RPE has a working range: about 1 to 6 reps, RPE 7 to 9.5.

The set of fifteen you rated an 8

Third set on the leg press, fifteen reps, and you wrote 8 in the log.

You were close to right about how it felt. Quads cooked, breathing wrecked, two more reps would have been ugly. Then the tracker converted that 8 into a percentage, folded it into an estimated 1RM, and quietly told you something about your squat that isn't true.

The exertion was real. The math downstream of it wasn't.

Every RPE-to-load chart in circulation, the ones printed in the back of programming books and the ones baked into tracking apps, performs a conversion whose accuracy depends on where in the rep range you took the reading. That dependency almost never appears on the chart. You get one grid, one number per cell, the same visual confidence at 2 reps and at 20. So lifters carry a reading taken well outside the instrument's range straight into next week's load selection, and then wonder why autoregulation feels like it's drifting.

RPE is an instrument, and instruments have working ranges

A luggage scale reads honestly at 20 kg and lies at 300 g. It still shows you a number at 300 g. That's the whole failure mode: the readout doesn't blank out when you leave the range, it just gets quietly wrong while looking exactly as confident as before.

RPE behaves the same way. Call it the working range: roughly 1 to 6 reps at RPE 7 to 9.5, the zone where a lifter's reps-in-reserve estimate maps tightly enough to actual proximity-to-failure that converting it into a load is defensible. Go past six reps and the estimate spreads. Drop below RPE 7 and it spreads further. Take the set to genuine failure and the number stops carrying load information at all, because everything that fails reads the same.

Those boundaries are an editorial read, not a published finding. No paper draws a line at six reps. The direction of the evidence is clear enough to act on; the exact coordinate is a judgment call, and it's worth saying so out loud rather than dressing a convention up as a result.

The accuracy debate keeps auditing the lifter instead of the reading

Most arguments about RPE accuracy are really arguments about the person holding the clipboard. Are you sandbagging. Do you have enough training years to know what two in reserve actually feels like. Both are fair questions, both have been chewed to the bone in every lifting forum for a decade, and neither one touches the problem here.

The measurement condition matters more than the character of the person measuring. A ten-year lifter calling a set of twenty on the hack squat is working with a degraded instrument. A second-year lifter calling a heavy triple is working with a good one. No amount of honesty closes that gap, because at twenty reps the sensation being rated has stopped being mostly about proximity to failure and started being about burn, breathing, and how much discomfort you happen to tolerate today.

Which makes the standard chart a strange artifact. It publishes one conversion table across every rep range with no indication of where it was ever validated, and lifters treat the whole grid as equally load-bearing because nothing on it says otherwise.

What the research supports, and what it doesn't

Helms and colleagues published "RPE vs. Percentage 1RM Loading in Periodized Programs Matched for Sets and Repetitions" in Frontiers in Physiology in 2018. The design compared eight weeks of squat and bench training in trained lifters, one group taking loads assigned by RPE, the other by percentage of 1RM, with sets and repetitions matched between them. Strength and volume outcomes came out comparable between the two groups. RPE-based loading held its own against the percentage-based standard.

That paper gets cited constantly as blanket proof that RPE works. Read the program shape it ran in: heavy, low-rep, powerlifting-style loading. It's evidence that RPE-assigned load holds up where powerlifters actually use it. It is not evidence that an RPE 8 on a set of fifteen means anything comparable.

The rest of the picture comes from the reps-in-reserve estimation work, which is where the range boundaries come from.

One Wednesday session, read twice

Lower body day. Top set of three at 145 kg, called RPE 8.5. Three back-off sets of eight at 120, called 8, then 8.5, then 9 as the sets accumulated. Two sets of twenty on the leg press to finish, both rated 9.

Six entries, one log, three completely different measurement worlds.

The triple is clean. Heavy, short, inside the range, and its 8.5 is the closest thing to a real instrument reading you produced all day. The sets of eight sit right at the edge, and their RPE is doing double duty: by the third one, the rise from 8 to 9 is reporting accumulated fatigue across the block as much as it's reporting reps left in the tank at that load. The twenties are a reading of discomfort tolerance wearing an RPE costume.

Now pool all six into one estimated 1RM, which is what happens by default. The estimate moves. It moved because you did leg press, not because your squat changed. Do that for a training cycle and the drift you're autoregulating against is partly an artifact of your own accessory work.

Tag the sets that fall outside the range

The fix is not to stop logging RPE on high-rep work. That number is real information about fatigue, and a hypertrophy block without it is blind. The fix is to stop letting one field carry two jobs.

Mark the sets that live outside the working range. A tag, a note field, a separate exercise entry, whatever your log supports. The rule is that anything outside the range still gets recorded in full and never gets converted into a load, an e1RM, or a percentage for next week. Sets inside the range do the load math. Sets outside it do the fatigue math.

Done consistently for a cycle, this changes what your history is able to tell you. Progression on the top singles and triples reads cleanly because only clean readings feed it. Fatigue accumulation reads separately, on the accessory entries, where it belongs. Two signals instead of one blurred average, from exactly the same set of sessions you were already logging.

A log that can tell you which sets it used

Most trackers pool every rated set into one curve and hand back a single number. The number looks the same whether it came from four heavy triples or from twelve sets of fifteen, and nothing in the interface tells you which. That's a design choice, and it's the wrong one for anybody past their second year of training, because by then the accessory volume is large enough to drown the signal it's sitting next to.

Platepusher records the set the way you performed it: load, reps, and RPE as separate fields that stay separate, so a distinction like this one can be applied to the sets rather than lost inside an average. Drop a CSV from your current tracker and every lift, every set, every date lands as native data, which means the working-range read runs against the three years you already have instead of starting from Monday.

Six reps is where the reading gets soft. If your calculator's answer changes when you add a set of twenty to the log, it was never reading strength in the first place. Keep those sets. Keep them out of the math.