Stability, Traction & Setup Repeatability in the Bench Press

A bench can feel stable while the athlete slowly slides. A grippy cover can resist sliding while a soft or mobile support still deforms. A rigid, high-friction surface can still produce inconsistent lifts if eye position, feet, grip and touch point change from set to set. These are not semantic details. Stability, traction and setup repeatability are separate variables with different measurements and different evidence.

Direct bench-press research mainly compares stable benches with intentionally unstable supports such as Swiss balls, balance cushions or unstable loads. Those studies show that instability can reduce force, power or six-repetition maximum under some protocols, yet at least one study found no difference in one-repetition maximum or selected muscle activation [1][2][3]. Stable-support evidence therefore clarifies task constraints, and reliability research shows how to standardize the comparison. Upholstery friction is a separate design variable until friction-specific bench trials measure drift and performance directly.

Three variables, three measurement plans

These measurements interact. More torso drift can change the shoulder-to-bar relationship, and a deformable support can alter both position and timing. Each variable still keeps its own measurement identity: a traction conclusion requires friction or torso drift to be manipulated or measured.

What stable-versus-unstable studies show

Instability changes the task. It may require more correction while reducing the load or speed that can be expressed through the prime movers. The size and even direction of an effect depend on how instability is introduced.

Koshida and colleagues compared dynamic bench pressing on stable and unstable support. Peak force, power and velocity were lower under the unstable condition—approximately 5.9%, 9.9% and 9.1%, respectively [1]. Saeterbakken and Fimland tested sixteen resistance-trained men using a stable bench, balance cushion and Swiss ball. Stable pressing supported a higher six-repetition maximum and greater pectoralis and triceps activation, while rectus-abdominis activity was higher on the Swiss ball [3]. The finding is a tradeoff, not evidence that instability is universally “bad”: the task shifted toward stabilization while limiting heavy pressing.

Marquina and colleagues tested sixty-three physically active young men across loads on a bench and Swiss ball. Peak force and power were generally impaired in the unstable condition, especially as the task moved toward heavier, lower-velocity pressing [4]. By contrast, Goodman and colleagues found no meaningful difference in one-repetition maximum, elbow range of motion or selected muscle activation between stable-bench and exercise-ball conditions in thirteen participants [2]. Protocols, familiarization, instability magnitude and measurement choices help explain why a single slogan cannot represent this literature.

Unstable loads are another exposure

Moving the instability from the support to the bar creates a different problem. Trained men pressing an unstable load showed altered activation of stabilizing musculature despite using a lower absolute load [5]. Analysis of the same general task showed a more variable and less predictable bar path in the mediolateral and anteroposterior directions [6]. Those studies support program-specific use of unstable loads. Bench-pad traction requires a separate direct comparison.

Pivotal study: instability can tax output, not just balance

The Koshida study is useful because it measured outputs lifters care about—force, power and velocity—during dynamic pressing rather than inferring difficulty from appearance. The unstable support reduced all three on average [1]. That result illustrates a task constraint: some of the system’s capacity is spent controlling an uncertain base, and the athlete expresses less external output under that protocol.

The study directly shows that an intentionally unstable support reduced acute force, power and velocity under its tested protocol. Its boundary is specific: that support condition differs from conventional changes in foam, texture or width, acute output is distinct from long-term adaptation, and athlete–pad friction was not measured.

Read beside Goodman’s null findings and the later studies, the pivotal lesson is methodological: define the instability, load and outcome before generalizing. “Stable” is not a product superlative; it is a condition that needs an operational definition.

A 2026 training signal—with an unequal-load caveat

Zhu and colleagues assigned 18 untrained men to 12 weeks of unstable-load or Smith-machine bench training. The unstable-load group progressed through nominal intensities of 55%, 65% and 75% 1RM; the Smith group used 65%, 75% and 85%. A group-by-time interaction favored unstable loading for the study’s 1RM outcome (p = .033; reported d = .82). The result is a preliminary longitudinal signal, not a clean demonstration that instability is superior: training loads were not matched, the sample was small and untrained, and the test was close to the practiced task [7].

The study adds a small longitudinal signal to the acute evidence base while preserving the broader task-specific finding that instability often limits immediate external output. Traction remains a separate research question.

Traction: clear mechanics, incomplete outcome evidence

At an interface, static friction resists the start of sliding up to a limit that depends on the materials, normal force and local conditions. Sweat, clothing fabric, cleaning products, upholstery wear, contour and pressure distribution may all alter the real contact. A laboratory coefficient measured with one material pair should not be treated as an athlete outcome.

In the bench press, sliding can matter because it changes coordinates that organize the lift: eyes relative to the bar, shoulders relative to the rack, pelvis on the pad and the timing of leg effort. A higher-traction surface is therefore rationally designed to resist drift. That is a design-intent claim unless a comparative test measures friction or torso displacement.

The stable-versus-unstable studies changed support conditions rather than cover material alone. Effects of a grippier pad on bar velocity, force transfer, shoulder load or injury outcomes remain direct test questions.

Repeatability determines what a test can detect

A measurement must have less noise than the change we want to interpret. Stock and colleagues tested free-weight bench-press velocity in trained men on two occasions. Reliability was moderate to high at lighter and moderate loads, but error increased at 80% and 90% of one-repetition maximum [8]. Heavy single-repetition velocity is therefore a noisy place to declare an equipment effect from one trial.

Zemková and colleagues compared chest-press power assessment on stable and unstable surfaces. Stable-bench measures were highly reliable, while reliability deteriorated for some unstable, heavy conditions; reported standard errors approached roughly one-fifth for some 80% measures [9]. The support condition can influence not only performance but measurement quality.

Instrumentation does not remove the need for protocol control. A linear position transducer showed high validity and reliability for bench and squat measures in trained men, but power variables carried more error than velocity [10]. Bar kinematics also drift with fatigue within a bench-press set [11]. Warm-up, load, rest, intent, pause, grip, touch point and repetition number must therefore be standardized.

Reliability is not the same as agreement—or validity

A device can rank athletes similarly across sessions while still giving values that differ enough to matter for an individual. That is why a reliability coefficient should be read beside absolute error, such as a standard error of measurement or limits of agreement. A small equipment difference that sits inside normal test–retest variation should not be promoted as a meaningful response simply because one repetition was higher.

Validity asks another question: did the method measure what the claim says? A transducer that validly measures bar velocity does not measure scapular motion, torso friction or internal joint load. Video landmarks can reveal drift but not the force required to prevent it. Pressure mapping can show contact distribution but not long-term adaptation. Each tool earns only the claim attached to its measured variable.

For a practical single-athlete comparison, collect several technically acceptable repetitions across at least two familiarized sessions per condition. Report the mean, spread and any excluded trials, and keep the repetition number consistent. Randomize condition order when feasible so that fatigue and learning do not always favor the same surface. If athletes and coaches cannot be blinded to the pad, outcome processing can still be blinded. These steps do not turn a gym comparison into a clinical trial, but they make the conclusion auditable.

Repeatability should never be confused with desirability. A lifter can reproduce a poor match to the task, and a new technique may be variable while it is being learned. The question is whether variability is acceptable for the intended use and whether it decreases with familiarization.

When two conditions differ by less than ordinary session-to-session error, the honest result is inconclusive rather than “no effect” or “the same.” More familiarization, more repeated trials or a more sensitive outcome may be needed. That language protects athletes from chasing noise and protects useful equipment ideas from being judged by an underpowered one-day comparison.

A weight-room repeatability audit

Use a simple audit before comparing programs, pads or techniques. The goal is not to freeze every athlete into one setup; it is to know what was held constant and what changed.

1. Identify the hardware. Log bench, pad configuration, bar, rack height, safety height and any attachment. Competition testing should use the applicable equipment and commands [12].
2. Record the athlete coordinates. Index eyes or forehead to the racked bar, grip marks, foot locations, head/shoulder/pelvis contacts and torso midline.
3. Define the repetition. Specify touch point, pause or touch-and-go standard, depth, intent and acceptable assistance from a handoff.
4. Control the test context. Match warm-up, load calculation, rest, time of day where practical, clothing at the interface and fatigue state.
5. Measure drift. Film from fixed views or use position sensors. Track head and pelvis displacement, touch-point spread, path variability and velocity across repeated sets.
6. Repeat before deciding. Compare multiple exposures. Report typical error or at least the range of repeated values, not only the best repetition.

When traction is the research question, hold foam, contour, dimensions, clothing and load constant; vary the cover condition; quantify static/kinetic friction and torso displacement; and blind outcome analysis where possible. That design can separate a real interface effect from expectation or setup variation.

Launch Pad relevance and the Joint Ops research path

Published Launch Pad studies evaluated a complete interface. In a randomized crossover study of ten resistance-trained men, five repetitions at 70% of one-repetition maximum on the novel pad condition produced higher measured pectoralis surface electromyography, bar velocity and range of motion; the power difference was not statistically significant [13]. The study measured the complete interface rather than isolating upholstery friction.

In a four-week supervised eccentric-overload program, intermediate-trained men improved one-repetition maximum more in the Launch Pad condition than on a flat bench [14]. In an eight-week supervised program with male collegiate football players, the Launch Pad group improved tested strength, repetitions and medicine-ball performance more than the flat-bench group [15]. These are complete-condition training outcomes, not proof that traction, stability or any single component caused the differences.

Joint Ops is an active AMM research-and-development program whose working prototype was publicly debuted at CSCCa in May 2026; it is not a published outcome study. The evidence plan should quantify cover friction, pressure distribution, torso drift, bar-path variability, force and velocity reliability, tolerance and training outcomes while holding other geometry constant. No Joint Ops performance, rehabilitation or injury-prevention outcome has been published or is claimed here.