RPE and Reps in Reserve: Autoregulating Load Without a 1RM Test

Reps in reserve is the number of repetitions you could still have completed when you stopped a set. Stop with two left and that set was 2 RIR, which on the resistance-training RPE scale is an 8. That is the whole system, and it exists because percentage-based prescription answers a question nobody actually has.

A percentage tells you what to lift based on a maximum measured on some other day, under different sleep, food and stress. RPE tells you how hard today's set was, today. The trade is that a percentage is objective and often wrong, while RPE is subjective and gets more accurate the more experienced you are. That means it works well for the people who need it least and poorly for beginners who need it most. What follows is where the scale came from, how the numbers map to loads, what the accuracy research actually shows, and how to use it without pretending it is a measurement.

From Borg to reps in reserve

The original scale is Borg's, described in Medicine and Science in Sports and Exercise (14:377-381, 1982), running from 6 to 20 and designed for endurance exercise, where it was built to track heart rate roughly across the range. It works reasonably for cycling and running and poorly for lifting, because the thing a lifter needs to judge is not how hard they are breathing. It is how close the set came to failure.

Zourdos and colleagues rebuilt the scale around that in the Journal of Strength and Conditioning Research (30:267-275, 2016). Their version anchors each rating to a specific number of repetitions remaining rather than to a general sense of exertion, running from 10 at momentary failure downward. They also validated it against average concentric velocity, finding that bar speed tracked the ratings. A lifter reporting a higher RPE was, in fact, moving the bar more slowly.

Helms and colleagues set out how to apply it in practice in the Strength and Conditioning Journal (38:42-49, 2016), including how to write RPE-based prescriptions and how to teach the scale to lifters who have never used it. Between those two papers the system acquired both a validation and a manual, which is why it spread from powerlifting into general programming over the following few years.

The scale, and what each number feels like

The physical cues matter more than the introspection. The most reliable signal is the change in bar velocity across the set: at 3 RIR the reps still look similar to each other, and by 1 RIR the concentric phase has visibly slowed even though you are trying to move it as fast as you can. That involuntary slowing is what Zourdos and colleagues measured, and it is the reason the scale is more than a vibe. If your last repetition looked exactly like your first, you were further from failure than you thought.

A prescription written in this language looks like "4 sets of 6 at RPE 8" rather than "4 sets of 6 at 80 percent." You select the load that produces six repetitions with two left, and if you have slept badly that load is lower than last week. The programme's intent survives a bad day instead of being broken by it.

The accuracy problem, stated plainly

This is the part that gets skipped in enthusiastic write-ups, and it is the most important thing about the method.

RPE is a trained skill, and novices are bad at it. Zourdos and colleagues found that more experienced lifters rated their sets more accurately than less experienced ones. Hackett and colleagues, in the Journal of Sports Sciences (30:1405-1413, 2012), asked trained lifters to predict how many repetitions they had left mid-set and found systematic underestimation: participants consistently thought they were closer to failure than they were, and the error was larger further from failure. A set called 3 RIR was often really 5 or 6.

Two patterns hold across this literature. Accuracy improves as you approach failure, because the sensation at 1 RIR is unmistakable while the difference between 4 and 6 RIR is genuinely subtle. And accuracy improves with training experience, because judging it requires having actually reached failure enough times to know what its neighbourhood feels like.

The practical consequences are specific. If you have been lifting for a few months, your RPE 8 is probably an RPE 6, and a programme written entirely in RPE will quietly under-load you for months. The remedy is calibration: occasionally take a set to genuine momentary failure on a safe movement (a machine, a dumbbell press, a leg extension, something where failure does not put you under a bar), predict beforehand how many reps you will get, and compare. A few of those exercises does more for your rating accuracy than a year of guessing.

There is a second consequence. RPE ratings drift over a session and over a training block as fatigue accumulates, and they are influenced by mood, caffeine, competition and how much you want the set to be finished. Two people can rate the same objective effort differently, and the same person can rate it differently in January and June. It is an internal instrument with a slow calibration drift, which is fine as long as you do not treat its output as data.

Where autoregulation genuinely earns its place

The case for RPE is not that it is more accurate than a percentage. It is that it targets the right thing.

Fixed percentages are derived from a maximum, and that maximum is itself an estimate. Take a lifter who does 100 kg for five clean repetitions and runs it through the one-rep max calculator: Epley returns 116.7 kg, Brzycki 112.5 kg, Lombardi 117.5 kg and Wathen 116.6 kg, averaging 115.8 kg. Five kilos separate the extremes before any programming has happened, and the reasons for that spread are the subject of estimating versus testing a one-rep max.

Here is the sharper problem, and it is a direct consequence of RIR. Every one of those equations assumes the set was taken close to failure. If that lifter actually stopped with three repetitions in reserve, a genuine 5 at RPE 7, then their real capacity at 100 kg is eight repetitions, and running 100 kg for 8 through the same calculator returns an average of 125.4 kg. The same load, the same lifter, the same day, and a 9.6 kg difference in estimated maximum depending entirely on how close to failure the set was. Any percentage table built on the lower figure prescribes loads roughly eight percent too light throughout.

So the two systems are not rivals with different error bars. Percentage prescription depends on knowing your proximity to failure, which is the thing RPE measures. You cannot escape RIR by using percentages; you can only leave it unstated.

The second argument is daily readiness. Strength on a given movement fluctuates meaningfully week to week with sleep, stress, food and residual fatigue. A fixed 85 percent is too heavy on a bad day and too light on a good one. Helms and colleagues compared RPE-based against percentage-based loading in periodised programmes matched for sets and repetitions, reporting in Frontiers in Physiology (9:247, 2018), and found RPE-based loading produced at least comparable strength outcomes. The literature here is not large and the effects are not dramatic, so the honest summary is that autoregulation is a reasonable alternative rather than a demonstrated improvement.

Running a programme on it

The mechanics are straightforward once the scale is calibrated.

Write the prescription as sets, repetitions and a target RPE, then select load by feel on the day. Log the load you used alongside the RPE you hit, because that pairing is your actual training record. The comparison across weeks is exactly what progressive overload asks you to demonstrate. If the same RPE is being produced by heavier loads over a block, you are getting stronger. If the load is static and the RPE is climbing, you are accumulating fatigue.

Most productive work sits between RPE 6 and 9. Below 6, the stimulus is thin for anything other than technique and speed. At 10, the fatigue cost rises faster than the benefit: Refalo and colleagues, reviewing proximity-to-failure studies in Sports Medicine (2023), found hypertrophy outcomes broadly similar across a range of proximities to failure, with little advantage to training at true failure and a clear cost in recovery. Leaving one to three repetitions is a defensible default, and the volume question that sits alongside it is handled in how many sets and reps.

Two adjustments matter. RPE targets should be lower on compound movements than on isolation work, because a squat taken to failure costs the whole system while a curl taken to failure costs an arm. And rating accuracy degrades badly when rest is too short. If you are still breathing hard from the previous set, what you are rating is systemic fatigue rather than proximity to muscular failure, which is one more reason rest periods are not the place to economise.

A common structure is to progress load at a fixed RPE within a block, then step back. Week one at RPE 7, week two at 8, week three at 9, week four lighter. That has the same logic as any planned variation in intensity, expressed in a currency that adjusts itself to how you actually turned up.

Where the method breaks

RPE is unreliable for very short, very heavy sets in inexperienced lifters, because a single at 1 RIR requires knowing what a single at 0 RIR feels like, and finding that out is a maximal attempt with everything that entails.

It is unreliable in long sets. At fifteen or twenty repetitions the limiting sensation is metabolic discomfort and breathing rather than force production, and people stop well before mechanical failure without noticing.

It is unreliable for anyone whose relationship with training intensity is distorted. Lifters who habitually stop early over-rate their sets; lifters who treat every session as a test under-rate them. The scale reports what you perceive, and perception carries whatever bias you brought with you.

And it does not substitute for recovery. A programme that autoregulates load downward every week is telling you something about sleep and energy intake rather than about programming, which is where the TDEE calculator and an honest look at food and rest do more than any change to the prescription.

One clinical note. Training close to failure with heavy loads involves a Valsalva manoeuvre and sharp transient rises in blood pressure. If you have uncontrolled hypertension, known cardiovascular disease, a spinal or joint condition, or you are pregnant or postpartum, how close to failure to train is a question for a clinician or a qualified coach who can assess you in person, not something to settle from a scale in an article.

The reasonable way to hold both

Use percentages for structure and RPE for the day. A block written as sets of five in the region of 80 to 85 percent, executed at RPE 8, has a defined shape and a self-correcting load. On days when 85 percent feels like an RPE 9.5, you take the lower end and the block survives.

And keep the scale honest by testing it. Occasionally predict a set and then take it to failure, on a movement where failure is safe, and see how far off you were. Most people discover they had more repetitions available than they believed. That discovery is worth more than any refinement to the programme it sits inside.