The science behind the game

Gummano was born in a physics research group at the University of Santiago de Compostela (Spain) and is developed by its spin-off MD.USE Innovations, S.L.. This is what sits underneath the score.

How it is measured

We use an open-source hand detector (MediaPipe Hand Landmarker) that locates 21 points of the hand in every frame and estimates their position in three dimensions. All computation happens on your phone; images are never sent to a server.

The score is the sum of the 3D distances between neighbouring fingertips, divided by the palm width (distance between the index and little-finger knuckles). The normalisation makes the result dimensionless and comparable across hands of different sizes; the flat-hand calibration at the start fixes the proportions of each hand and is used to detect a change of hand during the challenge.

Why 3D rather than 2D? A flat photo confuses “fingers far apart” with “fingers turned towards the camera”. Measuring in 3D, crossing one finger over another or bending the thumb backwards counts for what it really is. Our tests with a simulated kinematic hand show that a 2D metric underestimates alternating poses (fingers up/down) by up to 46 % — exactly the poses that win.

How much can you trust the camera

Finger joint mobility is measured in rheumatology and rehabilitation with a goniometer. Recent studies have validated that measurement using a phone camera and hand landmark detection, the same technique this game uses [1, 2]. Accuracy depends on light, angle and fingers not hiding each other; that is why the game warns when the hand is tilted or badly framed, and why repeating helps: the best of several attempts is more stable than a single one.

What it has to do with health

Finger hyperextension is part of the Beighton score, the standard scale for generalised joint hypermobility, which affects roughly 10–25 % of people depending on age, sex and ancestry [3]. Scoring high here is neither “better” nor “worse”: hypermobility can come with joint discomfort, and if there is pain it is worth discussing with a professional.

At the other end, losing hand mobility has known clinical associations. In long-standing diabetes, limited finger joint mobility was linked to a much higher risk of microvascular complications in a classic study of 309 patients [4].

Flexibility can be trained. In rheumatoid arthritis, a hand stretching and strengthening programme improved function in a trial with 490 patients [5]. In healthy hands, gentle regular stretching increases range of motion.

This game is not a medical device: it does not measure any clinical quantity, offers no diagnosis and does not replace a professional assessment. Your score depends on your anatomy, your technique and the camera.

Data for research

If you tick the consent box, the geometry of your final pose (21 points, no images) is stored pseudonymously together with your score, your country and, if you choose to give them, your age and sex. With it we build the statistics by age (“compare”) and improve the measurement. Details are in the privacy policy, and you can delete your data from Settings at any time.

References

Reliability and validity of using smartphone sensor and photography to measure hand and upper extremity joint range of motion: a systematic review. J Hand Ther, 2025. pubmed.ncbi.nlm.nih.gov/40581513/
Hand tracking for clinical applications: validation of the Google MediaPipe Hand (GMH) and the depth-enhanced GMH-D frameworks. arXiv:2308.01088, 2023. arxiv.org/abs/2308.01088
Beighton P, Solomon L, Soskolne CL. Articular mobility in an African population. Ann Rheum Dis 1973;32:413–8. doi.org/10.1136/ard.32.5.413
Rosenbloom AL, et al. Limited joint mobility in childhood diabetes mellitus indicates increased risk for microvascular disease. N Engl J Med 1981;305:191–4. doi.org/10.1056/NEJM198107233050403
Lamb SE, et al. Exercises to improve function of the rheumatoid hand (SARAH): a randomised controlled trial. Lancet 2015;385:421–9. doi.org/10.1016/S0140-6736(14)60998-3