Deep divefrom Episode 6Training10 October 202619 min

Stop the Set Early? Speed, Strength and Female Athletes

A deep dive into one paper: what was known before, what the researchers did, what the numbers say and what they don't.

The paper: Effects of 10% Versus 30% Velocity Loss Thresholds ... Trained Female Collegiate Athletes (J Sports Sci Med)

Chapters

Transcript

The script the two AI voices read from. The times are worked out from the script, not heard, so a line can start a few seconds early or late.

Read the whole episode

Cold open

Host 1You heard the trailer — this is the whole story. Welcome to the Hastrup Review of Science — where artificial intelligence reads real science, so you don't have to pretend you understood the paper. One paper today: twenty-eight female athletes, one squat machine, and a simple question about when to stop a set.

Host 2The paper is called Effects of ten percent versus thirty percent velocity loss thresholds on strength and explosive performance in trained female collegiate athletes. It is by Xuehaiyue Lv, Wei Han and colleagues at Shandong Sport University in China, published in the Journal of Sports Science and Medicine and accepted in September.

Host 1And the claim, in one sentence?

Host 2Stopping squat sets early, when bar speed had dropped by ten percent, gave slightly better sprint and long-jump gains, while pushing on to a thirty percent drop gave a bigger gain in squat strength. Both groups improved on everything they measured.

Host 1We will get to how big those differences really are. First, why does anyone measure bar speed at all?

Why measure bar speed

Host 2For most of the history of strength training, the dose was described by load and repetitions: three sets of ten at seventy percent of your one-repetition maximum. The problem is that the same prescription is easy on a good day and brutal on a bad one, and two people at the same percentage can have very different reserves.

Host 1So the number on the bar does not tell you how hard the set really was.

Host 2Right. As established background, coaches have long tried to fix that with effort scales, like how many more reps you think you could have done. Velocity-based training offers something more objective: a sensor on the bar that measures how fast you push it up, every single repetition.

Host 1Why would speed tell you anything about effort?

Host 2Because when you lift with full intent, the bar slows as you fatigue. A research group around Luis Sánchez-Medina and Juan José González-Badillo in Spain showed in twenty eleven that the percentage drop in speed within a set tracks how much fatigue has built up. That drop is called velocity loss, and this paper uses their formula.

Host 1Walk me through the formula.

Host 2Take the fastest rep in the set, usually the first. Velocity loss is how far the current rep has fallen below it, as a percentage. A ten percent threshold means you rack the bar the moment a rep is ten percent slower than your best one.

Host 1So the threshold is really a rule for when to stop.

Host 2That is exactly how the authors frame it. A low threshold gives you a few fast reps and stops well short of failure. A high threshold lets you grind out many more reps, with the later ones noticeably slow.

What was known

Host 1What did we know before this paper?

Host 2Quite a lot, in men. The paper cites a twenty twenty-three systematic review and meta-analysis by Ivan Jukic and colleagues, which found that maximal strength gains are often similar across thresholds. Higher thresholds tend to go with more muscle growth, and lower ones with better jumping and sprinting.

Host 1So stopping early costs you little strength.

Host 2Often, yes. Studies that roughly matched the total number of reps, by Myrholt in twenty twenty-three and Andersen in twenty twenty-four, found similar strength gains at low and high thresholds. And a dose-response meta-analysis by Zhang and colleagues suggested fifteen to thirty percent velocity loss, at seventy to eighty percent of max, as a generally favorable setup for one-rep max.

Host 1And in women?

Host 2Much thinner. The authors cite two training studies. Rissanen and colleagues in twenty twenty-two compared twenty and forty percent loss in men and women over eight weeks, and found no significant difference between thresholds, though effect sizes in the women leaned toward forty percent for some strength outcomes.

Host 1And the second?

Host 2Kambara and colleagues, published this year, compared ten and twenty percent loss in trained female basketball players over eight weeks of squats. Again, no significant group differences on strength, jumps or sprints.

Host 1So there is a gap.

Host 2A specific one. The evidence in women is small, and the threshold contrasts studied so far were fairly close together. Nobody had compared a very conservative ten percent against a fairly demanding thirty percent in trained female athletes.

The open question

Host 1So the open question here is?

Host 2Whether two thresholds, applied to the same exercise, load and number of sets, push trained women toward different kinds of adaptation. The authors' hypothesis, stated up front, was that ten percent would favor the fast, explosive tests and thirty percent would favor maximal strength.

Host 1A directional prediction, before the data.

Host 2Yes, which is to their credit. The study was not formally preregistered as far as the paper says, but the hypothesis is stated clearly.

Who took part

Host 1Who were the athletes?

Host 2Twenty-eight female collegiate athletes: six basketball players, five volleyball, six tennis, four track and field and seven rugby. To get in, they needed at least two years of structured strength training, a machine squat of at least one point two times body mass, and a top-three finish at a provincial-level competition in the past year.

Host 1What do they look like on paper?

Host 2About twenty years old, around one hundred sixty-eight centimetres tall, and a little over sixty kilograms. Their starting squat max averaged roughly seventy-nine kilograms in one group and eighty-one in the other, with just under three years of training experience.

Host 1Is that elite?

Host 2No, and the authors say so. Using a published classification framework, they put the sample at tier two: trained or developmental, not national or international level.

Host 1And twenty-eight people was the plan?

Host 2It was the minimum from their power calculation, assuming a medium effect size, eighty percent power and a correlation of zero point six between tests. They randomized one to one with a computer-generated sequence, fourteen per group, and the results tables still show fourteen per group, so no one appears to have dropped out.

The training

Host 1What did the training look like?

Host 2Eight weeks, two supervised sessions a week, on Tuesdays and Fridays, sixteen sessions in all. Every session was four working sets of squats on a Smith machine, the kind where the bar runs on fixed rails, at seventy percent of each athlete's max, with four minutes of rest between sets.

Host 1So the only thing that differed was?

Host 2When each set ended. One group stopped as soon as velocity had dropped ten percent from the fastest rep, the other at thirty percent. Everything else was the same, and every rep was meant to be pushed up as fast as possible.

Host 1How did they measure the speed?

Host 2With a GymAware linear position transducer, a cable sensor attached to the bar. The measure is mean propulsive velocity, the average speed during the part of the lift where the bar is actually accelerating upward, not coasting at the top.

Host 1And the lifters could see it?

Host 2Both groups got real-time velocity feedback on every rep, plus standardized verbal encouragement. That keeps effort high, and because both groups had it, it cannot by itself explain a difference between them.

Host 1How did they set seventy percent?

Host 2From an individual load-velocity profile. Each athlete lifted increasing loads, the speed at each load was recorded, and a straight line through those points gave the speed that corresponds to seventy percent of her max.

Host 1And that target speed was then used every session.

Host 2Yes. If the first rep of a session was more than zero point zero four metres per second off the target, they adjusted the weight, so the relative load stayed the same as the athletes got stronger.

Host 1Anything else going on in their lives?

Host 2No other lower-body strength training was allowed, but they kept up their normal sport practice and conditioning. That was not standardized, and it was a general preparation phase of the season.

The tests

Host 1How did they measure the outcomes?

Host 2Five tests before and after. Squat one-rep max on the same Smith machine, a squat jump from a static ninety-degree position, a countermovement jump with a quick dip first, a standing long jump, and a thirty-metre sprint.

Host 1Equipment?

Host 2The vertical jumps were on a contact mat, the long jump with a steel tape to the rearmost heel, and the sprint with infrared timing gates at zero and thirty metres on an outdoor track. Squat depth was controlled with an elastic band at each athlete's own ninety-degree height.

Host 1And the statistics?

Host 2For each outcome, a two-by-two mixed analysis of variance, group by time. The number that matters is the group-by-time interaction: did one group change more than the other?

Results: the training dose

Host 1Start with the training itself. Did the thresholds actually produce different sessions?

Host 2Very different. The ten percent group averaged six point three reps per set, about twenty-five per session. The thirty percent group averaged eleven point two per set, about forty-five per session.

Host 1So nearly twice the volume.

Host 2Nearly twice. And the average bar speed was zero point six zero metres per second in the ten percent group versus zero point five five in the thirty percent group. All of those differences had p-values below one in a thousand.

Host 1Keep that in mind, because it shapes everything that follows.

Results: strength

Host 1Strength.

Host 2Both groups got stronger. The ten percent group went from seventy-eight point nine to eighty-six point two kilograms, up nine point two percent. The thirty percent group went from eighty point nine to ninety-one point zero, up twelve point five percent.

Host 1So about seven kilograms against ten.

Host 2Yes, a difference in gain of just under three kilograms. The interaction was significant, p below zero point zero zero one, with a partial eta squared of zero point four five, which by the usual convention counts as large.

Host 1Large sounds big.

Host 2We will come back to that word. Expressed against how much athletes differed from each other at baseline, the authors' own standardized effect for strength is a Hedges' g of about zero point two three, which is small.

Results: jumps and sprint

Host 1Jumps.

Host 2Both vertical jumps improved in both groups. Countermovement jump went from thirty-five point one to thirty-eight point six centimetres with ten percent, and thirty-five point four to thirty-eight point one with thirty percent. The interaction was not significant, p equals zero point zero five eight.

Host 1Close, though.

Host 2Close, and the authors handle it well: they say it is insufficient evidence of a difference, not evidence of no difference. The squat jump was further from significance, p equals zero point one two seven, with gains of about two and a half versus two centimetres.

Host 1And the long jump?

Host 2Here the ten percent group came out ahead. It went from one hundred ninety-five point four to two hundred one point seven centimetres, the thirty percent group from one hundred ninety-four point seven to one hundred ninety-nine point six. Interaction p equals zero point zero three one.

Host 1In centimetres, the gap in improvement is?

Host 2About one and a half centimetres, on a jump of roughly two metres. The standardized effect is a Hedges' g of zero point one two.

Host 1And the sprint?

Host 2Thirty-metre time dropped from five point five two two to five point four six three seconds with ten percent, and from five point five zero three to five point four six nine with thirty percent. Interaction p equals zero point zero zero six.

Host 1So a difference in improvement of?

Host 2About two and a half hundredths of a second, roughly one percent versus zero point six percent. The standardized effect is a Hedges' g of zero point one three.

A word from our sponsor

Host 1Before the big picture, a quick word from our sponsor.

Host 1Does your hair still remember the gym, hours after you left? Then you are in luck. The Hastrup Review of Science is brought to you by dry shampoo.

Host 1A light powder spray absorbs oil at the roots in seconds, no water and no hairdryer required.

Host 2I use it after every heavy session. It gives me that lift without having to go all the way.

Host 1Available in tinted and untinted versions, widely in stock and very reasonably priced.

Host 2It keeps everything fresh between sessions, even when things get sweaty. A very dependable product.

Host 1Dry shampoo. Our referral link is in the show notes.

Why it matters

Host 1So, what is the principle here?

Host 2That a velocity-loss threshold is not a magic number. It is a stopping rule, and what it does depends on how many reps it lets you do, at what speeds, with which load and exercise.

Host 1Meaning ten percent and thirty percent are not speed and strength settings.

Host 2The authors say explicitly that assigning them to rigid speed and strength categories is the wrong reading. Under this protocol, ten percent at seventy percent load meant about six reps a set, and that was apparently not enough volume to maximize gains in the trained lift.

Host 1Which is a bit unusual, given the literature.

Host 2It is, and the authors point that out. Most earlier work found strength similar across thresholds, so their result is that a low threshold can cost you some strength when the volume it allows is this low.

Host 1And the sprint result is the other half of that.

Host 2Yes, and to me it is the most interesting point in the paper. The group that gained more squat strength gained less sprint speed, so a bigger strength gain did not automatically transfer to a faster sprint.

Host 1Who should care?

Host 2Strength coaches working with female team-sport athletes who already use velocity tracking. The practical version from the authors: thirty percent if squat strength is the immediate priority and the athlete can absorb the extra reps, ten percent if you want to keep reps fast, limit leg-training volume and support sprinting.

Host 1And what does it not mean?

Host 2It does not mean ten percent is optimal in general, which the authors say in their conclusion. It does not mean grinding to slow reps is good for strength in itself, because volume and fatigue were never separated. And it does not show a general advantage in explosiveness, since the two vertical jumps did not differ.

Host 1Nor anything about mechanisms.

Host 2No. They did not measure muscle size, nerve activation, force production or sprint mechanics, and they say so.

Caveats

Host 1Now the caveats.

Host 2The first is the gap between large and small. The partial eta squared values are called large, but they mostly reflect how consistently each group changed, measured against within-person noise. The authors' own Hedges' g values, which compare the difference in change to how much athletes vary, are zero point one two for long jump and zero point one three for sprint.

Host 1And those count as?

Host 2The paper's own forest plot shades everything below zero point two as trivial. So by the authors' own figure, the two explosive advantages that reached significance sit inside the trivial zone, and the strength difference is small.

Host 1Is two and a half hundredths of a second meaningful over thirty metres?

Host 2It could be in competition, but it is in the range where measurement matters. Sprinting was outdoors, wind and temperature were not recorded, and the authors list that as a limitation; as general background, single-beam timing gates also add some error of their own.

Host 1Second caveat?

Host 2Volume is confounded with the threshold. The thirty percent group did nearly twice the reps, and the authors say they cannot tell whether the strength gain came from the extra reps, the slower reps, higher effort or all three.

Host 1Third?

Host 2Specificity. They trained on a Smith machine and tested their max on the same Smith machine, so the group that did nearly twice as many reps also got twice the practice at the test. The authors flag this too.

Host 1Anything in the reporting?

Host 2One inconsistency. The confidence interval printed for the strength effect size runs from minus zero point three five to plus zero point one zero, which crosses zero, while the p-value is below one in a thousand; one of those is probably a typo. The figure also gives a slightly different raw gap in strength gain than the table.

Host 1Other limits?

Host 2Twenty-eight people from five different sports, whose sport training was not controlled. Menstrual cycle phase and contraceptive use were recorded but not controlled, and sleep and energy availability were not tracked.

Host 1And blinding?

Host 2The paper does not say whether testers knew who was in which group. As background, five outcomes were each tested at p below zero point zero five with no correction for multiple comparisons, so the two borderline-sized explosive results deserve some caution.

Host 1Anything reassuring?

Host 2Yes. The design is clean, the dose difference was documented session by session, the hypothesis was stated in advance, there were no apparent dropouts, and the authors' own discussion is unusually careful not to oversell. No funding and no conflicts were declared, and data are available on request.

What to watch

Host 1What should we watch for next?

Host 2A volume-matched version in women: same total reps, different thresholds, for example by adding sets to the ten percent group. That would separate the stopping rule from the amount of work.

Host 1And on the sprint side?

Host 2Split times and force-velocity testing, which the authors themselves call for, to see whether any difference sits in the start, the acceleration or top speed. And testing on a free-weight squat as well as the trained machine, to see how much of the strength gap was practice.

Host 1Anything bigger?

Host 2Replication with larger samples and with the menstrual cycle tracked properly. Taken together with the Rissanen and Kambara studies, this is the third small trial in women, and none of them yet shows a consistent pattern.

Host 1Bottom line?

Host 2A tidy, modest trial that shows a ten percent stopping rule at seventy percent load cost these athletes about three kilograms of squat gain and possibly bought them a little sprint and long-jump speed. Both approaches worked; the trade-off is real but small, and the volume difference may explain most of it.

Sign-off

Host 1That is the deep dive: Effects of ten percent versus thirty percent velocity loss thresholds, by Lv, Han and colleagues at Shandong Sport University, in the Journal of Sports Science and Medicine. The link is in the show notes.

Host 2Every Friday this same feed carries the roundup episode: five new papers as trailers, and you pick which become deep dives. This has been the Hastrup Review of Science.