Stop most working sets one to three reps before you would fail. For muscle growth, the research finds little or no advantage in going all the way to failure, and sets that end with a rep or two in the tank grow muscle almost as well. For strength, how close you go barely seems to matter. Failure costs more recovery, so save it for exercises where a missed rep is safe and cheap, and use it late in a session rather than early.

Three terms worth keeping apart

Momentary failure is the rep you start and cannot finish, however hard you try. Technical failure comes earlier: the rep would still go up, but only if your back rounds, your hips shoot or your elbows flare. Reps in reserve (RIR) is how many more good reps you could have done when you racked the bar. A set of eight where a ninth was possible but a tenth was not is a set at 1 RIR.

Researchers usually write RIR on a 10-point effort scale, where a 10 is 0 RIR, a 9 is 1 RIR, and so on [6]. You only need the reps.

What the studies compare

A 2023 meta-analysis pooled 15 studies that compared training to failure with stopping short [1]. When the authors defined failure strictly as momentary failure, they found no advantage for it on muscle growth. A looser definition produced a trivial advantage, with an effect size of 0.19 and a confidence interval that touched zero.

A 2024 analysis took a different approach [2]. Instead of sorting studies into "failure" and "not failure", it estimated how many reps each training group left in reserve and looked for a slope. Muscle growth rose as sets ended closer to failure. Strength gains were about the same across a wide range of RIR. The authors call the analysis exploratory, because the RIR values were estimated from how each study described its training, and nobody measured them.

The most direct test in trained lifters split each person's legs [3]. One leg trained leg press and leg extension to momentary failure; the other stopped at about 2 RIR on the leg press and 1 RIR on the extension. After eight weeks, quadriceps thickness had grown by 0.181 cm on the failure side and 0.182 cm on the RIR side. The failure side showed more fatigue in every session.

Put together, closer is somewhat better for size, up to a point, and the last rep before failure buys very little.

What failure costs across the week

Fatigue from one set is easy to ignore. It adds up over five sessions. In one study, ten trained men did bench press and full squats three ways [4]. Three sets of 10 to failure left their jump height and bar speed lower for longer than the same 30 reps done as six sets of five, each stopped five reps short. The failure protocol still showed impaired performance 24 to 48 hours later.

If you squat on Monday and deadlift on Wednesday, that recovery time is part of Wednesday's session.

Your estimate is probably off by a rep

RIR only works if you can judge it, and most people cannot do so exactly. Across 12 studies, people underestimated how many reps they had left by about one, on average [5]. They thought they were closer to failure than they were. Two things made estimates better: sets of 12 reps or fewer, and estimates made near the end of a set rather than at the start. Years of training did not help much.

So a set you log as 2 RIR may really be 3. Correct for it by a rep and keep using the scale.

Where to set RIR for each kind of lift

• Heavy compound lifts (squat, deadlift, bench press, overhead press): 1 to 3 RIR. A missed rep here is a stuck bar or a lost position, and the fatigue carries into other lifts.
• Machine and supported compounds (leg press, chest press, pulldown): 1 to 2 RIR. The machine holds your position and you can stop a rep safely, so you can go closer.
• Isolation work (curls, lateral raises, leg extensions, calf raises): 0 to 1 RIR, with the last set taken to failure if you want it. The fatigue is local and the risk is low.
• First weeks of a programme, or after a break: add a rep or two to every number above. You are still learning the movement, and a margin covers a bad guess.

How to calibrate

Pick a machine or isolation exercise where failure is safe. Choose a load you think you can lift about ten times. At rep six, say how many you have left. Then keep going until you fail and count. Do this on a couple of exercises for two or three weeks. The gap between what you guessed and what you did is your personal correction, and the research suggests it will shrink when sets are shorter and your guess comes late in the set [5].

Do not calibrate on squats or deadlifts with nobody to spot you.

How Agyl records it

Every movement in Agyl says whether it asks for RIR, and the set row only shows the box when it does. A plan written by Agyl prescribes 1 to 2 RIR on compound lifts and 0 to 1 on isolation work, with a failure last set only on the isolation work. When you start a set, the row shows last time's weight, reps and RIR on that movement, or an F if that set went to failure, so a change in effort sits next to the change in load. If you are new or coming back from a break, the plan adds reps in reserve rather than taking any away.

When this advice does not apply

These studies used healthy adults, mostly under 40, over eight to twelve weeks. They say little about competitive powerlifters peaking for a meet, or about anyone training around an injury. A coach who can watch you lift will judge your RIR better than you can at first. If an exercise hurts, stop it; this article is about effort, and pain is a separate question.