Why a Wrist-Worn Sensor Under-Reads Vertical Jump Height

A club volleyball setter's wristband has logged her countermovement jump at 38-40 cm for three straight weeks, and her strength coach is one bad session away from flagging her for an overtraining check. Then, almost by accident, someone runs the same jumps past a hip-mounted reference sensor and a slow-motion camera during a routine calibration. Both read 45-46 cm, matching what she jumped in preseason testing with the same reference method. She has not lost six centimeters of leg power. Her arm swing is the problem, or more precisely, where the sensor sits on that arm is. Every time she drives her arms overhead to help launch, her wrist accelerates, decelerates, and settles at a different moment than her center of mass does, and a sensor strapped to that wrist has no way to distinguish 'my wrist stopped rising' from 'the jump is over.' It reports the wrist's story, not the body's, and on any jump with a real arm swing those two stories split by several centimeters, almost every single rep.

A wearable jump sensor never measures jump height directly. It measures acceleration at the point where the device is strapped on, then turns that acceleration into a height figure through one of two methods: double integration, which runs the raw acceleration signal through two rounds of integration to get displacement, or a flight-time method, which watches for the acceleration signature of takeoff and landing and converts the gap between them into height using flight time squared times gravity, divided by eight. Both methods share one unstated assumption: that the motion recorded at the mount point is a faithful stand-in for the motion of the athlete's whole-body center of mass.

That assumption holds up reasonably well at the hip or waist. The pelvis sits close to the actual center of mass, and short of a hip-hike or heavy lean, it translates through space in roughly the same path the whole body's COM does. The wrist does not get that same free pass. It sits at the far end of a kinetic chain, three joints away from the trunk, and everything the shoulder and elbow do to help the jump also gets layered directly into whatever the wrist-mounted accelerometer records. The device cannot separate 'the body went up' from 'the arm swung.'

A countermovement jump with a free arm swing runs through phases the arms and the trunk do not share evenly. During the eccentric dip, the arms swing backward and down, often still gaining speed after the hips have already started reversing direction. During the concentric drive, the arms swing forward and upward hard enough that peak arm angular velocity is commonly reached before peak vertical velocity at the center of mass, since the arm's job is to help pull the trunk upward, not to match its timing. Once the feet leave the ground, the story keeps diverging: the center of mass is now in pure ballistic flight, governed by gravity alone, while the arms are still under active muscular control, decelerating toward full overhead extension well into the flight phase.

None of that arm motion is wasted effort. It genuinely helps the jump, contributing extra height through the added upward force during the drive and a following momentum transfer as the arms brake mid-flight. The problem is narrower than whether arm swing helps: it is that a sensor riding on the arm records the arm's own timeline instead of the body's, and those two timelines are never the same shape.

Three separate lines of research, none of which set out to study wrist-worn jump sensors directly, line up to explain exactly why this happens. Lees, Vanrenterghem, and De Clercq (2004, Journal of Biomechanics) used force-platform and full-body kinematic data to show that an arm swing raises jump height through two distinct mechanical effects: an early upward 'lift' during the drive phase, and a later 'pull' effect where the arms decelerate mid-flight and transfer momentum back into the trunk. Their kinematic traces show peak arm segment velocity clearly out of phase with peak center-of-mass velocity. Their study never touched a wearable device, but it confirms the underlying mismatch is real and well documented, not a quirk of one product.

Casartelli, Muller, and Maffiuletti (2010, Journal of Strength and Conditioning Research) tested a waist-mounted accelerometer against a force platform across squat and countermovement jumps and found acceptable test-retest reliability, but a validity problem: the device's error against the force platform grew larger as jump height increased. That is the COM-proximal baseline case. Even at the waist, close to the true center of mass, a sensor's algorithm can drift from ground truth, which sets a floor on error before a device is even moved out to a limb.

Balsalobre-Fernandez, Kuzdub, Poveda-Ortiz, and Campo-Vecino (2016, Journal of Strength and Conditioning Research) validated a wrist and forearm-mounted device against a linear position transducer during the back squat and found strong agreement for bar velocity. That result does not transfer to jump testing, and the reason is instructive: in a back squat, the hands stay fixed on a barbell, so the wrist moves rigidly with the load being measured. A free-arm-swing jump removes that rigid coupling entirely. The wrist is free to swing on its own schedule, which is precisely the condition the squat validation never tested.

Once a jump leaves the ground, the only external force acting on the whole body is gravity, so the center of mass has to follow a clean, unbroken parabola. Internal forces cannot change that path. But internal forces, the shoulder muscles braking the arm toward full elevation, can absolutely change how one segment moves relative to the trunk, even mid-air. A wrist that is still decelerating under active muscular control does not show a clean minus-one-g signature the instant the feet leave the floor. It shows gravity plus whatever braking torque the shoulder is applying, and that combination looks, to an algorithm scanning for a stable free-fall signature, less like a body already airborne and more like a body that has not quite let go of the ground yet.

That is the direction the error runs. A flight-time algorithm that delays marking takeoff because the wrist has not yet settled into a clean ballistic signature shortens the flight-time window it eventually counts. Landing detection is comparatively unaffected, since the wrist is usually done decelerating well before the feet return to the floor. Shorter measured flight time, run back through the same height formula, comes out as a shorter jump. A double-integration algorithm suffers the same directional bias from a different angle: the arm's own braking deceleration, layered on top of the true ballistic signal, subtracts from the integrated displacement rather than adding to it. Either way, the wrist sensor is not randomly noisy around the true number. It is systematically pulled toward reading low.

Run this before trusting a wrist-worn device for any jump-height trend, and repeat it whenever the device firmware or the athlete's jump technique changes noticeably.

1. Confirm the device's mount location and whether it uses flight-time detection or double integration to compute height; check the spec sheet or app settings if it is not obvious.
2. Have the athlete perform 4-5 countermovement jumps with hands fixed on the hips, arms swing removed entirely, while the wrist device and a COM-proximal reference (hip sensor, force platform, or 240 fps video) record simultaneously.
3. Have the same athlete perform 4-5 countermovement jumps with a normal, free arm swing, again with both measures recording at the same time.
4. Calculate the average divergence, wrist reading minus reference reading, separately for each condition.
5. A near-zero divergence in the hands-on-hips condition paired with a consistently negative divergence in the free-arm-swing condition confirms arm-swing placement error as the cause, rather than a generic calibration fault, which would show up in both conditions equally.
6. Decide the fix based on what the sport requires: relocate the sensor to the hip or waist if the hardware supports it and arm swing does not need to be tested; if the wrist is the only mount point available and free arm swing is sport-specific, keep it as a velocity or work-rate tool but stop treating its jump-height number as ground truth without a periodic COM-based cross-check, since the size of the under-read varies by athlete and by how vigorously each one swings their arms.

A collegiate basketball player performed both conditions in the same session, five reps each, with a wrist-worn accelerometer and a simultaneous 240 fps video reference. The table shows the average of each set of five reps.

With arm swing removed, the wrist sensor and the video reference agreed within half a centimeter, well inside normal rep-to-rep noise. With a full arm swing restored, both measures rose, since the arm swing genuinely added height, but the wrist sensor rose far less than the true jump did, opening a 6.1 cm gap that had nothing to do with the athlete's legs and everything to do with where the sensor was strapped on.