Fixing Contact-Time Threshold Misdetection in Reactive Jump Tests

Halfway through a pre-season testing block for a men's basketball roster, someone adjusts the contact-detection threshold on the force plate software from 50 N down to 20 N, hoping it will catch lighter athletes better. Nobody writes it down anywhere except a settings screen nobody screenshots. Three weeks later, reactive strength index across the whole roster has drifted up by roughly 18%, and the strength staff is in a meeting debating whether the off-season conditioning block worked better than anyone expected. It didn't. The threshold that decides when a foot officially counts as being on the ground moved, ground contact time got shorter across the board because of it, and RSI - jump height divided by contact time - inflated right along with it. Nothing about the athletes changed. The line the software draws between airborne and grounded changed, and every downstream number followed it.

Almost no device measures ground contact time directly. What it measures is a continuous signal - vertical force from a plate, acceleration from an IMU, a beam-break from a photocell mat - and then applies a rule for deciding when that signal crosses from flight into contact and back again. That rule is the threshold. A force platform typically flags touchdown the instant vertical force rises above some cutoff, often expressed in newtons or as a percentage of the athlete's body weight, and flags toe-off when force falls back below it. A shank- or foot-mounted accelerometer does something similar around the 1g baseline, watching for the signal to break out of a small deadband window as impact hits and settle back into it as the foot unloads.

The threshold exists for a good reason: raw signals never sit at a perfect, silent zero during flight. Vibration in the plate frame, drift in the accelerometer's baseline, and electrical noise all put small, real fluctuations into the trace even when the athlete's foot is nowhere near the ground. Without some cutoff, a device would trigger false contacts on noise alone. The problem is that the cutoff value is a judgment call, not a physical constant, and there is no single number that works for every athlete, every device, and every jump. Set it in the wrong place and the device stops measuring the athlete's ground contact time and starts measuring its own settings screen.

When the threshold sits well above the noise floor, it also sits well inside the real signal that a light or fast touchdown produces. Two things happen at once, both working in the same direction. Touchdown gets flagged late, because the force or acceleration has to climb past a higher bar before the device agrees contact has started. Toe-off gets flagged early, because force is already falling through the propulsive phase and crosses back below that same high bar sooner than the foot actually leaves the ground. The contact window gets squeezed from both ends, ground contact time reads shorter than what actually happened, and because RSI is jump height divided by contact time, a smaller denominator with the same numerator produces a larger, better-looking RSI that the athlete didn't earn.

Street, McMillan, Board, Rasmussen, and Heneghan (2001, Journal of Applied Biomechanics) demonstrated this mechanism directly, though for jump height rather than contact time specifically: recomputing the same countermovement jump trials under different threshold choices for identifying take-off and landing instants shifted computed jump height by several percent, with the largest threshold comparisons producing differences approaching 6%. Their work used the impulse-momentum method on a single force platform system and examined take-off and landing for jump height, not the touchdown-to-toe-off window inside a stretch-shortening cycle test - but the underlying mechanism is identical. Any metric built on a threshold-crossing timestamp moves when the threshold moves, and the direction of that movement is predictable: a higher bar shortens whatever window it's measuring.

Set the threshold too close to zero and the opposite failure appears, for a related but distinct reason. A stiff landing doesn't produce one clean spike - it produces a sharp impact followed by a brief ringdown, a few milliseconds of mechanical oscillation in the plate, shoe, or mounting bracket as the initial shock settles out. A threshold set near the noise floor can catch the leading edge of that ringdown before the foot has genuinely made contact, and catch the trailing edge of it after the foot has genuinely left, because the oscillation briefly re-crosses the low bar on its way down to true zero. Both ends of the contact window get pulled outward instead of squeezed inward, ground contact time reads longer than it actually was, and RSI reads lower than it should.

Bosquet, Berryman, and Dupuy (2009, Journal of Strength and Conditioning Research) compared two commercial optical timing systems measuring the same hopping and jumping trials and found strong correlations between them (ICC values generally above 0.90) alongside a consistent systematic bias in contact time, with one system reading shorter or longer than the other by a margin large enough to shift derived reactive strength metrics meaningfully. Their comparison treated both systems as black boxes rather than isolating the exact detection threshold inside each one, and it covered two optical systems rather than force plates or IMUs specifically - but the pattern of systematic, direction-consistent bias between two devices measuring the identical physical event is exactly what a mismatched threshold produces. Glatthorn et al. (2011, Journal of Strength and Conditioning Research) added a mechanistic piece to this picture, validating the Optojump photocell system against a force-platform criterion and reporting strong agreement on flight time (r > 0.95) but weaker, less consistent agreement specifically on contact time, attributing part of the gap to the photocell's beam sitting a few millimeters above the ground - a fixed positional threshold that, like a force or acceleration cutoff, does not always land where the true foot-ground boundary does.

The physical quantity being thresholded differs by device, but the failure mode doesn't. Use the table below as a starting range to check your current setting against, then confirm with the calibration protocol in the next section - drop height, athlete mass, and mounting stiffness all shift where the correct value actually sits for a given setup.

Run this once per device and mounting setup before trusting RSI trend data from it, and repeat it after any firmware update, software update, or manual threshold change - the roster drift in the opening example was entirely preventable with this check.

1. Record 3-5 reactive or drop jumps at the height and load you actually intend to test, capturing side-view video at 240 fps or higher alongside the device's own reading.
2. From the video, manually frame-count true touchdown (the first frame the foot visibly stops descending and begins to deform against the surface) and true toe-off (the last frame before the foot is fully airborne), then multiply the frame count by frame duration to get a criterion ground contact time for each rep.
3. Compare the device's reported ground contact time against the video criterion for each rep, not just the set average.
4. If the device reads shorter than the video criterion by more than roughly 5%, the threshold is likely set too high - lower it one increment and retest the same reps.
5. If the device reads longer than the video criterion by more than roughly 5%, or the raw trace shows a brief secondary bump just before touchdown or just after toe-off, the threshold is likely too low - raise it one increment and retest.
6. Repeat the adjust-and-retest cycle until the device's ground contact time sits within about 5 ms or 3% of the video criterion, whichever margin is larger, across at least three consecutive reps.
7. Log the final threshold value alongside the device's firmware or software version, the mounting location, and the drop height used - re-validate whenever any of those four things change, not just when numbers start looking strange.

A club-level sprinter performed a 30 cm drop jump onto a shank-mounted IMU, landing with an identical, video-confirmed jump height of 0.24 m across the comparison. The same raw acceleration trace from that single rep was then reprocessed three times, once at each of three acceleration deadband settings, alongside the 240 fps video criterion for ground contact time.

The physical jump never changed across these four rows - only the rule deciding where contact begins and ends did. The too-low setting pulled roughly 12 ms of post-landing ringdown into the contact window on each end, stretching GCT and dragging RSI down to 1.00. The too-high setting cut into both the early touchdown force and the late toe-off force, shaving 31 ms off contact time and pushing RSI up to 1.28 - a number that would read as a meaningfully stronger athlete than the video-confirmed 1.10, on the exact same jump. The calibrated setting landed within 3 ms of criterion, close enough that the small remaining gap falls inside normal frame-rounding error rather than threshold bias. A 0.18-point swing in RSI purely from a settings screen is large enough to change an athlete's ranking on a team leaderboard, and large enough to mask or fabricate a training effect if the threshold changes between a pre-test and a post-test without anyone noticing.