Fixing CMJ Eccentric-Concentric Phase Split Errors From Arm Swing and Unweighting Drift

A hitter posts a jump height barely different from last week, but the report flags her RSI as down 22% and her eccentric duration as up 35%. Nothing about the session felt different. The staff pulls the raw trace and finds the culprit isn't fatigue at all: she used a harder arm swing this session, and the software's phase-split logic caught the arm-driven dip in the force curve and called it the true braking minimum. The eccentric-to-concentric boundary got planted in the wrong spot, and every number downstream of that boundary — movement time split, RSImod, peak power windowing — inherited the error.

This isn't a rare glitch. Any algorithm that finds the eccentric-concentric transition by locating a minimum force value or a zero-velocity crossing is vulnerable to exactly this: a vigorous arm swing or an unusually fast or slow unweighting can create a force-time trace shape the algorithm wasn't built to expect, and the split point lands somewhere other than where the athlete actually transitioned from braking to driving. The number on the screen still looks precise to two decimal places. It just isn't measuring what it claims to. This guide walks through why phase splits drift, how to spot it on the raw trace, and a standardization protocol that keeps arm swing and unweighting style from quietly corrupting your RSI and power numbers.

Three Boundaries, Three Different Rules

A countermovement jump gets carved into phases using rules applied to the force-time and velocity-time curves, not by watching the athlete. Movement onset is typically flagged when vertical force drops (or rises) past a threshold set relative to quiet standing — commonly five standard deviations of the bodyweight signal, sometimes a fixed offset like ±20 N. The unweighting-to-braking transition is usually set at the point of minimum force during the descent. The braking-to-propulsion transition — the actual eccentric-concentric split — is set at zero velocity, the instant the center of mass stops descending and starts rising. Takeoff is flagged when force drops toward zero as the athlete leaves the plate.

Every One of Those Rules Assumes a Clean, Single-Peaked Curve

The zero-velocity crossing in particular assumes the velocity trace descends smoothly to a single minimum and then rises smoothly through zero. That assumption holds well for a controlled, hands-on-hips countermovement. It holds much less well once the arms are moving independently of the legs, or once the unweighting phase itself has an unusual shape — both of which change the force trace the algorithm is reading before it ever gets to the part it's trying to locate.

Arm Swing Adds a Second Force Event the Algorithm Wasn't Built For

A free arm swing doesn't just add momentum at takeoff — it adds its own force signature to the plate throughout the countermovement. As the arms decelerate at the bottom of the backswing and reverse direction to drive upward, they impose an additional downward force spike onto the trace, layered on top of the leg-driven braking force. In a hands-on-hips trial that spike doesn't exist, so the true minimum is unambiguous. In a free arm-swing trial, depending on timing, that arm-deceleration spike can sit close enough to the leg-driven minimum that an algorithm scanning for the single minimum force value picks the wrong one — sometimes an arm-driven local minimum a few dozen milliseconds away from where the athlete's center of mass actually reversed direction.

Unweighting Speed Shifts Where the Onset Threshold Gets Crossed

The unweighting phase — the brief drop below bodyweight as the athlete initiates the countermovement — isn't uniform across athletes or even across a single athlete's trials. An explosive, fast unweighting crosses the onset threshold quickly and produces a short, sharp dip. A slower, more deliberate unweighting crosses the same threshold more gradually, and can sit close to the threshold value for 30-50 milliseconds before committing to the descent. When onset detection is threshold-based, that gradual crossing changes exactly when the algorithm decides the movement began, which shifts every downstream phase boundary the same amount even though nothing about the braking or propulsion phases themselves changed. Two trials with an identical braking-to-propulsion transition can report different unweighting durations, different total movement times, and — because RSImod divides jump height by total movement time — different RSImod values purely from unweighting style.

The Two Errors Compound

A session with both a harder arm swing and a faster unweighting than the athlete's usual pattern can shift the reported split in two directions at once: an earlier apparent onset from the faster unweighting, and a misplaced eccentric-concentric boundary from the arm-driven force spike. The movement-time number that comes out the other end can look meaningfully different from a technically identical jump, with the difference sitting entirely in how the software carved up the curve rather than anything the athlete's neuromuscular system actually did.

Chavda et al. (2018): Threshold Choice Alone Changes the Numbers

Chavda, Bromley, Jarvis, Williams, Bishop, Turner, Lake, and Mundy (2018, Strength and Conditioning Journal) reviewed how different onset and phase-boundary detection methods applied to the same raw force-time data produce meaningfully different calculated values for phase durations and RSImod. Their central point is that the phase split isn't a property of the jump — it's a property of the detection method applied to the jump, and switching onset thresholds or minimum-force windows on identical raw data changes the reported unweighting duration and, downstream, RSImod, without the athlete doing anything different. The practical limitation the authors flag: there's no single universally adopted threshold standard across force plate software vendors, so a phase-duration figure reported by one system isn't automatically comparable to the same figure from another system or published study using a different detection method.

Lees, Vanrenterghem, and De Clercq (2004): Arm Swing Reshapes the Force Curve Near the Transition

Lees, Vanrenterghem, and De Clercq (2004, Journal of Biomechanics) examined how arm swing contributes to countermovement jump performance and found that a free arm swing meaningfully increases jump height relative to an arms-fixed condition, largely by extending force production into the later part of the propulsion phase and altering the vertical ground reaction force profile in the second half of the countermovement — precisely the region where eccentric-concentric transition detection operates. The effect was consistent and moderate-to-large in magnitude for jump height, though the study used a controlled biomechanics-lab protocol with reflective markers and a fixed testing order rather than a field-testing battery, so the exact force-curve shape change may look somewhat different on a portable plate with a simpler contact-based onset threshold than the marker-synchronized analysis the original study used.

The Combined Takeaway

Neither finding is really about fatigue or performance on its own — both are about measurement mechanics. Between them, they explain why a coaching staff can see RSI or phase-duration numbers move session to session for reasons that have nothing to do with the athlete's underlying neuromuscular state, and everything to do with arm position and unweighting tempo interacting with a threshold-based detection algorithm.

Pull Up the Raw Curve, Not Just the Summary Numbers

The summary screen that reports jump height, movement time, and RSImod won't show you a misplaced split point. You have to look at the force-time and velocity-time traces directly, with the software's marked phase boundaries overlaid, to catch this.

• A double dip in the force trace during the countermovement — one dip from arm deceleration, a second from leg braking — is the clearest sign the algorithm may have grabbed the wrong minimum. If the marked braking-to-propulsion boundary sits at the earlier, shallower dip rather than the deeper leg-driven one, the split is wrong.
• A velocity trace that flattens near zero for an extended stretch rather than crossing it cleanly usually means a slow, deliberate unweighting or a hesitant transition, and the exact millisecond the software picks as the zero crossing becomes somewhat arbitrary within that flat window.
• A reported eccentric duration that looks unusually short paired with an unusually long propulsion duration (or the reverse) relative to the same athlete's typical split, on a jump height that hasn't moved much, points at a shifted boundary rather than a genuine change in movement strategy.

Five Steps to Keep Arm Swing and Unweighting Style From Corrupting the Split

1. Fix the arm condition per test purpose. Use hands-on-hips for any session where phase durations and RSImod feed into a fatigue-monitoring trend, since it removes the arm-driven force spike entirely. Reserve free arm swing for sessions specifically assessing performance capacity, where the extra height matters more than clean phase splitting.
2. Standardize the onset threshold across your whole roster and don't let it change between software updates without a note in the athlete's file. A threshold change on its own can shift every reported phase duration by a consistent amount that looks like a real training effect if nobody flags it.
3. Visually audit the trace on the athlete's first few sessions to confirm the software's automatic split lands where it should. Once you've confirmed the pattern is clean for that athlete's typical technique, spot-check periodically rather than every session.
4. Cue unweighting tempo the same way every time. Quick down, quick up versus sink and drive produce genuinely different unweighting durations. Pick one cue per testing protocol and keep it identical across sessions and staff members.
5. Flag any session where the trace shape looks atypical — a double dip, an unusually flat velocity zone — before trusting the RSI or phase-duration number from that trial, and prefer the best of three trials by height rather than averaging a trial with a clean split against one with a questionable one.

Two Trials, Same Athlete, Same Day

An athlete performs two CMJ trials four minutes apart: one hands-on-hips, one with a free arm swing, both to a near-identical jump height.

Reading It

Total movement time and jump height barely moved, so RSImod — which uses only those two numbers — looks reasonably stable. But the eccentric-concentric split shifted hard in one direction: nearly 28% less eccentric time and 26% more concentric time on the free arm-swing trial, purely because the arm-driven force spike moved where the algorithm placed the braking-to-propulsion boundary. If a program is tracking eccentric duration specifically as an early fatigue marker — a reasonable thing to do, since braking phase lengthening is one of the earliest signs of accumulated fatigue — this pair of trials would show what looks like a dramatic improvement in braking speed that has nothing to do with the athlete's neuromuscular state and everything to do with switching arm conditions between trials.

Where This Goes Wrong in Practice

• Mixing arm conditions within the same tracking series. Alternating hands-on-hips and free arm-swing trials across sessions and then comparing eccentric duration or RSImod trend lines between them treats a measurement artifact as a training response.
• Trusting the summary RSI without ever looking at the raw trace. A misplaced split point doesn't announce itself on the results screen. It shows up as a phase-duration number that looks a little too good, or a little too alarming, relative to the athlete's history.
• Cueing arm swing loosely. A bare instruction to swing the arms produces wildly different timing between an athlete who initiates the backswing early and one who keeps the arms still until the last instant. That timing difference changes exactly where the arm-driven force spike lands relative to the leg-driven minimum.
• Changing onset threshold settings mid-season after a software update without documenting it, then wondering why every athlete's unweighting duration shifted by roughly the same amount on the same day.
• Averaging a clean-split trial with a questionable-split trial instead of visually confirming which trial has the reliable boundary and using that one for the phase-specific numbers, even if a different trial had the higher jump height.