Progressive overload for hypertrophy: adding load, reps and sets

You rack the bar after a third set of eight at 70 kg and the question is already waiting: what changes next week? For muscle there are three honest answers. Add 2.5 kg, add a rep at the same effort, or add a set to the week. Progressive overload for hypertrophy is picking one of those on purpose, writing it down, and checking a month later that the number moved. Done that way, a 70 kg bench for eight becomes 72.5 kg for eight at the same RPE in four weeks, which is worth 3.4 kg on the estimated one-rep max.

The rest of this guide takes the three levers one at a time, then walks one lifter through the four weeks with every number computed with the app's model.

Which lever moves first?

Over an eight-week block, reps and load grow muscle about equally well, and load is the better lever for strength. The set is what remains when both have stalled. That order comes from three places.

Plotkin and colleagues took 43 lifters with at least a year of lower-body training and split them for eight weeks: one group added load and held reps, the other added reps and held load, four sets of four exercises twice a week [1]. Quadriceps and calf growth were much the same. The rectus femoris modestly favoured the rep group, by 2.8 mm summed across sites, while squat strength slightly favoured the load group, by 2.0 kg, and the authors called both differences of questionable practical significance [1].

The load picture holds across the literature. A meta-analysis of 21 studies with every set taken to failure found hypertrophy similar whether the load was above or below 60 per cent of 1RM, while one-rep max strength favoured the heavier loads [2]. Heavier loads make a lifter better at lifting heavy, which is what a top set is for and what a hypertrophy block is not.

Sets have a dose-response of their own. Across 34 training groups from 15 studies, each extra weekly set was associated with 0.37 per cent more growth, and the higher-volume arm of a study beat its lower-volume arm by 3.9 per cent [3]. The same analysis could only show a trend when it split volume into fewer than 5, 5 to 9 and 10 or more weekly sets per muscle, so where the curve flattens is not settled by that paper [3]. Treat ten weekly sets per muscle as a floor with evidence behind it, and anything above it as a lever to pull slowly.

Lever one: load, in steps the bar can take

The smallest useful step on a barbell is the smallest pair of plates in the gym. The app's default increment for a barbell compound lift is 2.5 kg, computed with the app's model. A dumbbell compound steps by 2 kg and a machine by 1 kg. A 2.5 kg step is a 1.25 kg plate on each side, so 72.5 kg loads as one 25 and one 1.25 per side on a 20 kg bar. The plate calculator will do that sum for any target.

A load step counts as progress only when the reps and the effort stay put. Helms and colleagues had fifteen powerlifters work up to a one-rep max on squat, bench and deadlift while rating every set, and found RPE tracked percentage of 1RM with correlations of 0.88 to 0.91 on each lift [4]. In plain terms: eight reps at RPE 8 on 72.5 kg is a heavier eight at RPE 8 than it was on 70 kg, and that is the whole of the claim. Eight reps on 72.5 kg at RPE 9.5 is a different sentence.

The step also has to fit the lift. On a 40 kg overhead press, 2.5 kg is a far bigger relative jump than it is on a 140 kg squat. The app checks that ratio before writing the next weight and, when the jump is large for the load, tells you to bank reps beyond the range first rather than load up.

Lever two: a rep at the same RPE

RPE is a 1 to 10 rating of how hard a set was, where RPE 10 means no reps left and RPE 8 means two in reserve. The reserve is the whole trick of the rep lever. A rep added while the RPE stays at 8 is new strength. A rep added by pushing from RPE 8 to RPE 10 is the same strength shown with more grinding.

The numbers make that concrete, all computed with the app's model. A set of 70 kg for 8 at RPE 8 estimates a one-rep max of 94.7 kg. Make it 9 at RPE 8 and the estimate is 99.0 kg, and 10 at RPE 8 is 102.9 kg. Now log 70 kg for 10 at RPE 10 and the estimate is 94.7 kg again, exactly where the set of eight started. The two extra reps were bought entirely with the two reps in reserve, and the model knows it.

This is why a set logged without RPE reads as a max. The model treats a missing RPE as RPE 10, the same assumption every classic formula makes, so 70 kg for 8 with no rating estimates 89.1 kg instead of 94.7 kg. Log the effort and the rep lever starts telling the truth. The one-rep max calculator shows the same conversion for any set.

Lever three: a set in the week

A set is the lever to pull when load and reps have both sat still for a few weeks and recovery is fine. Growth responds to weekly sets in a graded way [3], but every added set also has to be recovered from, and the day that last set shows up is usually the day the first three get worse. Add one set to one lift, run it for two or three weeks, and look at whether the earlier sets held their reps and RPE before adding another.

Count the sets the way the app counts them: weekly sets per muscle, across every lift that trains it, with warm-ups left out. Top Sets sets weekly targets for 14 muscles and shows the gap in plain English, so a chest set added on Monday shows up against the same target as the ones already logged on Thursday. Three sets on a bench press and three on a dip are six for chest, not three on each of two exercises.

Four weeks on the bench, in numbers

Take a 75 kg lifter benching three sets of 8 to 10 at RPE 8 on the Double scheme, which adds load once every set reaches the top of the range and resets the reps to the bottom. Every estimate below was computed with the app's model.

Two things in that table catch lifters out. The first is week 2: only the best set reached nine, so nothing happens to the load yet. By default the scheme waits for every set to reach the ceiling, and you can switch it to fire on the best set instead; either way, the rule is written down before the session rather than decided under the bar.

The second is week 4, where the estimate drops from 102.9 kg to 98.1 kg. That is not lost strength. The model rates 70 kg for 10 at RPE 8 above 72.5 kg for 8 at RPE 8, because ten reps with two in reserve sits further down the chart than eight with two in reserve. The honest comparison is week 4 against week 1, the same reps at the same RPE: 94.7 kg to 98.1 kg, a 3.4 kg gain in a month.

Had week 4 come in at RPE 9 instead, the estimate would be 95.1 kg, barely above week 1. The load moved, but the effort moved with it, and the scheme would hold 72.5 kg until the reps came back at RPE 8.

What the scheme does when the week goes wrong

Top Sets carries seven progression schemes, labelled Auto, Double, Top set + back-off, Load, Reps, RIR and Manual, and all but Manual write the next load from what you logged. The rules they share are the ones worth knowing before you trust the suggestion.

Three sessions without progress flags a stall by default. The first flag is a hold with a reason; if it persists, the suggestion is a small drop in load to get the lift moving again, never applied without you. A planned deload week never emits an increase, however good the sets looked. More than 30 days off a lift resets it rather than resuming where it left off. On a cut, a load increase never goes through; the scheme holds the weight and lets the reps and RPE carry the progress instead. If every set finished two or more reps past the ceiling, the next jump is two steps.

When every set misses the floor of the range once, the scheme calls it an off day and repeats the weight. When it happens twice in a row, it eases the load down by one step and rebuilds. All of it runs on reps and RPE logged on every working set, which is what the home page means by automatic progressive overload.

Mistakes that look like progress

Reps added by spending the reserve. A set of 10 at RPE 10 is a set of 8 at RPE 8 wearing a bigger number, and the model prices them the same. If the reps climb while the RPE climbs with them, nothing has changed except how tired you are.

Load added while the reps fall through the floor. Going from 70 kg for 10 to 75 kg for 5 changes the rep range instead of overloading it. Take the step the bar allows and keep the reps inside the range.

Sets added while the first sets get worse. If a fourth set appears and the first three lose a rep each, the week has the same volume and more fatigue. Add the set only when the existing ones hold.

Comparing the wrong sets. Your fourth set of the day estimates lower than your first, so compare sets from the same position in the session, at the same RPE, before deciding a lift has stalled.

Testing to prove it. A one-rep max attempt costs a session and tells you about one day. The estimate from your working sets is updated every session and shows the trend, which is the thing progressive overload is actually about.

Questions

Should I add weight or reps first for muscle growth?

Either works. Over eight weeks in trained lifters, adding reps at a fixed load and adding load at fixed reps grew the quadriceps and calves about the same, with a slight edge to reps for the rectus femoris and to load for squat strength [1]. Pick the one your bar allows: add reps on lifts where 2.5 kg is a big jump, add load where it is small.

How often should progressive overload happen?

Not on a fixed schedule, and not the same lever every time. On an 8 to 10 rep range a rep usually moves from one week to the next and the load moves when every set hits ten, which in the example above happened once in four weeks. A lift that has moved neither reps, load nor RPE in three sessions is stalled by the app's default, and that is the point to change something deliberately.

What if I cannot add 2.5 kg?

Add a rep instead, up to two past the top of the range, or add a set to the week. The app does the same: when the next load step is large for the weight on the bar it asks you to bank reps first. For dumbbell lifts the default step is 2 kg and for machines 1 kg, so the load lever is smaller there and the rep lever carries more of the work.