Why Rep-Max Estimates Fall Apart Above Twelve Reps
The One-Rep Max Calculator refuses rep counts above twelve. That is not caution for its own sake. Past roughly a dozen reps both standard formulas stop describing anything real, and one of them fails spectacularly.
Watch Brzycki come apart
Brzycki estimates your max as weight × 36 ÷ (37 − reps). The denominator shrinks as reps rise, so the multiplier grows — and because it is a division, it does not grow gently:
| Reps | Multiplier | Estimated max from a 60kg set |
|---|---|---|
| 12 | 1.44× | 86kg |
| 20 | 2.12× | 127kg |
| 30 | 5.14× | 309kg |
| 35 | 18.00× | 1,080kg |
| 36 | 36.00× | 2,160kg |
| 37 | — | division by zero |
At 36 reps the formula claims your single is thirty-six times the weight you just used. At 37 it stops existing. Nothing in the equation prevents this — the guard has to live in the software.
Epley degrades more politely, since it is linear: every rep adds a thirtieth. A set of 30 projects double the weight lifted, which is wrong but not absurd. It just keeps climbing forever, which is also wrong, only less obviously.
The real reason, which is not mathematical
Even if the arithmetic behaved, high-rep sets would still be poor predictors, because a high-rep set is not primarily a test of strength.
A heavy triple is limited by how much force you can produce. Get to twenty reps and the limiter has changed: it is now local muscular endurance, breathing, blood flow, and your tolerance for a fairly unpleasant experience. Those qualities vary enormously between people, and they correlate only loosely with a maximal single.
Which is why two lifters who both squat 100kg for 20 reps might have genuinely different maxes — one is a good grinder, the other is strong — while two who both squat 100kg for 3 almost certainly do not differ much at all. The formula cannot see the difference. It only sees the rep count.
Where the estimates are actually good
Three to six reps is the useful window. Close enough to maximal that force production is the limiter, far enough from a true single to be safe and repeatable, and the range where Epley and Brzycki agree most closely.
Below three, you are approaching the risk of a max attempt without the benefit of actually testing one. Above eight the two formulas start diverging in ways that reflect their different shapes rather than anything about you — and they cross over at exactly ten reps, after which Brzycki becomes the more generous of the two.
If all you have is a high-rep set
Sometimes the situation is what it is: light dumbbells, a machine, a rehab context where heavy singles are off the table. A few options that beat forcing a bad estimate:
- Estimate in two steps. Use a high-rep set to find a working weight for a set of five, do that set next session, and estimate from it.
- Track the set instead of the max. If you can only train at 15 reps, then 15-rep performance at a given weight is your progress metric. It does not need converting into a hypothetical single to be useful.
- Use tonnage. The Training Volume Calculator does not care what rep range you work in. It compares this week against last week on whatever you actually did.
The general point
Every formula has a domain. These two were built from, and validated against, sets in the low rep ranges, and they were never claimed to work at thirty. Applying them outside that range does not give a rough answer — past a certain point it gives an answer with no relationship to reality, which is worse, because it still looks like a number.
The twelve-rep cap is where the tool draws that line. Below it you get an estimate worth acting on; above it you would get a plausible-looking figure that would quietly wreck the block built on top of it.
AMRAP sets: the honest way to use high reps
None of the above means high-rep sets are useless for measuring progress. It means converting them into a hypothetical single is the wrong move.
An AMRAP set — as many reps as possible at a fixed weight — is a genuinely good progress metric in its own right. Fix the weight, take the set to a consistent stopping point, and record the reps. If 80kg gave you 9 reps in January and 13 in March, that is unambiguous progress and it needed no formula at all.
Two conditions make it work:
- The weight must stay fixed. The moment both the weight and the reps move, the comparison stops being clean.
- The stopping point must stay consistent. Always to genuine failure, or always leaving one rep. Drifting between the two invents progress that is not there.
Many well-regarded programmes are built on exactly this, using a final AMRAP set as the week's feedback signal rather than testing maxes. It sidesteps the whole estimation problem: instead of guessing a single from a set, you compare a set to the same set.
Where each measure belongs
A rough map of which number to trust for which purpose:
- 1 rep: a true max. No formula needed, highest risk, and only worth doing if you compete.
- 2–6 reps: the estimation sweet spot. Use the 1RM calculator; the formulas agree closely and the input is safe to produce.
- 7–12 reps: usable but degrading. Epley and Brzycki begin diverging on shape rather than substance, crossing over at exactly ten.
- 13+ reps: stop converting. Track the set directly as an AMRAP, or track tonnage.
The wider habit worth having
Every formula has a domain of validity, and software that lets you leave it is doing you a disservice by being agreeable. A calculator that cheerfully accepts 30 reps and returns a number is not more flexible than one that refuses — it is just quieter about being wrong, and the number it hands back looks exactly as authoritative as a good one.
This is worth generalising beyond rep maxes. Bodyweight-based calorie estimates degrade at the extremes of body composition. Percentage tables assume a rep capacity that varies by lift and lifter. A CTR curve, a macro split, a Wilks coefficient — all of them are curves fitted to a population, useful in the middle and unreliable at the edges.
The habit that protects you is asking what range a number was built for before acting on it, and treating a tool that tells you where its limits are as more trustworthy than one that pretends it has none.