The Method N° 02 · How muscle grows

Effort

Finish your sets close to failure. One to three reps short builds nearly as much, for less.

Every study checked against the paper, 12 October 2026. Each section carries its grade. Run the numbers in the app.

Close to failure is where growth climbs

Strong

Failure is the rep you cannot complete. Across the trials, muscle growth rose as sets ended closer to it, while strength gains barely changed.¹ Going all the way, though, added at most a trivial amount over stopping short.²

One to three short builds the same

Strong

In trained lifters, stopping one or two reps short built the same quadriceps as going to failure: 0.181 cm against 0.182 cm over eight weeks.³ Two 2026 trials, one on arms and one on legs, found the same.^4,5

Failure costs more

Strong

Bar speed four minutes after a bench session fell 25% after failure sets, 13% at one rep short and 8% at three, though all three had recovered by 48 hours.⁶ Three sets of ten to failure slowed recovery by a day or two against the same reps stopped short.⁷ Failure sets also felt worse: more discomfort, a higher session effort, less enjoyment.⁸

Most people have more left than they think

Strong

Across 12 studies, people underpredicted their reps to failure by about one,⁹ and in 141 people the typical error was two to three reps.¹⁰ Trained lifters calling one or three reps left were off by about two-thirds of a rep.¹¹ Calls are most accurate close to failure and with heavier loads.⁹

Practice may sharpen it

Still open

In one six-session study, errors fell by up to 2.3 reps with practice;¹² another found experience made no difference.¹³

Effective reps is a theory

Still open

You may have heard that only the last five reps of a set count. It is a reasonable idea drawn from how muscle fibres are recruited, but no trial has tested it directly, so we do not lean on it.

How the Vanguard app runs this

After each set the app asks one thing: how many you had left. The target moves through the block, from one easier than written in the first week to a rep closer to failure at the peak. If your calls run generous, it reads your max lower, so a load never climbs on a guess.

The research

The 13 sources this chapter stands on, with what each one actually found.

Robinson, Pelland, Remmert et al., 2024 · Sports Med
Meta-regressions
Muscle growth rose as sets ended closer to failure; strength gains were similar across reps in reserve. Exploratory: reps in reserve were estimated from the studies.
Read the paper by Robinson, Pelland, Remmert et al., 2024 (opens in a new tab)
Refalo, Helms, Trexler et al., 2023 · Sports Med
Meta-analysis · 15 studies
Training to failure gave a trivial edge in growth (effect size 0.19); momentary failure against stopping short, 0.12, not significant.
Read the paper by Refalo, Helms, Trexler et al., 2023 (opens in a new tab)
Refalo, Helms, Robinson et al., 2024 · J Sports Sci
One leg each · n = 18 trained · 8 weeks
Quadriceps thickness rose 0.181 cm with sets to failure and 0.182 cm stopping one to two reps short.
Read the paper by Refalo, Helms, Robinson et al., 2024 (opens in a new tab)
Vasconcelos, Alves, Refalo et al., 2026 · Muscles
One arm each · n = 19 trained · 8 weeks
Elbow-flexor thickness differed by 0.032 cm between failure and stopping short, and the max by 0.59 kg.
Read the paper by Vasconcelos, Alves, Refalo et al., 2026 (opens in a new tab)
Vasconcelos, Ruivo, Refalo et al., 2026 · J Strength Cond Res
One leg each · n = 19 trained · 8 weeks
Knee-extensor thickness differed by 0.019 cm and the max by 0.2 kg.
Read the paper by Vasconcelos, Ruivo, Refalo et al., 2026 (opens in a new tab)
Refalo, Helms, Hamilton et al., 2023 · Sports Med Open
Randomised crossover · n = 24 trained
Bar speed 4 minutes after six bench sets fell 25% after failure, 13% at one short and 8% at three short; the differences were gone by 48 hours.
Read the paper by Refalo, Helms, Hamilton et al., 2023 (opens in a new tab)
Morán-Navarro, Pérez, Mora-Rodríguez et al., 2017 · Eur J Appl Physiol
Crossover · n = 10 trained men
Three sets of ten to failure slowed recovery for 24 to 48 hours against the same reps done as six sets of five.
Read the paper by Morán-Navarro, Pérez, Mora-Rodríguez et al., 2017 (opens in a new tab)
Refalo, Helms, Hamilton et al., 2025 · Eur J Sport Sci
Same 8-week trial · n = 18 trained
Discomfort 5.1 against 4.1 and session effort 5.4 against 4.3 for failure sets; feelings were more positive stopping short.
Read the paper by Refalo, Helms, Hamilton et al., 2025 (opens in a new tab)
Halperin, Malleron, Har-Nir et al., 2022 · Sports Med
Scoping review and meta-analysis · 12 studies · 414 people
People underpredicted their reps to failure by 0.95 on average; more accurate closer to failure, with heavier loads and in later sets.
Read the paper by Halperin, Malleron, Har-Nir et al., 2022 (opens in a new tab)
Steele, Endres, Fisher et al., 2017 · PeerJ
n = 141
People underpredicted their reps to failure, with a typical error of 2.6 to 3.4 reps.
Read the paper by Steele, Endres, Fisher et al., 2017 (opens in a new tab)
Refalo, Remmert, Pelland et al., 2024 · J Strength Cond Res
n = 24 trained · bench at 75% of max
Calling one and three reps left, the average error was 0.65 reps.
Read the paper by Refalo, Remmert, Pelland et al., 2024 (opens in a new tab)
Wiedenmann, Kunellis, Rappelt et al., 2026 · BMC Sports Sci Med Rehabil
n = 26 · 6 sessions
Errors in calling reps left fell by up to 2.3 reps with practice.
Read the paper by Wiedenmann, Kunellis, Rappelt et al., 2026 (opens in a new tab)
Droguett, Festa, Quintana et al., 2026 · J Hum Kinet
n = 16
Training experience made no difference to accuracy.
Read the paper by Droguett, Festa, Quintana et al., 2026 (opens in a new tab)

Research describes groups of people. Your coach will tell you how it applies to you.

Questions this chapter answers

Should you train to failure?