Comparison of bow-side kinematics in violinists is hindered by the scarcity of studies available. This makes meta-analysis impossible. This paper assesses the effect of music-based variables (bow stroke, tempo, and string played) on intra- and inter-participant variability in joint kinematics. The joint kinematics of nine high-level violinists were acquired via a motion capture system while they played a standardized piece of music involving contrasting bow strokes and strings at different tempi. Results were compared using linear mixed models using the root mean square (RMS) for each joint. We found highly individualized patterns of play, deduced from a low intra- but high inter-musician variability (4.2° vs 13.1° of normalized RMS) in joint kinematics. String played and bow stroke had the greatest effect on joint kinematics. The string played had the greatest impact on shoulder kinematics, and the bow stroke had the greatest impact on elbow and wrist kinematics. Based on these results, we propose guidelines for future research designed to study bow kinematics in the field of biomechanics of violin movements. For ease of comparison between studies and to limit the time and resources required, our main suggestions are to use repeated measures designs with a legato reference condition and to choose pieces of music spanning multiple strings.