Reps prescribes your next set from one rule set, and this page is that rule set in full. Hit every rep at a weight and the next session goes up one step. Miss reps and the weight repeats with the number to beat. Log a grind and it holds. Log pain and it drops. Take time off and it ramps you back. Stall three times and it resets you below the wall.
No page in this niche publishes its own math, so here is ours. Every rule below is the shipping behavior, not a simplification of it.
The default: double progression
Two variables move, one at a time. Reps climb to the target, then weight takes one step and reps start over. That is double progression, and it has been the backbone of intermediate lifting programs for decades because it progresses on performance instead of the calendar.
In Reps the rule reads exactly like this:
| What happened last time | What you get next |
|---|---|
| Every set hit the target reps | One grid step up, same reps |
| Any set missed the target | Same weight, with last time's number named |
| No history for this lift | The program or catalog weight, verbatim |
The step size is the grid, not a percentage. Pound-unit lifters step in 5 lb; kilogram lifters step in 6 lb, which lands on one 2.5 kg plate pair. A percentage would hand you 233.7 lb; a grid hands you a bar you can actually load.
When Reps moves you up it names the reason: Moved up — last time you owned 260 lb x 6. When it holds you it names the target: Beat last time: 260 lb x 6. A prescription without its reason is a command, and lifters correctly ignore commands they can't audit.
The effort gate
Double progression alone has a known failure mode: it will happily add weight to a set that was already a maximal grind, because it only sees reps. Reps reads your logged effort before it adds anything.
After a lift, the app offers four chips: Easy, Solid, Grind, Failed. They map to 4, 7, 8, and 10 on the RIR-anchored RPE scale Zourdos and colleagues tested, where RPE 10 means zero reps left in reserve, RPE 9 means one, and so on down.
Four choices instead of ten is deliberate, and the evidence for it is about where self-rating actually works. Hackett and colleagues had 81 lifters call their remaining reps and then take the set to failure to check. Estimates landed within about one rep when the lifter was zero to five reps from failure, and drifted past two reps of error when they were seven to ten away. Training experience did not change that.
So effort rating is trustworthy near failure and unreliable far from it, for everybody. A ten-point scale spends most of its resolution in the region where the reading is noise. Four chips put the resolution where the signal is: three of them describe the near-failure end, and one coarse bucket covers easy, because a lifter five reps from failure cannot honestly tell you whether that was a 4 or a 5.
The gate reads the average of your recent efforts on that lift:
| Effort trend | Decision |
|---|---|
| 7 or below | Progress allowed |
| Above 7, below 9 | Hold, repeat the weight |
| 9 or above, or pain flagged | Back off one step |
| No effort logged | Progress allowed (the historic behavior) |
The last row matters. If you never touch an effort chip, Reps behaves exactly like plain double progression. The gates only ever take away a jump you hadn't earned; they never invent one.
One caveat the app cannot see: effort readings are only as good as the rest that preceded them. Rush a heavy set at 45 seconds and you will honestly log a grind, and the engine will honestly hold you at a weight you could have moved. That is why the rest defaults matter to progression and not just to comfort.
Pain is a separate trigger from effort. Any set note containing words like pain, tweak, strain, or elbow flags the lift and backs the weight off regardless of what the reps said. A lifter who types "elbow hurt on set 3" and gets handed five more pounds has been failed by the software.
The comeback ramp
Two weeks off does not cost you your strength. A month starts to. Mujika and Padilla split their detraining review into two papers for exactly that reason: one on short interruptions, where strength is largely preserved, and one on longer layoffs, where the losses become real.
Reps encodes the split per lift, measured from that lift's own last-trained date rather than a global "last workout" date, because you might have benched three days ago and not squatted in a month.
| Days since this lift | What you get |
|---|---|
| 14 or fewer | No reduction |
| 15 to 28 | Minus 10 percent, one session |
| More than 28 | Minus 20 percent, then minus 10 percent, then normal |
The two-step ramp after a long layoff exists because the failure mode here is not physiological, it's psychological. The number you did in June is still in your head, and matching it on day one of your return is how people get hurt. The ramp gives you two sessions of runway to a weight your connective tissue agrees with.
The three-stall reset
Three consecutive sessions at the same weight without hitting the target reps is not a bad day. It's a wall. Reps drops you two grid steps below it and tells you why: Reset — run it back up past 260 lb.
Two steps, not one: one step is inside the noise of a bad night's sleep, and you'd be back at the wall next session with nothing changed. Two steps buys a real run-up. Reps counts stalls per weight, so a stall counter resets the moment you break through.
When rules collide
More than one reduction can trigger at once. You can come back from three weeks off, into a stall, having logged a grind. When that happens, the lightest prescription wins.
The reasoning is asymmetric risk. Prescribing too little costs you one sub-optimal session. Prescribing too much costs you a strain and three weeks. When the signals disagree, the cautious one is the one to trust.
A reduction that can't reduce still never advances. If you're already at the lightest bar on the grid, or the percentage rounds back to where you started, the prescription degrades to a hold. It keeps its reason and its note, but it will not hand you a jump through the back door.
Bodyweight lifts
Pull-ups and dips have no plates to add, so the same rules run on reps instead of weight. Hit the target and the next session asks for one more rep. Miss it and the target repeats. A grind holds the number where it is, and a pain flag drops the target by two reps.
One deliberate limit: rep-only progression runs only when the lift carries no load. The moment you clip on a 25 lb belt, Reps treats it as a loaded lift with no calibrated history and gives you the program's weight verbatim, because your bodyweight rep record tells it nothing about what you can do with a plate hanging off you.
What Reps does not do
It does not autoregulate mid-set, prescribe percentages of a tested 1RM, periodize your training blocks, or predict a date for your next PR. It looks at what you actually lifted, applies the rules above, and hands you a number for next time.
Every rule here is deliberately arguable. If your coach runs 3 percent jumps on a stall reset instead of two grid steps, they may well be right for you. You can edit any prescription in the app, and the next session progresses from what you actually did, not from what the app suggested.
How does Reps decide when to add weight?
When every set of a lift hits its target reps and your logged effort is 7 or below on a 10-point scale, Reps adds one grid step: 5 lb for pound lifters, 6 lb for kilogram lifters, which is one 2.5 kg plate pair. If you missed reps, the weight repeats with last session's performance named as the number to beat.
What happens if I log a set as a grind?
An effort trend above 7 holds the weight where it is rather than progressing. An effort trend of 9 or above, or any set note containing pain words, backs the weight off one grid step. If you log no effort at all, Reps progresses normally on reps alone.
How much weight does Reps take off after time off?
Nothing for two weeks or less. Between 15 and 28 days off that lift, it prescribes 10 percent less for one session. After more than 28 days it runs a two-session ramp: 20 percent lighter, then 10 percent lighter, then back to normal progression. The gap is measured per lift, not per workout.
What does Reps do when I stall?
Three consecutive sessions at the same weight without hitting the target reps triggers a reset: two grid steps down, with a note naming the weight to run back up past. Two steps rather than one, because a single step is inside the noise of a bad night's sleep.
Can I override what Reps prescribes?
Yes. Every prescribed weight and rep count is editable in the app, and the next session's prescription is built from what you actually lifted, not from what the app suggested. The rules describe a default, not a contract.
