How Close to Failure Should You Train?

The short answer

You do not need to train to failure to build muscle. Stopping one to three reps short works about as well, and leaves you able to do the rest of your session properly.

Where failure earns a place is on exercises where stopping short is genuinely hard to judge, and where the cost of going all the way is low. That is a smaller list than most people assume.

What the evidence actually shows

A review pooling 15 studies compared training to momentary failure against stopping short, and found no evidence that going to failure produced more muscle growth. The effect for failure versus non-failure was 0.12, which is trivial and not statistically significant. Even the comparison that did reach significance, using a broader definition of failure, came out at 0.19, which the authors also called trivial (Refalo et al., Sports Medicine, 2022).

Two honest caveats from the authors themselves. There were not many studies, and the field cannot yet measure how close to failure people actually got, because "failure" is defined differently across studies and lifters are imperfect judges of their own effort.

So the accurate summary is not "failure is bad". It is that the difference is small enough that it should not be the thing you organise your training around.

The part that is not small

Failure is cheap in a single set and expensive across a session.

Take a set of squats to genuine failure. The set itself was maybe slightly more productive than stopping at two reps in reserve. But your next set is worse, the set after that is worse again, and the lift you had planned for later in the session is compromised. You bought a small gain on one set with a measurable loss across five.

That trade gets worse as the exercise gets bigger. A set of leg press to failure costs a lot. A set of cable curls to failure costs almost nothing.

This is the whole argument, and it explains why the answer differs by exercise rather than being one number.

Where to stop, by exercise

The pattern is that you protect the sets that come after, and you spend freely when nothing is coming after.

Judging where you are

Here is the awkward part. Stopping two reps short requires knowing you had two left, and people are not very good at that. They tend to underestimate what they have in reserve, which means "two reps short" is often four.

Two practical fixes.

Take one set to failure occasionally, on a safe exercise, to recalibrate. Not as training, as measurement. Most people discover their nine was a seven.

Watch the bar, not the feeling. When reps start slowing noticeably despite full effort, you are close. That signal is more reliable than how hard it feels.

How RIR, RPE, and percentage of 1RM measure intensity goes further into how precise these numbers really are, which is less precise than the decimal points suggest.

The failure trap

There is a specific way this goes wrong, and it is common enough to name.

Training every set to failure feels productive. It is unambiguous, it is hard, and it produces a clear sense of having worked. What it also does is make every set after the first one worse, force you to cut volume to survive the week, and leave you carrying fatigue that hides your actual progress.

The signature is a log where the weights stop moving while the sessions keep getting harder. That is not a strength problem, it is a fatigue problem. Why your lifts stopped progressing covers how to tell those apart, and a deload week is often the cheapest way to test it.

Failure and safety

Two exercises deserve a specific note. Failing a barbell squat or a bench press without safety bars or a spotter is a genuinely dangerous way to find out where your limit is. Use a rack with pins, use a spotter, or do your hard sets on something you can bail out of.

Stop a set for sudden or severe pain rather than pushing to failure through it. If pain is worsening, changing how you move, or persisting, reduce the load or range of motion, switch to a variation you tolerate, and have it assessed by a qualified healthcare professional. This article is general education, not individual medical or coaching advice.

Recording effort so it means something

If you decide sets should sit around two reps in reserve, you need to know whether they did. Judged by memory a week later, every set was hard.

Lift, which publishes this page, takes an RPE value per set while you log it, so effort is recorded next to the weight and reps rather than remembered. That matters more than it sounds: the same 100 kg for 8 at RPE 7 and at RPE 10 are different sessions, and without the number they look identical in your history. You can also set a target RPE per exercise when you build a workout, which is how "two short on squats, all-out on curls" becomes a rule instead of a mood.

RPE logging is free. See how Lift handles progression, or start from the complete guide to progressive overload.

The thing worth remembering: failure is a tool for measuring your limit, not a requirement for approaching it.