how heavy is heavy enough.

Published September 24, 2026 training science 15 min read

the number on the dumbbell is not the gate. the last three reps are. here is the load range that builds muscle, the narrower one that builds strength, + the trials showing u cannot feel where your set actually ends.

why does every answer come with a percentage u cannot use

u are standing at the rack holding the 6s. the 8s are right there. u do not know if the 6s are a workout or a warm-up + the person next to u is doing something with the 20s that looks like it should be illegal.

so u search it. every answer says 70 to 85 percent of your one rep max. u have never tested a one rep max. u are not going to test one on a tuesday in a gym where the squat rack has a queue. the forum version is worse: if it is not hard u are not doing it right, which is true + also not a number.

i picked the wrong weight for years. not too heavy. too light, every session, because i could finish the set clean + finishing clean felt like the goal. it took a coach standing over me saying two more to find out my sets had been ending four reps before anything useful happened.

two camps sell u an answer here. the bro camp: go heavy or go home. the goat camp, the wellness accounts i actually rate: mike chang's 70 percent effort every day, be okay with being weak, flow before the program. troy casey walked out of the gym for good. kyle peche says walk instead of grind when u are depleted. eva says hard training dysregulates u.

the goats are right about more than the bros here, + they are wrong about the one thing that decides whether a set counts. the record says which is which.

does the weight itself even matter

for muscle size, less than everyone thinks. the 2017 meta-analysis pooled 21 studies that compared light loads [60 percent of max or less] with heavy [above 60], every set taken to failure. result: strength gains favoured heavy, muscle growth was the same in both.

the 2021 network meta-analysis went further. Lopez pooled 28 studies + 747 adults, split loads three ways [heavy = 8 reps or fewer, moderate = 9 to 15, light = more than 15] + found no difference in muscle growth between any pair, P values from 0.11 to 0.47. strength was a different story: heavy beat light by a standardised 0.60, moderate beat light by 0.34, + heavy edged moderate by 0.26, which did not reach significance.

then in 2023 the McMaster group ran the biggest one. Currier's Bayesian network pooled 178 studies + 5,097 people for strength + 119 studies + 3,364 people for muscle size, + roughly 45 to 47 percent of those people were women. loads above 80 percent of max maximised strength; every prescription promoted muscle growth comparably. the top-ranked recipe for size was multiple sets, twice a week, at a heavy load, but the gap to the other recipes was small enough that the authors said so.

the trained-lifter version is the one i send people who say the rules change once u are good. 49 trained men, 12 weeks, whole-body programme, one group doing 20 to 25 reps at 30 to 50 percent, the other 8 to 12 reps at 75 to 90 percent, everything to failure. lean mass + fibre size grew the same in both groups; strength was similar too, except bench press, where the heavy group added 14 kg to the light group's 9.

the verdict on load: for size, the weight is not the gate. for strength, heavy wins, + it wins by a margin worth planning around. the old repetition continuum, the strict zones where 1 to 5 reps meant strength, 8 to 12 meant size + 15 plus meant endurance, was re-examined by its own authors in 2021 + mostly retired.

so is there a weight that is too light

yes, + it is lower than the 8s-vs-6s problem. the study that found the floor is the one to remember.

thirty men trained one leg + one arm at 20 percent of max for 12 weeks, + the other limb at either 40, 60 or 80 percent, with total work matched so the light limb did more sets to make up the difference. everything to failure. thigh muscle grew 8.9 percent at 20 percent vs 20.5, 20.4 + 19.5 percent at 40, 60 + 80. upper arm: 11.4 percent vs roughly 25 at every other load.

in plain terms: a weight u could lift more than about 30 to 35 times in one go is under the floor, + no amount of sets fixes it. a weight u can lift 25 times or fewer is above it. that is the whole load rule for size. everything else is effort.

does this hold for women or is it 20-year-old men again

fair, because the load trials skew male. the women-only one exists + it is small but clean. 23 untrained women trained for 12 weeks to failure, 11 at 30 percent of max, 12 at 80 percent. leg-extension strength rose 32 percent, pulldown 25, leg curl 23, press 17, with no difference between the light + heavy groups on any lift.

honesty line: that trial measured strength + a DXA body scan, not muscle size directly, + neither group changed lean mass on DXA in 12 weeks. DXA is a blunt tool for 12 weeks of beginner training, so read it as light + heavy produced the same strength in women, not as a growth result.

for growth, the sex-difference meta-analysis is the receipt. across 10 studies that put women + men on the same programme, muscle growth did not differ between the sexes, effect size 0.07. relative upper-body strength gain actually favoured women, ES 0.60, P = 0.002. + the 2023 network above, with nearly half its 8,000-plus participants women, found the same load-does-not-decide-size result in the pooled data.

so the rules are not a men's thing. if anything the upper-body number says the shoulders + back u want respond faster than the internet told u.

if the weight is not the gate, what is

how close the set gets to the end. this is the variable the goats' 70 percent rule gets wrong + the one most people never measure.

the 2024 meta-regression took every failure + not-failure study, estimated how many reps in reserve each group stopped at [reps in reserve = how many more u could have done with the same form], + plotted the result against it. for strength, the line was flat across a wide range of reps in reserve. for muscle size, growth increased as sets ended closer to failure, + the confidence interval did not cross zero. the authors call it exploratory because the reps in reserve were estimated from study descriptions, so hold it loosely, but the direction is the direction every other line of evidence points.

going all the way to failure is not required. in 15 studies, true momentary failure vs stopping short gave a muscle-growth difference of 0.12, not significant; the pooled failure-vs-not effects were -0.09 for strength + 0.22 for size, neither significant, with a small edge for failure in trained lifters. the cleanest single trial: 18 trained lifters, 12 men + 6 women, one leg to failure, the other stopping 1 to 2 reps short for 8 weeks. quad thickness grew 0.181 cm vs 0.182 cm. identical, + the failure leg carried more fatigue for it.

so the target is 1 to 3 reps in reserve. not failure. not 70 percent. the neighbourhood of the end, on every working set, at any load above the floor. i keep mike's six-minute morning + i keep the walk. i do not keep his set rule.

can i actually feel where my set ends

this is the part nobody on page one tells u, + it is the reason the 2-in-reserve rule fails in practice. people cannot feel it.

the scoping review pooled 12 studies + 414 people who were asked, mid-set, how many reps they had left, then made to keep going. on average they stopped 0.95 reps early, with a confidence interval from 0.17 to 1.73 + enormous variation between studies. the largest single study had 141 people do full-body sets to failure after predicting: every group underpredicted, with a standard error of 2.6 to 3.4 reps. that is not a rounding error. that is a whole set's worth of growth signal left in the rack.

the accuracy depends on where u are in the set. in 81 adults, 28 of them women, the guess was off by about 1 rep when 0 to 5 reps from failure, + more than 2 reps when 7 to 10 out. the leg press was harder to judge than the chest press, women were less accurate than men at certain points on the leg press, + training experience did not improve the guess. knowing the exact load did not help either: 20 trained men guessed the same whether or not they were told what was on the bar, + the first set was always the worst guess, improving by the third.

the reversal in one line: the rep u think is your last is usually one to three reps before your actual last. when u stop at what feels like 2 in reserve, the trials say u are at 3 to 5. so honest 2-in-reserve means going past the point where quitting feels reasonable.

there is one cue the body cannot fake: bar speed. in 29 squatters, how fast the bar moved tracked how hard the rep felt at r = -0.88 in experienced lifters + -0.77 in novices. the last honest reps of a set are visibly slower than the first, even when u are trying to move fast. if the tenth rep looks like the first, u are not near anything. that is the cue i use on myself + on every person i coach, because it works in a mirror + it works on video.

how heavy, then. the protocol

u do not need a one rep max + u do not need heavier dumbbells. u need one honest test per exercise + a rule for what happens next. this is it.

above the floor, pick by the joint + the lift. below the floor, nothing saves it. on every row, end within 1 to 3 reps of failure.

a note for the person who reads all this + hears push harder: effort is not pain. a joint that hurts is a stop sign, a muscle that burns is a set. failure is a tool, not a moral, + a week where every set ends at 2 in reserve is more than most people lifting next to u are doing. mike's line stays: be okay with being weak. the 6s taken to the end are the same tissue signal as the 20s taken to the end. the person with the 20s just has a longer history of ending sets honestly.

questions i get at 1am

how do i know if i am lifting heavy enough

Count what happens at the end of the set, not the number on the dumbbell. If the last 2 to 3 reps visibly slow down + u could not have done more than about 3 more with the same form, the weight is heavy enough to build muscle. If u finish the set at the same speed u started, u stopped too early or the weight is too light. Across 28 studies muscle growth did not differ between light, moderate + heavy loads when sets were taken close to failure.

is it better to lift heavy or do more reps

For muscle size it does not matter, as long as the set gets within a few reps of failure + the weight is above roughly 30 percent of your max. For strength, heavy wins: in a 28-study network meta-analysis, heavy loads of 8 reps or fewer beat light loads of 15 or more by a standardised 0.60 on strength gains. Most people are best served by a mix, with the big lifts in the 5 to 10 rep range + the smaller ones in 10 to 20.

do i have to train to failure

No. In 15 studies, training to complete failure did not add reliable muscle growth over stopping a rep or two short, + in 18 trained lifters an 8-week comparison found identical quad growth at failure versus 1 to 2 reps in reserve. Stopping 1 to 3 reps short is the target. The catch is that people underestimate how many reps they have left, so an honest 2 in reserve usually means pushing further than feels natural.

how much weight should a beginner start with

Start with a weight u are sure u can lift 12 times with good form, then take one set until the reps slow + count. If u pass 15, add 2 to 10 percent next session, which is the official progression rule. If u cannot reach 6, take some off. In a 12-week trial of 23 untrained women, both 30 percent + 80 percent of max produced the same strength gains once every set went to failure, so the starting weight matters less than finishing the set honestly.

keep reading

• when am i supposed to add weight. — the 2 to 10 percent rule above, with the whole month written out
• what do i even do at the gym. — the first session, so the load test has somewhere to live
• why did i stop making progress. — sets ending 4 reps early is on that list, near the top
• how jacked can i actually get. — the ceiling, once the floor in this post is sorted
• am i going to hurt my back deadlifting. — for the reader who read heavy + felt her spine flinch

Published September 24, 2026 · sources are the published meta-analyses + trials named in each line; effect sizes + percentages are quoted as printed in the abstracts, with confidence intervals where the paper gave them.

this is education, not medical advice. nothing here diagnoses or treats a condition. if u have an injury, a heart condition, are pregnant, or are working with a clinician or physio on your training, they own the call on load + effort.