Training to failure

Almost every page on this subject now tells you the same thing: don't go all the way, stop 1 to 3 reps short. It's reasonable advice and the research broadly supports it. There's just a problem with it that none of those pages mention. You can't tell when you're 2 reps short.

The short version

What training to failure means

Failure is the rep you try and can't finish. Not the rep that's slow, not the one where your form goes, and not the point where you decide you're done. If you racked the weight without a rep beating you, you stopped short of it.

The other half of the vocabulary is reps in reserve, usually shortened to RIR. That's how many reps you think you had left when you stopped. Two in reserve means you believed 2 more were possible. Keep both in mind, because the gap between them is what this page is about.

The answer depends on what you're training for

In 2024 Robinson and colleagues pooled the research differently from everyone before them. Instead of asking whether failure beats not-failure, they treated how close you get as a sliding scale and looked for a trend.

For strength, there wasn't one. Gains were similar whether sets finished near failure or a long way from it.

For muscle size there was. Growth improved the closer sets were taken to failure, at roughly half a percentage point of muscle size for each rep left unused. Over a few months that's the difference between a change you notice and one you don't, and it's a long way short of the effect the word "failure" implies.

A separate review by Grgic and colleagues, covering 15 studies, found no overall difference for either strength or size. One subgroup stood out. In people who were already trained, going to failure did produce more growth. The effect was small and it was real, and it's the group most people reading this are in.

So the honest summary is that failure isn't required, and isn't harmful either. Those authors said as much themselves.

You're probably wrong about where failure is

Here's what none of the pages giving you a number will tell you.

In 2020 Armes and colleagues took lifters with at least a year of training and asked them to stop a set at the point they were sure the next rep was impossible. Then they told them to carry on anyway.

They averaged 2 more reps. The researchers' own conclusion is that people with a year of training behind them aren't accurate enough at judging their own effort.

Read that against the advice and it falls apart. If you're told to stop 2 reps short, and you're 2 reps out in the direction of stopping early, then you're at failure while believing you're 2 away from it. Someone else with the same instruction is 4 reps clear. The same words produce two different sets.

The research has the same problem, and to their credit the researchers say so. Robinson's team state in their own summary that the reps in reserve behind every point on their curve were estimated from written descriptions of the studies, not measured. So a precise curve has been drawn through a quantity nobody can pin down. They call their own analysis exploratory. Almost nobody quoting it does.

None of this makes reps in reserve useless. It makes the precision fake. Treat "2 reps short" as a rough dial rather than a setting.

What it costs you

In 2023 Refalo and colleagues had 24 trained lifters, 12 women and 12 men, do 6 sets of a barbell bench press three separate ways: to failure, stopping 1 rep short, and stopping 3 short. Then they measured how fast the bar moved.

The cost is mostly today, not this week. A day later the gap had almost closed, and after 2 days it was gone. What failure really buys you is a much worse sixth set.

That's the link to how long you rest between sets. If you take a set to failure, the next one needs longer, and if you don't give it longer you'll do fewer reps at a lower speed. Half the arguments about failure are really arguments about not resting enough afterwards.

The men in that study lost more speed than the women did, both across the sets and straight after. It's one study, so don't build a plan on it, but it's worth knowing that the cost isn't identical for everyone.

The same study found soreness rose the closer sets got to failure. That's the other sensation lifters treat as a score, and soreness is a poor guide to whether a session worked for much the same reason.

What to do with your sets

• Don't chase failure on your heavy compound lifts. Strength doesn't need it, and a squat taken to failure ruins the sets after it.
• Save it for the last set of an isolation exercise, where growth is the point and the weight is light. A lateral raise or a curl costs you little when it goes wrong.
• Take one set to genuine failure occasionally, on purpose, to find out where it is. Most people have never met it and are calibrating against a guess.
• Count reps rather than trusting the feeling. The feeling is the thing that was measured and found wanting.

That last one is what the workout builder is for. It logs each set against a weight, so next week you're comparing 8 reps at 20 kg against 9 reps at 20 kg instead of comparing two memories of how hard something felt. A number survives the week. A feeling doesn't.

And if you're deciding how many sets to give an exercise in the first place, that's the other half of the same decision.

Common questions

How do I know if I am training to failure?

Mostly, you do not. Lifters with at least a year of training behind them were asked to stop when they were certain the next rep was impossible, then made to keep going. On average they had 2 more reps in them. Real failure is the rep you attempt and cannot finish, so if you have not failed one, you have not reached it.

Should you train to failure?

You do not have to. Across 15 studies, training to failure produced no more strength and no more size than stopping short. In lifters who were already trained, going to failure did produce slightly more muscle growth, so it is worth some of your sets rather than none of them.

Does training to failure build more muscle?

A little. Pooling the research, muscle size improves the closer a set is taken to failure, at roughly half a percentage point of growth per rep you leave unused. Strength does not follow that pattern at all: it came out the same whether sets stopped near failure or well short of it.

Is 4 reps to failure too little?

For size, stopping 4 reps short is toward the weak end, because growth improves as you get closer. For strength it is fine, since strength gains were similar across a wide range. Given that most people underestimate what they have left by about 2 reps, a genuine 4 short is probably nearer 6.

What does training to failure cost?

The rest of your session, mostly. When trained lifters took all 6 sets of a bench press to failure, their lifting speed fell 22% from the first set to the last. Stopping 3 reps short cost 6%. A day later the difference was small, and after 2 days it was gone.

How many times a week should you lift to failure?

Nobody has measured a weekly dose, so any specific number you are given is invented. What has been measured is the cost per session, and it is short-lived: the speed you lose is largely back within 24 hours and gone by 48.

What are reps in reserve?

The number of reps you believe you could still do when you rack the weight. Two in reserve means you stopped believing you had 2 left. It is the standard way to describe how hard a set was, and its weakness is that it depends on a judgment most people get wrong in the same direction.

Is it better to train to failure or leave reps in reserve?

For strength, leave reps in reserve: you get the same result for less fatigue. For size, closer to failure is better by a small margin, so it suits your last set of an exercise more than your first. Either way, count the reps rather than trusting how the set felt.

Where these numbers come from

Every figure on this page comes from one of 4 papers. We store the fetched record for each one, so the author list and the population you're reading are the ones the paper actually has.

• Robinson ZP, Pelland JC, Remmert JF and colleagues, 2024. Exploring the Dose-Response Relationship Between Estimated Resistance Training Proximity to Failure, Strength Gain, and Muscle Hypertrophy. Sports Medicine 54:2209 to 2231. The source of the strength and size split, and of the estimate problem.
• Armes C, Standish-Hunt H, Androulakis-Korakakis P and colleagues, 2020. "Just One More Rep!" Ability to Predict Proximity to Task Failure in Resistance Trained Persons. Frontiers in Psychology 11:565416. Two experiments, 14 and 24 people, all with at least a year of training. The source of the 2 rep error.
• Grgic J, Schoenfeld BJ, Orazem J, Sabol F, 2022. Effects of resistance training performed to repetition failure or non-failure on muscular strength and hypertrophy. Journal of Sport and Health Science 11:202 to 211. Fifteen studies, all in young adults.
• Refalo MC, Helms ER, Hamilton DL, Fyfe JJ, 2023. Influence of Resistance Training Proximity-to-Failure on Neuromuscular Fatigue in Resistance-Trained Males and Females. Sports Medicine Open 9:10. 24 trained lifters, 12 women and 12 men. The source of the speed figures.

What this page does not claim: that training to failure is dangerous, that it causes injury or overtraining, or that it does anything to your hormones. None of that was measured here. It doesn't tell you how many failure sets a week is safe, because nobody has measured a weekly dose. The Armes experiment used a machine leg exercise rather than a free weight, which is worth knowing, though what it measured was people judging their own effort rather than the exercise itself. Grgic's 15 studies were all in young adults, and those authors ask for more work in older and more experienced lifters.

If a lift hurts rather than being hard, stop and see a physiotherapist.