The research team, led by Tsuyoshi Nagatani of Edith Cowan University alongside colleagues including Christopher Latella, Paul Comfort, Eric Drinkwater and G. Gregory Haff, compared three alternatives to the traditional set, in which repetitions are performed consecutively with rest only between sets. Cluster sets insert brief rest periods between individual repetitions or small groups of repetitions within a set, adding total rest and lengthening the session. Rest-redistribution strategies take a different route: instead of adding rest, they rearrange the rest already in the program. Intra-set rest-redistribution carves a slice out of the between-set rest and moves it inside the set, while inter-set rest-redistribution splits a large set into many smaller ones, for example turning three sets of ten into six sets of five, keeping total rest constant.

Because only a handful of studies have directly pitted cluster sets against rest-redistribution, the authors turned to a network meta-analysis, a statistical framework that combines direct comparisons with indirect ones linked through a common comparator, in this case the traditional set. They searched PubMed and Web of Science, screened 1,121 records, and ultimately included 37 studies encompassing 611 unique participants, 557 of them male and 54 female. The outcomes of interest were mean velocity, mean propulsive velocity and peak velocity, the kinematic signatures that velocity-based training practitioners use to gauge fatigue and readiness. Analyses were run in a Bayesian framework using Markov Chain Monte Carlo simulation, with posterior standardized mean differences and 95 percent credible intervals, and treatments were ranked with SUCRA scores, which express the probability that each set structure is the most effective.

The headline finding is that all three alternative structures beat the traditional set at preserving mean velocity. Cluster sets showed a standardized mean difference of 0.60 (95 percent credible interval 0.42 to 0.80), intra-set rest-redistribution 0.41 (0.08 to 0.75) and inter-set rest-redistribution 0.42 (0.26 to 0.61). For peak velocity, cluster sets (0.48, 0.20 to 0.76) and intra-set rest-redistribution (0.39, 0.04 to 0.73) again outperformed traditional sets, but the credible interval for inter-set rest-redistribution crossed zero, leaving its effect on peak velocity uncertain. In the SUCRA rankings, cluster sets held the highest probability of being the best structure for both mean velocity (99.7 percent) and peak velocity (95.6 percent), followed by intra-set rest-redistribution and then inter-set rest-redistribution.

Yet the more consequential result is what the analysis did not find. When the alternative structures were compared directly with one another, the credible intervals for mean and peak velocity crossed zero in every case. Cluster sets versus intra-set rest-redistribution produced a credible interval of −0.19 to 0.58 for mean velocity, and cluster sets versus inter-set rest-redistribution ranged from −0.05 to 0.41. In plain terms, although cluster sets carry the highest probability of ranking first, the evidence does not support a practically meaningful advantage over either redistribution strategy. The physiological logic behind the cluster advantage is straightforward: brief intra-set pauses allow partial resynthesis of phosphocreatine, blunting the metabolic disruption, lactate accumulation and neural drive decrements that cause velocity to crater in uninterrupted sets. But rest-redistribution achieves a similar effect without adding a single second to the session, which is precisely why the question of whether cluster sets are meaningfully better matters so much to practitioners.

The authors also tested whether the benefits depend on context, using network meta-regressions. Exercise type, comparing lower-body movements such as squats and clean pulls with the bench press, did not significantly moderate the effects, with a regression coefficient of 0.10 and a credible interval spanning zero, although there was a tendency toward larger benefits in lower-body exercise, plausibly because larger muscle masses and longer ranges of motion generate greater metabolic stress. Relative training intensity, split between moderate loads of 60 to 79 percent of one-repetition maximum and high loads of 80 percent or more, likewise failed to moderate the effects (coefficient 0.15, credible interval −0.12 to 0.43). The practical implication is that alternative set structures appear robust across exercise selection and loading zones, rather than being niche tools for specific lifts or intensity zones.

Peak velocity told a more nuanced story. The uncertain effect of inter-set rest-redistribution was driven partly by the small number of studies and by the exercises examined. The one study reporting a clear peak-velocity benefit examined the clean pull from the floor, a full-range movement with large barbell displacement and high metabolic demand, whereas the two studies showing no benefit used partial weightlifting derivatives, the hang pull and countermovement shrug, which involve shorter displacement and less time under tension and may simply not generate enough fatigue for redistributed rest to matter. The authors caution that the apparent superiority of cluster sets and intra-set redistribution for peak velocity may therefore be inflated by these study characteristics, and they call for more research on full-range, multi-joint movements.

For coaches, the practical takeaway may be the most valuable part of the paper. Because intra-set rest-redistribution produced velocity outcomes statistically comparable to cluster sets without extending session duration, it emerges as a time-efficient alternative for athletes training under tight schedules. Inter-set rest-redistribution, while effective for mean velocity, carries logistical friction: some implementations redistribute rest between every single repetition, forcing athletes to rack and un-rack a barbell dozens of times, and in team environments one athlete must finish all their scattered repetitions before partners can rotate through. The authors suggest that cluster sets and intra-set redistribution are the more practical choices in real-world settings, with the final decision guided by session time, equipment and training context rather than by any large performance differential.

The review is not without limitations, which the authors acknowledge candidly. Several studies comparing intra-set rest-redistribution with traditional sets failed to equalize total rest duration, inadvertently giving the intervention more recovery and potentially inflating its benefit. Evidence for peak velocity rested on far fewer studies than mean velocity, some consistency checks could not be performed for certain comparisons, and the participant pool was overwhelmingly male, preventing any analysis of sex as a moderator. Publication bias was detected for the mean velocity model, though not for peak velocity. Registered prospectively on the Open Science Framework, the review nevertheless represents the first analysis to treat the two rest-redistribution variants as distinct constructs, and its central message is likely to reshape how velocity-focused programming is discussed: the way rest is arranged matters far more than the label attached to it, and almost any thoughtful departure from the traditional set will keep the bar moving faster.

Subject of Research: A systematic review and Bayesian network meta-analysis comparing cluster set and rest-redistribution set configurations for maintaining movement velocity during resistance training.

Article Title: The Effectiveness of Cluster vs. Rest-Redistribution Set Configurations to Maintain Movement Velocity During Resistance Training: A Systematic Review and Bayesian Network Meta-analysis

Article References: Nagatani, T., Latella, C., Yang, J., Kendall, K. L., Vial, S., Comfort, P., Drinkwater, E. J., & Haff, G. G. (2026). The Effectiveness of Cluster vs. Rest-Redistribution Set Configurations to Maintain Movement Velocity During Resistance Training: A Systematic Review and Bayesian Network Meta-analysis. Sports Medicine – Open, 12(1), Article 141. https://doi.org/10.1186/s40798-026-01106-w

Image Credits: AI Generated

DOI: 10.1186/s40798-026-01106-w

Keywords: resistance training, cluster sets, rest redistribution, movement velocity, velocity loss, neuromuscular fatigue, network meta-analysis, strength and conditioning, velocity-based training, set structure, phosphocreatine, sports science

The study, published in Sports Medicine – Open, was led by Simone Montenegro of the German Sport University Cologne, together with Pamela Wicker, Tim Wiedenmann, Ludwig Rappelt and Lars Donath. The team drew on the Open Powerlifting database, restricting their sample to drug-tested, unequipped “classic” powerlifting competitions sanctioned by the International Powerlifting Federation and its national affiliates between January 2015 and March 2026. To ensure a homogeneous contemporary population, they limited analysis to full-power results comprising the squat, bench press and deadlift, and to athletes aged 21 to 49 years, a window designed to encompass the age of peak competitive performance in the sport, reported at roughly 27 years.

A crucial methodological decision shaped everything that followed. Large athletic databases are noisy: a single athlete may appear dozens of times, and most of those entries capture sub-maximal performances rather than career-best efforts. To solve this, the researchers applied what they call a “Best-Ever” filter. Each unique athlete was linked by name and sex, and only the single competition containing their highest recorded total was retained. Anchoring on the best-total competition, rather than stitching together best lifts from separate events, preserved the internal coherence of the three lifts as one simultaneously achieved performance. The final sample comprised 101,898 athletes, of whom 67,789 were male and 34,109 were female.

The statistical engine of the study was quantile regression, a technique that estimates specific points of the performance distribution, in this case the 50th, 75th, 90th and 99th percentiles, independently, rather than focusing only on the average lifter. Standard errors and confidence intervals were derived through bootstrap resampling to ensure robust inference at the extreme quantiles, and the researchers labeled the four tiers Intermediate, Advanced, Elite and World Class, following terminology common in strength and conditioning practice. The analytical unit was the bodyweight multiplier, the ratio of load lifted to body mass, but the authors were explicit that the ratio served as a diagnostic tool, not a benchmark: the entire point was to test whether a single fixed multiplier could mean the same thing across the body-mass spectrum.

It cannot. The analysis revealed a highly significant, systematic, non-linear decline in relative strength with increasing body mass across all lifts, all percentiles and both sexes. The most striking illustration comes from the male world-class tier: the smoothed 99th-percentile deadlift multiplier fell from 4.37 times bodyweight in the 59-kilogram class to just 2.65 times bodyweight in the 120-plus-kilogram category. In other words, a numerical standard that an elite lightweight could achieve comfortably sits at the absolute frontier of what an elite super-heavyweight can manage. The decay coefficients are small but relentless, and the researchers quantify their meaning precisely: a coefficient of -0.01 means that for every 10 kilograms of additional body mass, an athlete’s expected multiplier drops by exactly 0.1 times, for instance sliding from a 2.1-times to a 2.0-times squat.

Perhaps the most novel finding is that this decline is not uniform across skill levels. Pairwise comparisons of the quantile slopes, conducted with Wald tests for equality of coefficients, confirmed that the rate of relative-strength decay becomes significantly steeper as performance level rises. World-class athletes at the 99th percentile lose relative strength with each kilogram of body mass considerably faster than intermediate lifters at the 50th percentile. The authors suggest a likely explanation rooted in training physiology: muscle hypertrophy follows a linear-log trajectory, so advanced lifters experience diminishing returns in muscle accretion. When elite athletes gain weight, a larger proportion of that new mass may be non-contractile tissue, bone, connective tissue and fat, that inflates the denominator of the strength-to-mass ratio without contributing to the force-producing numerator.

This mechanism connects back to one of biology’s oldest mathematical constraints, the Square-Cube Law. As a body scales up, its mass increases with the cube of its linear dimensions, while the cross-sectional area of muscle, the tissue that actually generates force, increases only with the square. Heavier athletes are therefore structurally constrained to lower strength-to-mass ratios than lighter athletes of equivalent training status, a principle long established in the biomechanics literature. The observed declines in this dataset often exceed the theoretical predictions of pure geometric scaling, and the authors point to the fat-free mass index ceiling as a contributing factor: as lifters approach their natural hypertrophic potential, further mass gain becomes progressively less functional.

The study also documented a systematic shift in what the authors call the internal lift ratio, the percentage contribution of each lift to the competitive total. In the lightest weight classes, the deadlift dominates, contributing roughly 42 to 45 percent of the total. As body mass increases, that dominance erodes: in the heaviest male categories, the gap between deadlift and squat contribution narrows from an eight-percentage-point difference to roughly two points, at approximately 39 versus 37 percent. Regression models indicated that each 10-kilogram increase in body mass is associated with about a 0.44 percentage-point reduction in the deadlift’s share and a 0.28 percentage-point increase in the squat’s share. The likely culprit is soft tissue interference: greater abdominal and thigh girth can compromise the deadlift starting position or lockout, while added body volume may assist the squat through passive joint stability and a reduced effective range of motion.

Sex differences added another layer of complexity. Female athletes exhibited a steeper mass-dependent decline in bench press and squat performance than males. At the 99th percentile, the female squat decay coefficient was -0.0136 against the male -0.0125, meaning that for every 10 kilograms of body mass, a world-class female lifter loses 0.136 times bodyweight in relative squat strength compared with 0.125 for her male counterpart. The female bench press proportion of the total remained essentially static across the mass spectrum. The authors note this upper-body discrepancy may reflect both biological differences in regional muscle distribution and sociological factors, such as lower prior exposure to upper-body training among women, citing meta-analytic evidence that females make larger relative upper-body strength gains in response to training. The practical implication is blunt: applying male-derived multipliers to female athletes sets targets that simply do not correspond to the observed female performance distribution.

To make the findings usable on the gym floor, the researchers converted their statistical models into absolute-load normative atlases, stratified by sex, lift and current IPF weight class, expressed in raw kilograms with no calculation required. A coach can now determine whether an 83-kilogram athlete’s 180-kilogram squat sits at the 50th or the 90th percentile of drug-tested competitors worldwide, and whether the deadlift and bench press sit at comparable percentiles, something no aggregate score like the Wilks or IPF Good Lift coefficient can reveal, since those formulas compress the three lifts into a single number optimized for cross-weight-class ranking. The atlases also offer an empirical anchor for weight-class transitions: because relative multipliers fall even when absolute strength is maintained, an elite athlete moving up a class faces a disproportionately steeper relative decline than an intermediate lifter making the same move. The authors caution that the atlases are current reference values rather than permanent benchmarks, noting that squat and deadlift medians can rise by roughly 2.5 to 7.5 kilograms within five-year periods, and recommend periodic re-derivation. They also acknowledge limitations, including weigh-in manipulation through weight cutting, the absence of anthropometric data such as limb lengths, and the possibility that name-based athlete matching allowed a small number of duplicate records. Still, the message to the strength community is clear: the era of one-size-fits-all bodyweight multipliers is over, and the data have finally caught up with the barbell.

Subject of Research: Percentile-based normative standards for assessing relative and absolute strength across body mass categories in competitive classic powerlifting

Article Title: Percentile-Based Normative Standards for Strength Assessment Beyond Linear Bodyweight Multipliers in Competitive Powerlifters

Article References: Montenegro, S., Wicker, P., Wiedenmann, T., Rappelt, L., & Donath, L. (2026). Percentile-Based Normative Standards for Strength Assessment Beyond Linear Bodyweight Multipliers in Competitive Powerlifters. Sports Medicine – Open, 12(1), Article 136. https://doi.org/10.1186/s40798-026-01111-z

Image Credits: AI Generated

DOI: 10.1186/s40798-026-01111-z

Keywords: powerlifting, bodyweight multipliers, normative standards, quantile regression, relative strength, squat, bench press, deadlift, weight classes, sexual dimorphism, strength and conditioning, IPF