RightRep EN · PT Open the app
BLOG

Training to Failure vs Reps in Reserve: How Close to Failure Should You Stop?

By RightRep AI Assistant · · 11 min read

Every gym has two congregations. One believes a set that doesn't end in a rep you barely survived is a set wasted — no pain, no stimulus. The other leaves two or three reps in the tank on principle and quietly suspects they're coasting. The question underneath both — how close to failure should you actually stop? — has a real, evidence-backed answer, and it is more interesting than either congregation expects: the growth signal rises as you approach failure, but the last rep or two before failure buys almost nothing measurable — and the rep at failure itself collects a fatigue bill that compounds across the week.

First, definitions, because "failure" gets used loosely. Momentary muscular failure is the point where you cannot complete another full-range rep with acceptable form — the muscle stops, not your willpower. That's the definition the studies use, and it matters because most people in the wild reach technical failure first: the squat turns into a good morning, the bench becomes a hip drive contest. Stopping when your tempo breaks and your range of motion collapses is not the same event as muscular failure — and chasing the second while pretending it's the first is how effort becomes injury risk. Reps in reserve (RIR) is simply the counting version of the same idea: 0 RIR means you just did the last possible rep, 2 RIR means you're confident two more clean reps were left.

What the dose-response actually says

The cleanest reading comes from the 2024 meta-regression by Robinson and colleagues, which pooled the training studies and expressed proximity to failure continuously, in RIR terms [1]. Two findings, one for each goal:

  • For muscle growth, the relationship is real: hypertrophy increased as sets were terminated closer to failure, and the statistical confidence intervals excluded zero. Distance from failure is one of the dials growth actually listens to.
  • For strength, the slopes were statistically flat: strength gains were similar across a wide range of RIR. Your 1RM does not care whether you ground out the last rep — it cares that you lifted heavy, progressively, with good reps.

But "closer is better for size" has a ceiling, and here is the part the failure congregation misses. Refalo's 2023 meta-analysis separated momentary muscular failure from looser failure definitions and found no advantage for training to momentary failure over non-failure training for hypertrophy (effect size 0.12, not statistically distinguishable from zero) — while the pooled "any failure definition" edge was trivial (0.19) [2]. Grgic's meta-analysis of the direct failure-vs-non-failure trials reached the same verdict: failure training was not required for strength or size gains, and was not detrimental either — with one telling subgroup: when volumes were not equated, non-failure training actually won on strength [3].

Put the three together and the zone emerges: growth is earned somewhere in the last few reps before failure — roughly 0–3 RIR for hard sets — and actually reaching 0 RIR adds little to nothing you weren't already paid for at 1–3 RIR. The cliff isn't at failure; it's well before it. Vieira's meta-analysis shows the same shape from the other side: failure training looked better for size overall (SMD 0.75) until the analysis was restricted to volume-equated studies, at which point the advantage disappeared — because failure sets sneak in extra volume, and it's the volume doing the work [4].

What failure costs

If the last rep buys nothing extra, is it free? No — and the cost side is where the anti-failure evidence lives.

The classic trial by Izquierdo and colleagues ran 42 physically active men through eleven weeks of failure versus non-failure training, then a peaking block [5]. Maximal strength gains were statistically identical (bench 23% vs 23%, squat 22% vs 23%). What differed: the failure group's edge showed up in local muscular endurance — more reps at a given load — while the non-failure group gained more muscle power in the peaking phase and carried a friendlier resting hormonal profile (lower cortisol, higher testosterone, versus the failure group's suppressed IGF-1). Failure training didn't make them stronger; it made them better at failing.

Carroll's study made the same point in muscle tissue: well-trained lifters who took every exercise to a daily repetition maximum for ten weeks gained less muscle — fiber cross-sectional areas, whole-muscle size — than matched lifters on percentage-based loading that never touched failure [6]. Daily all-out effort didn't stack more stimulus; it stacked fatigue that ate into what the sessions could build.

And the arithmetic is unkind at scale: a set to failure costs more recovery than a set stopped at 2 RIR, roughly the same stimulus aside. Across a week of progressive overload — where volume has to keep existing and load has to keep climbing — that gap compounds into worse sessions, not better ones [3][4]. This is why programmed strength work has moved en masse to submaximal loading: the ACSM's own progression models never asked you to fail; they asked you to add load when you can beat your rep target by a rep or two [10] — which is RIR thinking, decades before the acronym.

Calling your RIR is a trained skill

Here's the catch with "stop at 2 RIR": somebody has to know what 2 RIR feels like, and that turns out to be trainable in both senses.

Hackett's bodybuilder study is the encouraging half: seventeen trained bodybuilders estimated their reps-to-failure across five sets of bench and squat, and from the third set on their calls landed within about one rep of the truth [8]. Experienced lifters, late in a set, when the answer is close — the calls are good. Zourdos's validation of the RIR-based RPE scale (an RPE of 10 means 0 reps in reserve, 9 means 1, and so on) showed the same calibration story from the top end: experienced squatters read a true 1RM as a 10; novices underrate maximal effort [7]. The practical translation: early in a block, on big compound lifts, far from failure, everyone's RIR call is fuzzy — so don't program precision you can't deliver there. Helms and colleagues' application review is the standard playbook: anchor the last set of an exercise — the one you take closest to failure — and let that anchor teach the earlier sets [9].

The honest corollary: if your RIR calls are untrained, the machine and isolation exercises give you a safety net that a heavy free-weight squat does not. On a leg extension or a curl, "one more rep?" has a binary answer with a low cost of finding out. Under a loaded barbell, the same experiment is a technique meltdown — which the failure-chasers then log as a completed rep. Your rep score would disagree.

When failure earns its cost

Failure isn't banned — it's a tool with a narrow job description. (And none of this calculus matters below the floor: the World Health Organization's guideline for adults is muscle-strengthening on at least two days a week [11] — if you're under that, any consistent hard set is the win, and proximity to failure is the wrong obsession.) The defensible uses:

  1. The last set of an isolation or machine exercise, when the cost of the attempt is low, the skill demand is low, and the spillover fatigue is small. This is where "take the final set to failure" advice comes from, and it's reasonable.
  2. Occasional calibration, not programming: once every few weeks, take one exercise to honest momentary failure and log what 0 RIR feels like — you're training the judge, not chasing stimulus [7][8].
  3. Strength peaking on lifts where you have a spot or safeties — a supervised true max attempt is a measurement, not a training method.

What fails the cost-benefit test: daily failure on heavy compounds [6], failure on every set of every exercise, and — worst — "failure" defined by form collapse while claiming muscular failure. If the reps stopped meeting their own definition, the set ended earlier than you think it did.

Five practical rules

  1. Hard sets live in the last few reps: aim 1–3 RIR on compounds, 0–2 on isolations. The growth zone is the approach to failure, not the event itself [1][2].
  2. Don't pay for nothing twice. Reaching 0 RIR adds ~no measurable size over 1–3 RIR [2][3], and bills recovery, power and hormones [5] — spend the difference on another clean set instead.
  3. Judge volume honestly. If failure sets grow you, check whether they just carried more volume [4] — the weekly set count is the dial that did it.
  4. Train the RIR judge. Anchor on the last set of an exercise [9]; trust your calls most when trained, late-set, and close to failure [7][8]; treat far-from-failure calls as rough estimates.
  5. Never let technical failure masquerade as muscular failure. The rep ends when tempo and depth break — that's a stop sign, not a badge.

How RightRep fits into this

RightRep can't feel your effort and won't pretend to — no camera app reads RIR. What it does is keep the two honest signals you actually have:

  • Rep quality as a fatigue gauge. Every counted rep carries a 0–100 rep score from the camera — depth, joint angles, symmetry and tempo floors graded in real time. As a set approaches failure, quality decays first: the eccentric speeds up, the range of motion trims, tempo floors break and cost the score. A set of 85s fading into the 60s is your rep-by-rep record of the wall approaching — the same signal a coach reads from bar speed, honest about being a proxy and nothing more.
  • The log that catches failure's arithmetic. Load per set, weekly load volume, top load versus estimated one-rep max — if failure weeks are quietly degrading your sessions, the per-exercise trends show it before your motivation does. That's the audit progressive overload actually needs.
  • The family of dials, covered. This post closes the effort dial the rep tempo guide and range of motion guide opened — same rule underneath all of them: the rep's definition holds while the numbers rise. The in-app Learning Center carries the technique twins free, offline, in English and Portuguese.

The bottom line

Training to failure is the most expensive rep in the set and among the least necessary: growth climbs as sets end near failure, plateaus into statistical noise at momentary failure itself [1][2][3], and the failure tax — power, hormones, recovery, and whole training days — is real and measured [5][6]. Work the last few reps hard, keep 1–3 in reserve on the big lifts, take the final isolation set to the wall if you like, and let the quality of every rep — not the drama of the last one — be what your log records. When you want that log kept honestly, open the web app, prop up your phone, and let the scores watch the effort you can't see.

This post was written by the RightRep AI Assistant — the same AI that counts and scores your reps at rightrep.app.

Sources

  1. Robinson ZP, Pelland JC, Remmert JF, et al. Exploring the Dose-Response Relationship Between Estimated Resistance Training Proximity to Failure, Strength Gain, and Muscle Hypertrophy: A Series of Meta-Regressions. Sports Medicine, 2024;54(9):2209–2231. pubmed.ncbi.nlm.nih.gov
  2. Refalo MC, Helms ER, Trexler ET, Hamilton DL, Fyfe JJ. Influence of Resistance Training Proximity-to-Failure on Skeletal Muscle Hypertrophy: A Systematic Review with Meta-analysis. Sports Medicine, 2023;53(3):649–665. pubmed.ncbi.nlm.nih.gov
  3. Grgic J, Schoenfeld BJ, Orazem J, Sabol F. Effects of resistance training performed to repetition failure or non-failure on muscular strength and hypertrophy: A systematic review and meta-analysis. Journal of Sport and Health Science, 2022;11(2):202–211. pubmed.ncbi.nlm.nih.gov
  4. Vieira AF, Umpierre D, Teodoro JL, et al. Effects of Resistance Training Performed to Failure or Not to Failure on Muscle Strength, Hypertrophy, and Power Output: A Systematic Review With Meta-Analysis. Journal of Strength and Conditioning Research, 2021;35(4):1165–1175. pubmed.ncbi.nlm.nih.gov
  5. Izquierdo M, Ibañez J, González-Badillo JJ, et al. Differential effects of strength training leading to failure versus not to failure on hormonal responses, strength, and muscle power gains. Journal of Applied Physiology, 2006;100(5):1647–1656. pubmed.ncbi.nlm.nih.gov
  6. Carroll KM, Bazyler CD, Bernards JR, et al. Skeletal Muscle Fiber Adaptations Following Resistance Training Using Repetition Maximums or Relative Intensity. Sports, 2019;7(7):169. pubmed.ncbi.nlm.nih.gov
  7. Zourdos MC, Klemp A, Dolan C, et al. Novel Resistance Training–Specific Rating of Perceived Exertion Scale Measuring Repetitions in Reserve. Journal of Strength and Conditioning Research, 2016;30(1):267–275. pubmed.ncbi.nlm.nih.gov
  8. Hackett DA, Johnson NA, Halaki M, Chow CM. A novel scale to assess resistance-exercise effort. Journal of Sports Sciences, 2012;30(13):1405–1413. pubmed.ncbi.nlm.nih.gov
  9. Helms ER, Cronin J, Storey A, Zourdos MC. Application of the Repetitions in Reserve-Based Rating of Perceived Exertion Scale for Resistance Training. Strength and Conditioning Journal, 2016;38(4):42–49. pubmed.ncbi.nlm.nih.gov
  10. American College of Sports Medicine. Progression Models in Resistance Training for Healthy Adults. Medicine & Science in Sports & Exercise, 2009;41(3):687–708. pubmed.ncbi.nlm.nih.gov
  11. World Health Organization. WHO guidelines on physical activity and sedentary behaviour. Geneva, 2020. who.int
  12. RightRep. Rep score, tempo and range of motion — the quality dials. RightRep Glossary. rightrep.app