RIR vs RPE is a difference of direction, not of substance: RIR (reps in reserve) counts how many more reps you could have done, and RPE (rate of perceived exertion) rates how hard the set felt on a 1–10 scale. On the lifting-specific scale the two are tied together by one line of arithmetic — RIR = 10 − RPE — so a set at RPE 8 is a set with 2 reps in reserve.
Two scales, one idea
RPE started life as a general effort scale for endurance work. The version lifters use was re-anchored to something you can count: how many reps were left in the tank when you racked the bar. That re-anchored scale — the RIR-based RPE scale — was formally described and tested in 2016 by Zourdos and colleagues, who had 29 experienced and novice squatters rate sets at 60%, 75% and 90% of their one-rep max while bar velocity was measured1. Ratings tracked bar speed closely in both groups (r = −0.88 in the experienced lifters, r = −0.77 in the novices), which is the evidence that the number people report is measuring something real1.
The idea is simple:
- RIR answers "how many more could you have done?" — 0, 1, 2, 3…
- RPE answers "how hard was that, out of 10?" — 10 is nothing left, 9 is one rep left, and so on.
They meet at the top of the scale. RPE 10 is RIR 0. Every step down the RPE scale is one more rep in reserve. Below about RPE 6 the published scale stops giving single-rep precision and groups 5–6 as "4–6 reps in reserve", because nobody can reliably tell four reps left from six2.
In plain terms: RIR is the number you feel after the set; RPE is the same number written from the other end.
The conversion table
This is the table the RIR / RPE converter runs, and the one Zyra Workouts uses when it turns a logged RPE into the reps-in-reserve label you see on a set.
| RPE | RIR | What it felt like |
|---|---|---|
| 10 | 0 | Nothing left; a technical-failure set |
| 9 | 1 | One more rep was possible |
| 8 | 2 | Two more reps were possible |
| 7 | 3 | Three more reps were possible |
| 6 | 4 | Four more; the app counts anything easier as this |
| below 6 | 4 (clamped) | Warm-up territory; not distinguished |
Half points work too. The app rounds first, then converts: RPE 8.5 rounds to 9 and becomes RIR 1; RPE 9.5 rounds to 10 and becomes RIR 0. If you want half-rep precision, log the RPE rather than the RIR label.
Why the app clamps RPE at 6
The conversion in the source is one line:
RIR = 10 − clamp(round(RPE), 6, 10)
The clamp does two things. It caps RPE at 10 (you cannot have fewer than zero reps left), and it floors it at 6, so RPE 4 and RPE 6 both come out as RIR 4. That floor is deliberate. The scale's own authors present RPE 5–6 as a range of 4–6 reps in reserve rather than exact counts2, and a set that easy is not the kind of set a progression rule should be reasoning about. Collapsing everything easy into "at least 4 left" keeps the number honest instead of pretending to a precision nobody has.
The reverse direction is clamped the same way: RPE = 10 − clamp(round(RIR), 0, 4). Type RIR 6 into the converter and it reads back RPE 6, not RPE 4.
In plain terms: the app trusts your effort rating between "four left" and "nothing left", and counts anything easier as a warm-up.
How Zyra Workouts logs it
Zyra Workouts stores effort per set as an RPE value. From that it derives a short RIR label — 2R for two reps in reserve, 1R for one — and shows it on the reps cell; a set at RIR 0 shows no label, because there was nothing in reserve to report. Change the value after the set (uncheck, edit, re-check) and the new number is what Smart Progression sees next time.
That last part is why the rating matters beyond the log. Two of the app's rules read it directly:
- The progression ceiling for a set target is
round((repMin + repMax) / 2) + RIR target— the effort target is baked into the number of reps the app expects before it adds weight. That rule is walked through in double progression. - The "Deload needed" flag on the dashboard fires only when an exercise's estimated 1RM has fallen by 2% or more *and* the average logged RIR over the window is 1 or less — see training plateau or time to deload.
Neither rule can work from "it felt hard". They need the number, which is why the app asks for it on every working set.
A caveat that the research supports: RIR is a skill. The novice group in the validation study rated less consistently than the experienced group1, and the scale's application paper suggests newer lifters practise recording RIR before letting it steer intensity on its own2. Log it from day one; lean on it once your ratings stop surprising you.
How accurate are the ratings?
The scale measures something real, but a rating is an estimate, not a count. Researchers test accuracy by asking people to call their reps in reserve, then letting them continue to failure and counting what was actually left.
Closer to failure, better calls. In a 2021 study, 25 trained men did one squat set to failure at 70% of their 1RM, which came to about 16 reps, and called out when they thought they were at RPE 5, 7 and 93. On average their calls were off by 5.15 reps at RPE 5, 3.65 at RPE 7 and 2.05 at RPE 9. Sets with more reps gave less accurate calls far from failure, and training age did not predict accuracy3.
How far off RIR calls were, by how close to failure they were made · reps
The pooled picture. A 2022 exploratory meta-analysis pooled 12 studies with 414 participants4. On average people underpredicted how many reps they had left by 0.95 reps (95% CI 0.17 to 1.73), with very large differences between studies. Predictions were slightly better closer to failure, in later sets, and in sets of 12 reps or fewer; training status and upper versus lower body made little difference4. A 2023 study of 58 trained and untrained men and women on curls, pushdowns and seated rows found the same pattern: more accurate near failure and in later sets, with no significant effect of sex or training experience5.
That adds nuance to the novice caveat above. The 2016 validation found novices' ratings tracked bar speed less tightly1, yet the later accuracy studies did not find training experience made a significant difference to how well people predicted reps left. Either way, the ratings that matter most for the app, RIR 0 to 3, are also the ones people call most accurately.
In plain terms: expect your RIR to be off by a rep or two, less so near failure. That is one reason the app folds anything easier than RIR 4 into a single bucket.
Same load, different RPE by exercise
RPE also depends on the exercise, not only on the load. A 2026 study followed 104 people through 8 weeks of training, 3 sessions a week, with 5 or 6 exercises at 50% or 80% of 1RM for 6 or 14 reps, and recorded RPE after every set6. At the same assigned percentage of 1RM, exercises rated differently: the shoulder press had the highest average RPE (17.2 ± 2.3) and the calf raise the lowest (13.3 ± 2.6). Heavier relative loads and fewer reps went with higher RPE, and RPE rose from set to set at different rates for different exercises6.
Mean RPE at the same percentage of 1RM: highest vs lowest exercise · RPE
Note the scale. The abstract does not name it, and values like 17.2 only exist on the older 6–20 Borg scale, not on the 1–10 lifting scale this page is about. So the study measured general effort, not reps in reserve. Its authors concluded that percentage of 1RM "might be the most appropriate method" for setting training intensity6. That is a reading about using RPE to set loads, and it is the question the RPE vs percentage article covers. For logging effort per set, the lesson is narrower: the same RPE on two exercises need not mean the same load.
In plain terms: the rating belongs to the exercise. An RPE 8 on shoulder presses and an RPE 8 on calf raises are each true for that exercise, and the app keeps them apart because it tracks effort per exercise.
What the studies found
| Study | Who was studied | Key result | Year |
|---|---|---|---|
| Zourdos et al., *J Strength Cond Res*1 | 29 experienced and novice squatters | RIR-based RPE tracked bar speed (r = −0.88 experienced, −0.77 novices) | 2016 |
| Zourdos et al., *J Strength Cond Res*3 | 25 trained men, one squat set to failure at 70% of 1RM | Calls off by 5.15 reps at RPE 5, 2.05 at RPE 9 | 2021 |
| Halperin et al., *Sports Med*4 | Meta-analysis, 12 studies, 414 people | Reps left underpredicted by about 0.95; better near failure and in sets of 12 or fewer | 2022 |
| Remmert et al., *Percept Mot Skills*5 | 58 trained and untrained men and women, 3 machine exercises | More accurate near failure and in later sets; sex and experience not significant | 2023 |
| Lyristakis et al., *J Strength Cond Res*6 | 104 people, 8 weeks | At the same %1RM, shoulder press RPE 17.2 vs calf raise 13.3 | 2026 |
Limits worth knowing
- Accuracy was measured to failure. Every accuracy study had people finish the set to failure to check the call, which is not how most working sets are done.
- High between-study spread. The 2022 meta-analysis found very large heterogeneity and called itself exploratory4.
- Different scales. The 2026 exercise study reports values that fit the 6–20 Borg scale, not the RIR-based 1–10 scale6.
- Unknowns in the 2026 study. The abstract does not say who the 104 people were, their training status, or the RPE of the exercises between the two extremes.
- Abstracts only for the four newer papers; the 2026 study's full text is behind a paywall.
Latest research
The newest study we opened is the 2026 program study in 104 people6. Across 8 weeks, heavier relative loads (β = 0.6 to 2.0) and fewer reps (β = −0.1 to −0.3) went with higher RPE, and at the same assigned percentage of 1RM the shoulder press rated about 4 points higher than the calf raise on the study's scale. It is a reminder that an effort rating is specific to the exercise and the set, and it says nothing about how accurately people count reps in reserve, which the 2021 to 2023 studies measured directly.
The short version
RIR is reps left; RPE is effort out of ten; on the lifting scale RIR = 10 − RPE. Zyra Workouts stores RPE, rounds it, clamps it between 6 and 10, and shows the result as an nR label — and both the progression ceiling and the deload flag are computed from that number. In studies, people's reps-in-reserve calls were off by about one rep on average and got more accurate near failure4, and the same load can rate differently on different exercises6. The converter does the arithmetic either way round. Zyra Workouts is in development and not yet on the App Store; the rules quoted here are read from its source code and may change before release. Training guidance, never medical advice. Educational overview only — not medical advice.
