Skip to content
Wiki/training principles/Training to Failure
training principles 6 min read

Training to Failure

The meta-analysis found no advantage to training to actual failure. It also found that stopping far short does not work, which is the half everyone forgets.

Last updated: 30 September 2026

Overview

Momentary muscular failure is the point where you cannot complete another repetition despite trying. Not "it got hard". Not "I put it down". The rep did not happen.

Whether you need to get there is one of the few arguments in training where a good meta-analysis exists, and the answer it gives is more interesting than either side wanted.

Quick Answer

A 2023 meta-analysis found no advantage for training to momentary muscular failure over non-failure training for muscle growth. Training close to failure does matter — roughly the last one to three repetitions in reserve — but the final, failed rep adds nothing worth its cost.

Key Points

  • No measurable advantage to actual failure over stopping just short.
  • Being close to failure does matter. Stopping at 5 reps in reserve does not work.
  • One to three reps in reserve is the working range for most sets.
  • Failure costs fatigue that reduces the next set and the next session.
  • Save it for machines and isolation, where it is safe and cheap.
How much a set contributes as it approaches muscular failure reps left in reserve 5 3 2 1 0 and a great deal more fatigue
The last few reps do the work. The last one, the failed one, adds fatigue rather than stimulus.

What the meta-analysis found

Fifteen studies, comparing training to failure against stopping short, and comparing different velocity-loss thresholds:

ComparisonResult
Set failure (any definition) vs non-failureTrivial advantage, effect size 0.19
Momentary muscular failure vs non-failureNo advantage, effect size 0.12, not significant
High (>25%) vs moderate (20 to 25%) velocity lossNo advantage, effect size 0.08

Volume load did not moderate the result, and neither did relative load.

The conclusion the authors reached is worth reading precisely: there is no evidence that training to momentary muscular failure is superior to non-failure training for muscle growth, and closer proximities to failure do not always produce more growth. They describe the relationship as potentially non-linear.

The half people forget

"You do not need to train to failure" gets repeated as though it means effort is optional. It does not.

The studies finding no difference were comparing failure against training close to failure — a rep or two or three short. Nobody is claiming that stopping five reps early works, and the evidence for that is not ambiguous: sets terminated well short of failure produce clearly less growth.

So the finding is narrow and specific: the last rep, the one you fail, is not doing anything the rep before it did not already do.

Reps in reserve, in practice

RIRWhat it feels likeUse it for
0Failed the next repRarely. Last set, isolation
1 to 2Certain you had 1 or 2 leftMost working sets
3Comfortable, could clearly do moreEarly sets, heavy compounds
4 plusToo easy to drive growthWarm ups only

RPE and RIR covers how to judge this, and the honest caveat that beginners are bad at it — most people with 3 reps left believe they have none.

Why failure costs more than it gives

It wrecks the next set. Take set one to failure and set two is several reps worse. Across four sets that is less total volume than stopping one short would have produced.

It is systemically expensive. A set to failure on squats takes far longer to recover from than the same set on a leg extension, for a stimulus that is barely different.

Technique goes first. The rep before failure on a barbell lift is usually the rep where form changes. See ego lifting and technique.

It is harder to progress from. You cannot beat a logbook you filled in by emptying the tank.

Where failure is genuinely useful

  • -Isolation work. A failed lateral raise costs nothing.
  • -Machines. Safe to fail on alone. See machines vs free weights.
  • -The last set of an exercise, occasionally, when nothing follows it.
  • -Calibrating your own judgement. Going to true failure once in a while tells you how wrong your RIR estimates are, which is usually very.

Where it is a bad idea

  • -Heavy barbell squats and deadlifts, especially alone.
  • -Overhead pressing with a heavy bar above you.
  • -Every set of every exercise, which is how people end up unable to train three days a week.
  • -Anything where a failed rep means the weight lands on you.

Common Mistakes

MistakeWhy it causes problemsBetter approach
Every set to failureCosts volume and recovery1 to 3 RIR on most sets
Reading "not necessary" as "effort optional"Stopping 5 short does not grow muscleStay close to failure
Failing heavy squats aloneGenuinely dangerousFail on machines
Trusting your RIR estimateMost people underestimate what is leftTest it occasionally
Failure on the first set of fourThe other three sufferSave it for the last

Frequently Asked Questions

Do I need to train to failure to build muscle?

No. A 2023 meta-analysis found no advantage for momentary muscular failure over training close to it. Being close still matters.

How close to failure should I train?

One to three reps in reserve for most working sets, nearer one on isolation and nearer three on heavy compounds.

Is training to failure bad?

Not bad, just expensive. It adds fatigue without adding much stimulus, which costs you the rest of the session and sometimes the week.

Should beginners train to failure?

Occasionally, on machines, mostly to learn what failure actually feels like — because most beginners stop several reps earlier than they think.

Sources

Refalo et al. (2023), influence of resistance training proximity-to-failure on skeletal muscle hypertrophy: systematic review with meta-analysis - https://pubmed.ncbi.nlm.nih.gov/36334240/

National Strength and Conditioning Association - https://www.nsca.com/

American College of Sports Medicine - https://acsm.org/

training to failurereps in reservedo you need to train to failure

Spotted an error or want to improve this article? Sign in to suggest a change