent scoring method is very desirable to quantify the motion impairment and to evaluate the result of rehabilitation. Traditional internationally recognized clinical scores (e.g. Barthel Index, Berg Balance Scale, etc.) are widely used but fail to assess the functional mobility or to assist clinicians in the planning of personalized therapy. For this reason, a new score (the ReHabX score) has been developed, which evaluates 6 functional gait motion criteria, namely “trunk movement”, “arm movement”, “leg movement”, “gait speed”, “gait fluency”, and “stability/risk of falling”. The gait of 59 stroke patients was measured using motion capture technology, and evaluated by a team of medical experts using the ReHabX score. 168 parameters (so-called “features”) were identified and extracted from each measurement. This list of features included conventional gait metrics, discretized joint angles, and non-traditional gait parameters designed to represent specific pathological gait characteristics. Due to the good reliability and normal distribution of the scorings, a linear model was chosen for the reconstruction of the clinicians’ ratings from the features. By a stepwise multilinear regression algorithm, features which allowed the reproduction of the functional scores provided by the medical experts were selected. For each scoring criteria, 3 to 6 features were inserted in the corresponding linear model. The accuracy of the numerical models was evaluated using both the traditional measures of the goodness of fit, and a novel technique. In this context, a novel so-called “bubble plot” was developed, which allows the visualization of the level of agreement among the raters together with the comparison of the numerical scoring with the average medical score. It was thereby shown that the determination of the individual criteria using the gait features accurately corresponds to the reliability of the medical experts. The numerical scores obtained are a valuable contribution to the diagnostics of motion impairment that ensure objectivity and reproducibility. They may support the creation of a personalized therapy and speed-up the post-stroke recovery process.