How hard should a set be? Effort, failure and RIR

No, you don't need to train to failure: the point where you can't complete another rep. End most working sets close to it instead. A common target is 2–3 reps in reserve (RIR), which means you could have done 2 or 3 more reps when you stopped. That's the effort an expert panel proposes as enough.^[1]

The hard part is the guessing. RIR is an estimate you make at the end of a set, and on average people are off by about a rep. They tend to guess they have fewer reps left than they do.^[2] Experienced lifters miss too.^[2][3]

So there are two jobs: picking how hard a set should be, and getting better at telling.

The article on progressive overload left you with a placeholder rule. The last rep of each set should be hard but still look like the first. Here's what "hard" means in numbers.

Why it matters

Effort is what makes a set count. The American College of Sports Medicine's main advice is to lift with high effort. Its panel argues that a set needs enough effort to cause change.^[1] You can miss in both directions.

Too easy. The plan says 10 reps, so you stop at 10. But the weight was light enough for 16. You did the reps and left most of the work behind. Sets that end far from failure seem to build less muscle than sets that end close to it.^[4]

Too hard. Every set goes to failure, because that feels like the honest way. It costs more: in one trial, sets taken to failure caused more fatigue within the session, and later sets lost more reps.^[5] For that extra cost, the studies found little or no extra gain.^[6][7]

For a beginner, that's good news: the target is a range, not a knife edge. You only need to get roughly close.

What the research says

A few terms first. A position stand is an official summary of the evidence by a group of experts. The main one here is the American College of Sports Medicine (ACSM) position stand from 2026. A meta-analysis pools the results of many studies on the same question into one estimate.

You may also see effort written as RPE (rating of perceived exertion). The version used for lifting runs from 10 down and maps onto RIR. RPE 10 means 0 RIR, RPE 9 means 1 RIR, and so on.^[8] So "RPE 8" and "2 RIR" mean the same set. This article sticks to RIR.

Failure isn't required

The ACSM panel looked at 137 systematic reviews (studies that gather all the research on one question), covering more than 30,000 people. Training to failure "did not consistently impact training outcomes."^[1] In its words, failure is "not necessary for benefits to occur."^[1]

Strong evidence. A position stand and two meta-analyses point the same way: going all the way to failure didn't consistently add strength or muscle.

How close is close enough

The ACSM panel proposes that ending sets "near-failure", or at about 2–3 RIR, is enough effort.^[1] It also says the evidence can't yet pin down an exact RIR target.^[1] Expert consensus. This is the panel's view, not a tested finding.

So 2–3 RIR is a sensible target, not a precise one. The evidence can't yet say whether 2 beats 3, or the other way round.^[1]

Close seems to matter more for growth than for strength

A 2024 study pooled earlier trials in a meta-regression: a meta-analysis that checks whether results change as one factor changes. Here the factor was how many reps in reserve people had when their sets ended. Muscle growth improved as sets ended closer to failure. Strength gains were about the same across a wide range of RIR.^[4]

Limited evidence. The authors call the analysis exploratory. They estimated each study's RIR from how it described its training, rather than measuring it. They say the exact relationship is still unclear.^[4]

Put that next to the failure studies and a picture forms. Getting close matters for growth. The final rep or two, all the way to failure, adds little.

The relationship probably isn't a straight line, with each rep closer adding the same amount of muscle.^[7][9] The studies point the same way but don't agree on exactly how close. Limited evidence for the exact shape.

Strength cares least. When the failure and non-failure groups didn't do the same total work, stopping short even gave slightly bigger strength gains.^[6] Limited evidence: it's one subgroup of one analysis.

Experience may change the picture

The International Universities Strength and Conditioning Association (IUSCA) has its own position stand. It says beginners can build plenty of muscle without training close to failure, and that effort seems to matter more as you gain experience.^[9] Expert consensus. This is the panel's view, not a tested finding. The IUSCA paper was written for athletes.

In people who already trained, one meta-analysis found a small growth advantage for failure.^[6] A 2025 trial in 42 trained men and women, doing one set per exercise, also found growth tended to favour failure, though the differences were modest. Strength came out the same.^[11]

Limited evidence. Only two of the meta-analysis's growth studies used trained people.

What failure costs

One 2024 trial tested this in 18 people who already lifted: 12 men and 6 women. Each person trained one leg to failure and the other leg stopping short. On the leg press the short leg stopped at 2 RIR, and on the leg extension at 1 RIR. They trained twice a week for 8 weeks.^[5]

Both legs grew about the same. Thigh muscle thickness went up about 0.18 cm in each.^[5] Within each session, the failure leg lost more speed and more reps from set to set.^[5] Limited evidence. It's one small, short trial, and it only trained legs.

The ACSM panel adds that going to failure may be a poor idea for some groups, such as older adults. The reasons it gives are strain on blood vessels and the risk of injury when form breaks down.^[1] Expert consensus. That's a caution, not a measured risk.

RIR guesses are off by about a rep on average

A meta-analysis of 12 studies with 414 participants compared people's guesses of how many reps they had left with how many they could actually do. On average they guessed about one rep too few. They had more left than they thought.^[2] Results varied a lot between studies, and the authors call the analysis exploratory. Moderate evidence.

This isn't a beginner problem. In the same meta-analysis, training experience didn't change how accurate people were.^[2] A study of 27 men and 31 women found the same: sex, training experience and past practice with RIR didn't predict accuracy.^[3] Moderate evidence.

Guesses were a little more accurate closer to failure and in sets of about 12 reps or fewer, and slightly more accurate in later sets.^[2][3] Moderate evidence. A guess of "1 left" tends to be closer to the truth than a guess of "5 left".

It's also a skill you can practise. In a study of 26 younger and older adults, the size of people's errors fell by up to 2.3 reps over 6 sessions in 6 weeks.^[10] In an 8-week trial with 42 trained men and women, people's guesses got more accurate over the study, most of all on the bench press.^[11] Limited evidence. Both studies were small.

How to apply it

The steps below build on the ACSM's 2–3 RIR proposal and the studies above. This is coaching practice, not a tested finding.

1. End most working sets with 1–3 reps left. The ACSM panel proposes 2–3. Stopping at 1 is fine when the reps still look clean.
2. Ask one question at the end of each set. "How many more could I have done with the same form?" Write that number next to the weight and reps.
3. Assume you're guessing low. If you call 2, you may have had 3 left. On average, people underestimate by about a rep.^[2]
4. Check your guesses now and then. On a safe exercise, such as a machine or a dumbbell curl, take the last set to failure and count. Compare the result with what you guessed. Don't test this on a heavy squat or bench press without safety bars or a spotter.
5. Let the reps decide the weight. If you reach the top of your rep range with 3 or more reps left, the weight is too light. Add a little next time.
6. Keep failure rare. If you go to failure, save it for the last set of an isolation exercise. That's one that moves a single joint, like a curl or a leg extension.

In your first few weeks, while you're still learning how a lift should look, it's fine to stop further from failure. The IUSCA panel's view is that beginners can grow without going close.^[9] Move toward 1–3 RIR once the movement feels steady.

A worked example

A seated cable row, 3 sets of 8–12 reps, aiming for 2 RIR.

• Set 1: 45 kg × 10. You think you had 2 left. Write "RIR 2".
• Set 2: 45 kg × 10. Same feeling. "RIR 2" again.
• Set 3: you check. Same 45 kg, all the way to failure. You get 14 reps.

So even after two sets, you had 4 left at rep 10, not 2. Your first set probably had more. Next session, go up to 50 kg and aim for 10 reps at 2 RIR. Check again in a few weeks.

Heavy or light: rep ranges covers how the weight on the bar changes this, and why light sets have to be hard. A set that ends this close to failure is what coaches call a hard set. Hard sets are what you count when you plan how many sets per week to do.

What works in practice. In one exploratory analysis, growth improved the closer sets ended to failure, so 1 RIR may build more than 3.^[4] But you'll repeat these sets every week for months. Pick the end of the range you can keep doing with clean reps. Coaching practice.

Common mistakes

Stopping at the number on the plan. The plan says 10, so 10 feels like the job. But if 16 were possible, the set was easy. Fix: treat the rep number as a range and let RIR tell you when the weight needs to go up.

Taking every set to failure. It feels like the only honest effort. In the studies it didn't consistently add strength or muscle, and in one trial it cost more fatigue.^[6][5] Fix: keep most sets at 1–3 RIR and save failure for the odd isolation set.

Trusting your first guesses. A set that feels heavy can still have reps left in it. On average, people have about one more rep left than they think.^[2] Fix: check yourself on a safe exercise every few weeks. In small studies, guessing improved with practice.^[10][11]

Counting ugly reps. A rep where your back rounds or the bar crawls up crooked isn't the same rep. Fix: the set ends when your form changes. Count your RIR from there.

Arguing over 2 versus 3. The evidence can't pin down an exact target.^[1] Fix: aim for the range, write down your guess, and spend the energy on adding reps over time.

How Gymbers fits

The Gymbers app lets you log RIR next to each set, from 0 to 4, beside the weight and reps. Programs in the app set a target RIR for each working set, and you can mark a set as a failure set. Some programs raise the target over the weeks, for example from about 4 RIR early on to 2–1 later. That ramp is a coaching choice, not a research finding.