How close to failure should you train?
Not every set has to end in a grimace. Here is what training closer to failure actually buys you, what it costs, and why most people are further away than they think.
By Max van Strydonck
Founder of VERTEXPublished
"Train to failure" is the most repeated instruction in lifting and one of the least examined. It is treated as a virtue rather than a variable, which is odd, because it is one of the few things in a program you can dial up and down set by set.
The research is now clear enough to give you a rule. It is not "always" and it is not "never".
0–3 RIR
Reps in reserve — the productive window
3–5 reps
How far most lifters stop short when told to go to failure
What "failure" actually means
Momentary muscular failure is the point where you cannot complete another rep with good technique despite trying. That last clause matters. Stopping because the bar slowed down is not failure. Stopping because your form collapsed is not failure either — it is a different set, performed by different muscles.
The useful way to talk about this is reps in reserve (RIR): how many more you could have done. Zero RIR is failure. Two RIR means you left two in the tank. The scale is a decade old now, it maps cleanly onto RPE, and it is far more actionable than "go hard".
The dose-response, such as it is
Proximity to failure behaves like most training variables: it matters, and it matters less than the volume of shouting about it would suggest.
Proximity to failure
Closer to failure, a little more growth.
Relative hypertrophy by reps left in reserve, indexed to sets taken to failure. The curve rises, but the last two reps are the flattest part of it.
5+ RIR
warm-up territory
3–4 RIR
real stimulus
1–2 RIR
the sweet spot
0 RIR
failure
Two findings from that literature are worth stating plainly, because they point in different directions.
For size, closer is slightly better. Hypertrophy tends to increase as sets get nearer to failure. The gradient is real but gentle — going from four reps in reserve to one buys you a lot; going from one to zero buys you very little.
For strength, closer is not better at all. Meta-analyses comparing failure and non-failure training find no strength advantage to failure, and sometimes a small advantage to stopping short. Maximal strength is largely a skill, and skills are not practised well in a state of near-collapse.
"If you're not going to failure, you're not really training."
Overstated
What the last rep costs
The reason the answer is not simply "always go to failure" is that the last rep is not free. It is the most expensive rep in the set by a wide margin.
The last rep
Small return. Large bill.
Roughly how the growth stimulus and the recovery cost divide between the first reps of a hard set and the final grinding one.
Stopping at 2 RIR
95%
of the stimulus
55%
of the cost
Grinding to failure
100%
of the stimulus
100%
of the cost
Failure sets take longer to recover from, degrade the quality of every set that follows in the session, and raise the amount of rest you need before the next one. Do them on squats and you have not just had a harder set — you have had a worse third, fourth and fifth set, and possibly a worse Thursday.
That is the entire trade. A few percent more stimulus per set, in exchange for fewer good sets overall.
You are probably further away than you think
Here is the part that undermines everyone's confidence, including yours. When untrained and moderately trained lifters are asked to take a set to failure, and are then pushed by a coach or a machine, they turn out to have had several reps left. Estimates cluster around three to five.
This means two things at once, and they are both true:
- People who think they train to failure usually train at 3–5 RIR. Which is at the bottom edge of the productive band, and explains a lot of stalled progress.
- People who genuinely train to failure every set are rarer than gym folklore suggests. If you actually did, you would know, because you would be recovering from it.
Calibrate
Once a month, take one isolation set to genuine failure — a machine or a dumbbell exercise where nothing can fall on you. Not to train, but to recalibrate. Almost everyone finds that their "failure" was two reps early.
Where to spend it
The sensible policy is not uniform. Proximity to failure should track how much a movement costs you to recover from.
By movement
Where to spend your last reps.
2–3 RIR
Heavy compounds
Squat, deadlift, bench, overhead press. High systemic cost, high technical demand, and the movements where a failed rep is a safety event.
1–2 RIR
Secondary compounds
Rows, pull-ups, dips, leg press, hip thrusts. Enough stimulus to matter, contained enough to recover from.
0–1 RIR
Isolation work
Curls, lateral raises, leg extensions, cable flyes. Small muscles, small fatigue bill, nothing to drop. Take the last set to failure if you like.
3–4 RIR
Anything technical or new
A movement you are still learning is a skill session. Practising it exhausted teaches you the exhausted version.
A workable default: leave one to two reps on most working sets, and take the final set of an isolation exercise to failure when you feel like it. That gets you essentially all of the growth available, and leaves you able to train again on Wednesday.
When failure is actually the answer
Three situations flip the calculation.
- Light loads. If you are training at 30–40% of your max, the early reps recruit almost nothing. You have to get close to failure for the set to count at all. Low load and low effort together is just movement.
- Very low volume. If you only have time for one hard set per exercise, push it. The cost calculus changes when there is no next set to protect.
- Blood flow restriction and machine work near the end of a session. Nothing to fail under, nothing left to protect.
The short version
One to two reps in reserve on almost everything. Failure on isolation work when it suits you, rarely on the big lifts, never on a movement you are still learning.
And once in a while, actually go to failure on something safe — not because that set matters, but because it will tell you how honest your other sets have been.