Neither has clearly won. In trials that compared them, effort-based (RPE) and percentage-based loading built squat strength that was not significantly different, and a pooled analysis of six trials found similar 1RM gains overall2.
The details are more interesting than the tie. Below are the terms, what each study measured, and why a percentage of an old max can drift.
The terms, briefly
- 1RM (one-rep max): the heaviest load you can lift once with good form. See what an estimated 1RM is.
- Percentage-based training: each set's load is a fixed share of a 1RM tested earlier, for example 80% for five reps, unchanged until the next test4.
- RPE (rating of perceived exertion): a 1–10 rating of how hard a set felt. In lifting it is usually tied to reps in reserve (RIR), how many more reps you could have done. An RPE 8 set leaves about two reps in the tank. The two scales are compared in RIR vs RPE.
- Autoregulation: adjusting the day's load to how you are performing that day, instead of following a fixed number. RPE is one way. Others measure bar speed with a sensor (velocity-based training), or use a rep-out set, where the number of reps done to failure in one set decides the next load4.
In plain terms: a percentage trusts the old test; RPE and the other autoregulation methods trust today's performance.
What the direct trial found
The cleanest head-to-head is a 2018 trial in resistance-trained men aged 19 to 351. Both groups squatted and bench pressed three times a week for 8 weeks with the same sets and reps. The only difference was how the load was chosen: 11 men used percentages of their 1RM, and 10 chose loads to hit target RPE ranges.
Both groups got stronger and grew muscle1:
- Squat 1RM: +13.91 kg with percentages, +17.05 kg with RPE.
- Bench press 1RM: +9.64 kg and +10.70 kg.
- Muscle thickness of the chest and outer thigh rose in both groups by similar amounts.
The between-group differences were not statistically significant. Using a probability-based method, the authors estimated a 79% chance that RPE loading gave a small squat advantage on average, and concluded that "both loading-types are effective"1. That probability method is not a standard significance test, so the plain reading is: a lean toward RPE, not a proven edge.
Strength gain over 8 weeks: percentage vs RPE loading · kg on the bar
Percentage of 1RM RPE
In plain terms: with sets and reps matched, both ways of choosing the load built strength and muscle. The RPE group was slightly ahead, but not by enough for the trial to call it.
What the pooled evidence says
Pooling many small trials gives a wider view, and here the reviews do not fully agree.
A 2022 meta-analysis combined six studies of autoregulated load (RPE-based and bar-speed methods) against percentages in resistance-trained people2. The difference in 1RM strength was 2.07 kg in favour of autoregulation, with a 95% confidence interval from −0.32 to 4.46 kg. Because that range includes zero, the difference was not significant. The authors concluded both approaches "produced similar improvements in strength".
A 2026 network meta-analysis compared five ways of setting the load across 27 randomised trials and 694 healthy, active people aged 13 to 35, mostly trained or athletes4. A network meta-analysis compares several methods at once, including pairs never tested head to head, by linking them through shared comparisons. For squat 1RM, against fixed percentages:
- Rep-out set: standardised effect 0.59 (95% CI 0.16 to 1.02), ahead.
- Bar speed: 0.41 (0.10 to 0.73), ahead.
- RPE: 0.12 (−0.31 to 0.56), "little-to-no difference".
- Bands and chains (not autoregulation, the load rises through the lift): 0.62 (0.29 to 0.95), ahead.
A standardised effect expresses a difference in units of the spread between people; the paper calls 0.2 to 0.5 small, 0.5 to 0.8 medium and above 0.8 large4. Fixed percentages ranked last. Only 5 of the 27 trials tested RPE, so its estimate is the least certain.
Squat 1RM gain vs fixed percentages, by loading method · SMD
A 2025 network meta-analysis ranked the methods differently for the back squat3. It found no moderate or large effect sizes between methods, and its ranking put the rep-out set first, RPE second, bar speed third and percentages last. Its authors still concluded that all three autoregulation methods, RPE included, were "significantly more effective" than percentages for maximal strength.
In plain terms: percentages never came out on top, but RPE's place against percentages ranges from "slightly ahead" to "can't tell apart", depending on which trials are pooled.
Why a percentage can drift: the max is not fixed
A percentage is only as good as the max it is based on. A 2026 study tested how much that max moves within one session5. Twenty-seven resistance-trained young men each did three sessions of the hexagonal-bar deadlift. They tested their 1RM, then did 30 minutes of rest, moderate sets, or sets to failure at 70% of 1RM, and tested again.
The max from the start of the session overshot the real max at the end by5:
- 4.1 kg after rest only,
- 8.9 kg after moderate sets,
- 15.7 kg after sets to failure.
Bar-speed estimates taken at the end of the session did not show that growing error. The best one, based on speed at a heavy load, was off by about 4.7 kg on average5.
How far the start-of-session max overshot the end-of-session max · kg
That is the mechanism behind autoregulation, stated plainly: a fixed percentage assumes today's max equals the tested max, and fatigue can make that untrue. This study measured the drift, not long-term strength gains, and it used the deadlift, not the squat.
In plain terms: 80% of last month's max is not always 80% of today's.
What this means in practice
This describes what the trials used, not a prescription:
- Both approaches built strength. In the direct trial and the 2022 pooled analysis, neither was clearly better12.
- RPE depends on judging effort. The trials that used it gave lifters target ranges, such as RPE 7 to 94. How well people rate their reps in reserve is covered in RIR vs RPE.
- Percentages need a current max. The drift study shows why an old or fatigued max can mislead5.
- Effort targets can be planned too. An RIR ramp sets a reps-in-reserve target for each part of a training block.
What the studies found
| Study | Who was studied | Key result | Year |
|---|---|---|---|
| Helms et al., *Front Physiol*1 | 21 trained men aged 19 to 35, 8 weeks, same sets and reps | Squat 1RM +17.05 kg (RPE) vs +13.91 kg (percentage); not significantly different | 2018 |
| Hickmott et al., *Sports Med Open*2 | Meta-analysis, 6 load-autoregulation studies in trained people | 1RM difference 2.07 kg (−0.32 to 4.46), not significant | 2022 |
| Huang et al., *J Exerc Sci Fit*3 | Network meta-analysis of autoregulation methods | No moderate or large squat differences; ranking rep-out set, RPE, bar speed, percentages; authors concluded all autoregulation beat percentages | 2025 |
| Bao et al., *Front Physiol*4 | Network meta-analysis, 27 trials, 694 active people aged 13 to 35 | Squat 1RM: rep-out set and bar speed ahead of percentages; RPE effect 0.12, not different | 2026 |
| Janicijevic et al., *Sports Med Open*5 | 27 trained young men, hex-bar deadlift, crossover | Pre-session max overshot the end max by 15.7 kg after sets to failure | 2026 |
| Barnes et al., *J Strength Cond Res*6 | 8 trained men, crossover, two 8-week blocks | Rep-out-set loading vs percentages: similar squat and deadlift gains | 2026 |
Limits worth knowing
- Small trials. The direct trial had 21 men1; the newest crossover had 86.
- Few RPE trials. Only 5 of 27 trials in the 2026 network tested RPE4.
- Mostly young, trained men. The 2026 network covered ages 13 to 35 and says its results cannot be extended to clinical, sedentary or older people4.
- The reviews disagree on rankings. The 2025 and 2026 networks ranked RPE differently34.
- The drift study is about one session and one lift. It did not measure strength gains over weeks5.
- What we read. The 2026 network and the drift study in full text; the other four as abstracts.
Latest research
The newest study we opened is the August 2026 drift study5, which shows why a fixed percentage can overshoot on a tired day. The newest training comparison is a small 2026 crossover in 8 trained men6. Daily loads set by a rep-out set at 85% of the starting 1RM gave squat and deadlift gains with no significant difference from fixed percentages. Bench press favoured percentages in that study. The 2026 network meta-analysis authors describe their own ranking as "probabilistic tendencies rather than definitive superiority"4. When a larger RPE-versus-percentage trial appears, this page will be updated and its date will change.
The short version
- RPE-based and percentage-based loading built similar squat strength in the trials that compared them.
- The direct trial leaned toward RPE (+17.05 vs +13.91 kg), not significantly.
- Bar-speed and rep-out-set methods ranked ahead of percentages in one 2026 network; RPE could not be told apart.
- A tested max can drop within a session, which is the case for adjusting loads to the day.
Educational overview only — not medical advice.
