The Ball Is the Result. The Body Is the Cause.

By Marcell Balogh, CTO / Lead Data Scientist & ML Engineer, Dinkmate.ai

Everyone building sports tech right now seems focused on the same thing: where did the ball go. Placement, speed, spin, whether it landed in or out. All useful. All, we'd argue, downstream of the thing that actually matters.

A ball can travel from point A to point B and land well or land badly, but the ball doesn't decide that. Your body does. The stance you set up in, the angle of your knees, how far your feet are spread, how your hips and shoulders rotate through contact — that's the actual cause. The ball is just the readout.

Why we started here

It's a little strange to us that none of the competitors out there focus much on body motion. At Dinkmate, this has been core to how we think about the sport since day one — most of our features are built on top of human motion data, not on top of the ball. That commitment shows up in two decisions we made early: we track the whole motion, not just the moment of contact, and we track it in 3D, not 2D.

The whole motion, not one frame

A stroke isn't a single frame; it's a motion that starts before contact and doesn't finish until the body has settled again. So we capture joint angles, leg spread, and feet distance not just at contact, but at the end of the shot too — the follow-through and the recovery stance. A clean contact point followed by a scrambled recovery tells you something very different than the same contact followed by a controlled reset, and only one of those shows up if all you capture is a single moment.

3D, not 2D from a locked angle

The second decision is just as important. A 2D pose estimate from a fixed camera angle can only tell you what's visible from that one point of view — and it flattens depth into guesswork. A knee that looks bent at a certain angle from one camera can look completely different from another. Turn your hips slightly away from the lens and a 2D system loses the very information you actually need. Real skeleton kinematics — actual joint angles, in actual degrees, relative to the ground and to each other — only exist in 3D. Anything less is an approximation dressed up to look like a measurement.

Why this matters for coaching

Ball data tells you the outcome. Body data tells you why. Two players can hit the same shot with wildly different footwork and joint angles — one repeatable, the other one bad step from an injury or a mishit. A coach who only sees where the ball landed is guessing at the cause. A coach who sees the actual kinematics — angles, spread, how the shot resolved, in 3D and across the whole motion — is working from the cause forward.

Eadweard Muybridge understood half of this over a century ago, settling a famous bet by breaking a galloping horse's stride into sequential photographs and proving a single glance at movement misses almost everything. But he was still working in two dimensions, from a fixed camera. He could show that the legs moved through a set of positions — not the actual angles and depth producing that motion.

That's the gap we're closing. The sequence matters, the same way it mattered to Muybridge. But the sequence has to be in 3D, or you're still just watching a shadow of the motion instead of measuring it.