Type "I did 100 kg for 5 reps" into three different one-rep-max estimators and you will get three different maxes — maybe 112, maybe 116, maybe 120. If they all use the same inputs, how can they disagree? And which one should drive your training percentages? This article explains the source of the divergence and gives you a rule for choosing.
This disagreement trips people up because they expect math to be definitive. If a formula spits out 116 kg, it feels like a fact — precise, authoritative, final. But that precision is an illusion created by the decimal places. Behind every clean number is a statistical model that's making an educated guess, and different models guess differently. Once you understand that, the divergence stops being confusing and starts being informative: it tells you exactly how much uncertainty is baked into any single estimate.
Every formula is a model, not a measurement
A rep-max formula does not observe your strength; it models the relationship between how much weight you lift and how many times you can lift it. Each formula was created by measuring many lifters, plotting reps against load, and fitting a curve through the cloud of data points.
The key phrase is "the cloud of data points." Different researchers used different lifters, different exercises, and different rep ranges. Each curve is an honest summary of its own data — but those datasets weren't identical, so the curves aren't either. When you plug in your set, you are asking each model to extrapolate from its particular history. Disagreement is the natural result.
Where the formulas diverge most
The formulas agree closely at low reps and spread apart as reps climb. At 2–3 reps, most estimators land within a kilogram or two of each other, because you are extrapolating only a short distance from an actual near-max effort. At 10+ reps, the spread balloons, because the curve is now guessing across a wide gap — and high-rep sets are limited by factors the formulas can't see.
This is the single most important practical fact: estimate accuracy is highest in the 2–8 rep range. Above that, you are asking the model to predict a max from a set that is as much a test of grit and conditioning as of pure strength.
The hidden variable: rep endurance
Here is the deepest reason two people — or two lifts — produce different "true" answers. How many reps you can do at a given percentage of your max is not universal. It depends on muscle-fiber makeup, the exercise, and training history.
- A lifter with more fast-twitch fibers may manage only 3 reps at 90% of max, while another grinds out 5 at the same percentage.
- Squats and deadlifts allow more reps at a given percentage than bench press, because larger muscle mass sustains effort differently.
- A lifter who trains heavy singles builds different rep endurance than one who lives in the 8–12 range.
A single fixed formula cannot capture all of that, so it encodes an average. If your rep endurance differs from that average — and most people's does on at least one lift — the formula will systematically over- or under-shoot for you.
So which number do you trust?
You do not need the "true" max. You need a consistent, useful anchor for programming. That reframing dissolves most of the problem:
- Pick one formula and stick with it. Consistency matters more than which curve you chose. If you always use the same estimator, its bias cancels out — your percentages stay internally coherent even if the absolute number is slightly off.
- Estimate from 3–5 rep sets whenever possible, where all formulas are most accurate and agree most closely.
- Calibrate to your own lift. Note how many reps you can actually hit at a computed percentage. If your "80%" feels like a comfortable 8, your real max is higher than the estimate; if it's a grinding triple, it's lower. Adjust your anchor accordingly.
- Re-test the anchor periodically with a fresh submaximal set every few weeks.
For most trainees, a mid-range formula like Epley is a fine default. Powerlifters who live at low reps sometimes prefer Brzycki, which tends to read slightly more conservative at higher reps. The differences are small once you standardize on 3–5 rep inputs.
Why the anchor is only half the equation
An accurate max is worthless if your body can't recover to express it. Percentage-based programming assumes you show up recovered enough to hit the prescribed loads. That recovery is built on sleep and protein.
Keep protein around 1.6–2.2 g per kilogram of bodyweight so the training stimulus actually turns into strength. A precise estimated max paired with chronic under-recovery just gives you an accurate measurement of a stalling lift.
A calibration drill to personalize your anchor
Since every formula encodes an average rep endurance, the real upgrade is to measure your own. This simple drill turns a generic estimate into an anchor tuned to you and your specific lift, and it takes only a few sessions.
- Pick one lift and estimate a max from a clean 3–5 rep set using your chosen formula. Call it your working max.
- On a later session, test a prescribed percentage. Load 80% of that working max and do as many clean reps as you can, leaving nothing sloppy.
- Compare against the expectation. As a rough guide, 80% of a true max is often good for about 6–8 reps. If you cranked out 10 clean reps at your computed 80%, your real max is higher than the estimate — bump your anchor up a few percent. If you barely scraped 4, it's lower — trim the anchor down.
- Repeat on your other main lifts. Each has its own rep endurance, so calibrate the squat, bench, and deadlift separately.
After a couple of these checks, you'll know your personal rep-to-percentage relationship far better than any published table. Some lifters discover they're "grinders" who get lots of reps at high percentages (deep rep endurance), while others are "poppers" who fade fast past a few reps. Neither is better — but knowing which you are transforms how well your programming fits.
This is also the honest answer to "which formula is right?" The best formula is the one you've calibrated to yourself. Once you've measured how your prescribed percentages actually feel, the small differences between Epley, Brzycki, and Lombardi wash out — you're no longer trusting a stranger's average, you're trusting your own data. Re-run the drill every couple of months or after a long training phase, since rep endurance shifts as your training style changes. A few minutes of testing buys you a percentage table that finally tells the truth about your body instead of the average one.
Putting it together
Rep-max formulas disagree because each is a curve fit to a different dataset, and because how many reps you can do at a given load is personal. Rather than hunting for the "real" formula, choose one, feed it clean 3–5 rep sets, and calibrate the output against how your prescribed percentages actually feel. Do that and the exact algorithm stops mattering — you'll have a stable, personalized anchor that makes your programming coherent, which is the only job a 1RM estimate ever needed to do.