How to choose a progression scheme for every lift you log

You finish three sets of lateral raises with the 8 kg dumbbells, twelve reps each, and the next pair on the rack is 10 kg. That is a 2 kg jump on an 8 kg weight. The squat you did an hour earlier moves in 2.5 kg steps on 140 kg, a jump you barely feel. The same rule cannot progress both lifts. So choose a progression scheme per lift, not per program, from three facts you already know: the rep range, the smallest load step, and whether you rate effort.

The short version: a lift with a real rep range and a small step gets Double. A lift you train at a fixed rep target gets Load. A heavy compound you rate by RPE gets Top set + back-off. A small isolation lift with a chunky step gets Reps. A block where effort should climb week by week gets RIR. Anything you want to drive yourself gets Manual. If you would rather not decide, Auto makes the first two calls for you, and the rest of this guide shows exactly how.

The three questions that pick a scheme

Three facts about a lift decide its scheme, and you know all three before you start the session.

The first is the rep range. A range like 8 to 12 leaves room to add reps at one load before adding weight. A fixed target like 5 does not. That single fact is the first thing Auto reads.

Next comes the smallest load step. The app's default step is 2.5 kg for a barbell compound lift, 2 kg for a dumbbell compound lift, and 1 kg for a machine or a loaded bodyweight lift, all computed with the app's model. Plate maths decides what a barbell can actually hold, and the plate calculator will show you for any target. A step that is tiny against the working weight can be taken often. A step that is a large share of the working weight needs reps banked first.

Last comes effort: do you rate your sets? Schemes that read RPE or RIR (reps in reserve, ten minus RPE) can tell a hard 5 from an easy 5 and adjust. Schemes that read only reps cannot. Autoregulation, adjusting training to measured performance or perceived capability on the day, has been a training idea since the 1940s, and a 2020 review found the research on it scattered across specific practices and held back by inconsistent terminology [1]. If you log RPE honestly, the effort-based schemes have something to work with. If you do not, pick a scheme that reads reps.

None of the three is about which method builds more muscle. Progressing reps at a fixed load and progressing load at fixed reps produced similar results in an eight-week trial in 43 trained lifters, with quadriceps growth slightly favouring reps and strength slightly favouring load, both by margins the authors called of questionable practical significance [2]. A meta-analysis of 13 studies found linear and daily undulating periodization grew muscle about equally [3]. The scheme is a rule for deciding the next weight. Pick it for the lift, not for a promised result.

All seven schemes, side by side

Each scheme answers one question differently: what has to happen in a session before the load goes up. The table puts the rules in one place.

Double progression holds the load and adds reps until every working set reaches the top of the range, then adds the step and resets to the bottom. That default trigger is all sets. You can loosen it to the average of the sets or to the best single set, which progresses sooner and with less proof. The double progression guide runs six weeks of it on a barbell row.

Top set + back-off anchors on the heaviest set. When that set carries an RPE, the next top set is worked out from how hard it was against the target. Back-off sets follow at 88 per cent of the top set by default, and they move whenever the top set does. A top set logged as a grinder, well above its target RPE or clearly short on reps, holds the load and does not drag the back-offs down with it. The top set and back-off guide covers how many back-offs to do.

Load is plain linear progression. Hit the target reps on every set and the step goes on. Miss, and the load repeats. Miss at the same load across sessions and the scheme takes the step back off and suggests smaller jumps or a switch to Double. It is the simplest rule and the first one to break, because a fixed rep target leaves no buffer when a session goes badly.

Reps holds the load and adds one rep at a time toward a ceiling, then adds the step and resets. If you give it a fixed target with no range, it opens a ceiling five reps above that target so it has room to work. That makes it the right home for lifts where the step is clumsy.

RIR ties progression to effort across a block. By default it ramps from 3 reps in reserve in the block's first week to 0 in its last. If a set comes in well under the week's effort, the load goes up. If it comes in well over, the load holds so you do not burn out mid-block. When a set is logged without RPE, it falls back to a rep rule so the block still moves.

Manual is never autoregulated by anything: no suggestions, no stall flags, no deload maths. Use it for a peaking week or a test day, or for a plan someone else wrote.

Auto, with the rules spelled out

Auto is not a separate algorithm. It picks Double or Load from the shape of the rep range, then adjusts that choice with your estimated 1RM trend over recent sessions.

The first rule is the range. If the top of the range is at least two reps above the bottom, Auto runs Double progression with the all-sets trigger. If the range is narrower than that, or a single number, it runs Load. So 8 to 12 becomes Double and a straight 5 becomes Load.

Then Auto reads the trend. If your estimated 1RM on the lift has been sliding across sessions, Auto turns an increase into a hold and tells you to consolidate. If the trend is climbing fast and your hardest set was at least two RPE points easier than the target, it takes a double step instead of a single one.

Pace is the last rule, set per exercise as conservative, standard or aggressive. Conservative never takes the double step. Aggressive switches the Double trigger from every set to the average of the sets, so one short set no longer blocks an increase.

Auto never chooses Top set + back-off or RIR. Both imply a set structure or an effort plan you have to write on purpose, so Auto leaves them for you to pick. That is the main reason to step off Auto: you want a top set, or you want effort to ramp across a block.

A worked example on a 140 kg squat

Take a 90 kg lifter squatting on Top set + back-off with a target of 5 reps at RPE 8. Last session's top set was 140 kg for 5 at RPE 8. In the app's model that set is worth an estimated 172.6 kg for one, computed with the same model behind the one-rep max calculator.

The next top set depends on the RPE logged. Computed with the app's model, the scheme's suggestion for each rating the effort control offers looks like this.

An easy set earns a bigger jump, a set on target earns one 2.5 kg step, and a hard set repeats. The RPE 10 row is the one worth noticing. The general next-weight engine would cut that load to 132.5 kg, but the top set scheme refuses to drop the back-offs off one bad day. It repeats 140 kg and asks you to check recovery instead.

The same set logged without any RPE tells a different story. The model treats it as RPE 10, the assumption every classic formula makes, and the estimate falls to 162.2 kg. The scheme also loses its effort signal and falls back to a plain rep rule. If you squat on Top set + back-off, rate the top set every time.

Research on RPE-based loading supports the idea without proving it is better. In an eight-week study of 21 trained men, a group that set loads by RPE gained 17.1 kg on the squat and a group on fixed percentages of 1RM gained 13.9 kg, a difference that was not statistically significant [4]. RPE did no worse, and it adapts on the days you are under-recovered.

Now the lateral raise from the opening. Three sets of 12 with 8 kg dumbbells on Double with a range of 8 to 12 would trigger an increase to 10 kg. The step is 2 kg, which is more than a tenth of 8 kg, so the scheme reads it as a chunky jump and asks you to bank reps beyond the range first. Put the same lift on Reps with a ceiling of 15, and it does that by design: hold 8 kg, chase 13, 14 and 15, then step up to 10 kg and reset to the bottom of the range.

Matching schemes to the lifts in a week

A typical week mixes all of these, and each lift can carry its own scheme. A program row stores the scheme next to the sets and reps, so the choice travels with the lift through the block.

Heavy barbell compounds in the 3 to 6 rep range suit Top set + back-off if you rate effort and Load if you do not. The step is small against the working weight, and the low reps leave little room for a rep range to grow.

Compound lifts in the 6 to 12 range, barbell or dumbbell, suit Double. Heavy and light loads build muscle about equally when sets are taken hard, while heavier loads build more one-rep max strength [5]. So a hypertrophy range is a reasonable home for most compound accessories, and Double gives it a stopping rule.

Machine work suits Double or Auto. A 1 kg step on a stack is rarely chunky, so the default rules fire cleanly.

Small isolation lifts with dumbbells, cables or light plates suit Reps. Raises, curls, rear-delt work and calf raises all have steps that are a large share of the load.

Use RIR when the block itself is the plan: a hypertrophy block where weeks one and two sit at 2 or 3 reps in reserve and the last week goes close to failure. Use Manual for anything you are peaking or testing.

Rules every scheme shares

Some rules sit above the scheme and override it. Knowing them saves you from thinking a scheme is broken.

A planned deload week never emits an increase, whatever the last session said. More than 30 days off a lift resets it, so the first session back starts lighter. On a cut, a load increase never leaks through. The scheme turns it into a hold so you keep strength while losing weight.

Stalls are flagged after three sessions at the same load without progress, by default. A rising estimated 1RM at a fixed load does not count, because that is progress in reps or effort. One flag means hold and troubleshoot. A stall that persists past that, or a trend that is falling, gets a small deload suggestion, never applied to the plan without you. The stall guide walks through what to check first, and the deload guide covers how to run one.

Mistakes that make a scheme look wrong

The most common one is running an effort scheme without effort data. Top set + back-off and RIR both fall back to rep rules when a set has no RPE, so they behave like a slower Load. If you will not rate sets, choose Double or Load and the app will do exactly what the reps say.

A range too narrow for Double is next. A range of 10 to 11 is one rep wide, so Auto treats it as fixed and runs Load. If you want rep progression, give the range at least two reps of room.

Switching schemes after one bad week is another. Every scheme holds on a missed session before it reduces anything, and the stall rule needs three sessions. Changing the scheme resets that evidence. Give a scheme a block before you judge it.

Then there is the wrong step for the equipment. If your gym's dumbbells go up in 2.5 kg pairs or your machine stack jumps 5 kg at a time, set the exercise's increment to match in the program editor. A scheme can only suggest the next weight you can actually load if it knows the step.

Questions

Which progression scheme should a beginner use?

Auto, or Double if you want to choose. A beginner adds strength fast, and Double's rule is easy to follow at the rack: hold the weight, add reps until every set reaches the top of the range, then add the smallest step. It needs no effort ratings to work. Once you rate RPE reliably and train heavy compounds at low reps, Top set + back-off becomes worth trying on those lifts.

Can I use different progression schemes in one program?

Yes, and you usually should. The scheme is stored per exercise, not per program, so a squat can run Top set + back-off, a dumbbell row can run Double and a lateral raise can run Reps in the same session. The shared rules for deload weeks, layoffs, cuts and stalls apply to every one of them, whichever scheme each lift uses.

Does Auto progression use RPE?

Partly. Auto runs Double or Load, which decide on reps, so it progresses without any ratings. When you do log RPE, Auto reads it in two places: a session where the hardest set was well under target can earn a double step if the trend is climbing, and a falling estimated 1RM, which RPE makes more accurate, turns an increase into a hold.