Strength programs write the same instruction two ways: "3 × 8 @ RPE 8" or "3 × 8, 2 RIR". Both ask for a set that stops with about two good reps left. The only difference is which way the scale counts, and that matters far less than whether your rating is honest. This guide covers how the two scales line up, how accurately lifters actually use them, how close to failure you need to train, and how to turn a rating into a decision about your next session.

RPE and RIR in one line each

RIR (reps in reserve) counts down from failure: how many more clean reps you could have done when you racked the bar. 0 RIR means you could not have done another.

RPE (rating of perceived exertion) counts up to a maximum: how hard the set was on a 1–10 scale, where 10 is a maximal effort.

Older, general RPE scales asked how hard exercise felt, which is vague for a set of squats. A lifting scale anchored to reps in reserve first appeared in Reactive Training Systems' powerlifting manual in 2008. Zourdos and colleagues brought it into the research literature in 2016, and Helms and colleagues described how to program with it the same year. That anchoring is why the two terms are now interchangeable in most programs.

How the two scales line up

• RPE 10 = 0 RIR: nothing left.

• RPE 9 = 1 RIR.

• RPE 8 = 2 RIR.

• RPE 7 = 3 RIR.

• RPE 5–6 = 4–6 RIR. The scale groups these because a range is easier to judge once you are more than three reps from failure.

• RPE 3–4 = light effort, and RPE 1–2 = little to no effort: warm-up territory.

Half points exist too. An RPE 8.5 means you might have managed two more reps and could certainly have done one. Most lifters rate more consistently in whole numbers.

So which should you use? Whichever makes you rate honestly. RIR asks a concrete question, "how many more?", which many lifters find easier to answer with a bar on their back. RPE is the language most published programs use. Converting between them takes a second, so there is nothing to gain from switching programs over it.

How accurate are lifters at judging reps in reserve?

A rating is only as useful as it is accurate. The short answer from the research: reasonably accurate close to failure, much less accurate far from it.

• On average, people underestimate slightly. A 2022 meta-analysis by Halperin and colleagues pooled 12 studies with 414 participants. People underpredicted the reps they had left by about one rep on average, with large variation between studies. Predictions were a little better closer to failure and much better in sets of 12 reps or fewer.

• Far from failure, errors are large. Zourdos and colleagues (2021) had 25 trained men, averaging nearly five years of training, squat 70% of their max to failure, which took about 16 reps. They called out when they thought they had 5, 3 and 1 reps left. At their "1 left" call they were off by about two reps; at "3 left", by about 3.7; at "5 left", by about 5.2. More reps in the set meant worse predictions, and training age did not predict accuracy.

• In moderate-rep sets on a familiar lift, it gets tight. Refalo and colleagues (2024) tested 24 trained men and women on the bench press at 75% of their max. Their calls of 1 and 3 reps in reserve were off by about two-thirds of a rep on average, with no clear link to sex, experience or strength.

• It improves with practice. In an eight-week trial by Hermann and colleagues (2025), trained lifters got better at estimating reps in reserve over the study, especially on the bench press, and were more accurate on bench than on squat.

The practical reading: an RPE 8 on a set of 6 on a lift you know well is a useful number. An RPE 6 on a set of 20 is closer to a guess.

How close to failure do you need to go?

It depends on what you are training for, and the answer is less extreme than gym culture suggests.

• Strength. A 2024 series of meta-regressions by Robinson and colleagues found strength gains were similar across a wide range of reps in reserve. Leaving reps in the tank did not measurably cost strength.

• Muscle size. The same analysis found muscle growth increased as sets ended closer to failure. The authors caution that reps in reserve had to be estimated from how each study described its training, so the exact relationship is still uncertain.

• Failure itself is not required. In an eight-week trial, Refalo and colleagues (2024) had 18 trained lifters train one leg to failure and the other with 1–2 reps in reserve on the leg press and leg extension. Quadriceps growth was essentially identical between legs.

• Failure costs more fatigue. In an earlier study (Refalo and colleagues, 2023), six sets of bench press to failure cut bar speed by about 25% four minutes afterwards, against about 13% when sets stopped at 1 RIR and 8% at 3 RIR. A day later, the failure and 1-RIR sessions were still about 3% slower, while the 3-RIR session had fully recovered. By 48 hours the differences were gone.

For most working sets that points to a sensible default: finish around 1–3 reps in reserve (RPE 7–9), take isolation work closer to failure if size is the goal, and keep true failure for occasional sets on exercises where missing a rep is safe, such as machines.

How to calibrate your ratings

You will not get better at rating sets by guessing more often. A few habits help:

1. Rate straight after the set. A rating given five minutes later drifts towards how you feel in general.

2. Check yourself now and then. On the last set of a machine or isolation exercise, call your reps in reserve, then keep going to failure and count. If you said two and got five more, you have learned your bias.

3. Calibrate in the rep ranges you use. Accuracy is worse in long sets, so check a 15-rep set separately from a 6-rep set.

4. Do not test yourself to failure on a barbell squat or bench press alone. Use a spotter or safety arms, or test on a lift where failing is harmless.

Turning a rating into a decision

A rating earns its place in your log when it changes what you do next. Here is a worked example using double progression. The plan is bench press, 3 sets of 6–8 at 80 kg (176 lb), finishing each set around RPE 8.

• Week 1: 8, 7, 6 reps, rated 7, 8, 9. Normal. Stay at 80 kg and add reps.

• Week 2: 8, 8, 8, rated 7, 8, 8. Every set reached the top of the range without a grind. Add weight next time.

• Week 3 at 82.5 kg: 7, 6, 6, rated 9, 9.5, 10. The reps are inside the range, but every set was close to a max. Keep the weight, and let the ratings come down before chasing more reps.

The other pattern to watch plays out across weeks rather than within a session. If the same workout slowly goes from feeling like a 7 to a 9 while the weights and reps stay flat, the extra effort is not producing anything. That is one of the signs covered in our guide to when to deload. The reverse matters too: if everything feels like a 6 and you are still hitting your reps, you probably have room to push.

Adjusting by feel holds up in trials. Helms and colleagues (2018) had 21 trained men follow the same eight-week squat and bench program, with sets and reps matched, choosing loads either by RPE or from fixed percentages of their max. Both groups gained strength and muscle, the differences between them were not statistically significant, and the RPE group showed a small probable edge in strength.

Rating every set or rating the session

Rating every set gives the most detail, and it is worth the effort if your program sets a target RPE for each one. It is also more to log mid-workout. A single rating for the whole session is quicker. The "session RPE" method was validated by Foster and colleagues in 2001 against heart-rate measures during cycling and basketball, and it still captures the trend that matters most to a lifter: whether training is getting harder without getting better.

How Adensa uses effort

Adensa does not ask for an RPE or RIR on every set. Its prescriptions are driven by reps: each exercise has a rep range, and once every working set reaches the top of it, Adensa adds 5% on compound lifts or 2.5% on isolation work, rounded to the smallest weight jump you set for that exercise. How close to failure you take those sets is still your call, and everything above applies.

Effort comes in once per workout. When you save a session you can rate it from 1 (easy) to 10 (all out), and Adensa uses that rating in three ways:

• Your next session. Rate a workout 8 or higher after missing your targets, and Recover is pre-selected next time: 90% of the load, about 60% of the sets, and your progression does not move. Rate it 4 or lower after hitting everything, and Push is pre-selected, which sets the rep target at the top of your range without adding weight early. You can change either choice.

• The trend. Insights compares your recent rated sessions of a routine with the ones before them, leaving out Recover days. When effort has climbed by 1.5 points or more while at least half your lifts have not moved, it tells you the routine is getting harder, not heavier, and suggests a Recover session or two.

• Imported history. If you import from Hevy or Strong and your old sets carry RPE, Adensa uses those in the effort trend for any session without a session rating.

If you want an app that sets a target RIR for every set and adjusts the load from it, Adensa is not that app. Some others, Fitbod among them, log reps in reserve set by set. Adensa keeps each set to weight and reps and asks one honest question at the end.

If that trade suits the way you train, see how Adensa turns your log into the next session's targets.