One-Rep Max Testing: Estimate It or Test It?

You can estimate a one-rep max from a set you have already done, and for nearly every training purpose that is the better choice. The estimate costs nothing, carries no injury risk, and lands close enough to prescribe loads from. A genuine single is worth attempting only when the number itself is the point: a competition, a formal test week, a standard you have to clear on paper.

That is the short answer. The longer one is more interesting, because the four equations in common use do not agree with each other, and the size of their disagreement tells you something about how much confidence the estimate deserves.

Four equations, four different curves

The equations were developed independently, from different data, by people solving the same practical problem: a lifter performs a set to near-failure, and a coach needs a maximum to program against.

Boyd Epley, then a strength coach at Nebraska, published the version most widely used in 1985. It is linear in repetitions: the estimated maximum equals the weight lifted multiplied by one plus reps divided by thirty. Matt Brzycki, writing in JOPERD (64:88-90, 1993), took a different route, dividing the weight by a shrinking factor of 1.0278 minus 0.0278 times reps. Lombardi (1989) used a power function, scaling the weight by repetitions raised to the 0.10. Wathen (1994) fitted an exponential decay, dividing one hundred times the weight by 48.8 plus 53.8 times e to the power of negative 0.075 times reps.

Those are four different mathematical shapes, not four versions of one formula. A straight line, a reciprocal line, a power curve and an exponential curve can be tuned to overlap across a narrow band and will inevitably separate outside it. That is exactly what they do. Between roughly three and eight repetitions they cluster tightly. Push past ten and they scatter.

Brzycki's structure carries a hard ceiling worth knowing about: its denominator shrinks toward zero as repetitions climb, so the estimate inflates sharply and eventually breaks down entirely. It is a well-behaved equation inside its intended range and nonsense outside it.

What they say about one real set

Take a lifter who squats 100 kg for five clean repetitions, stopping with something left. Run that through the one-rep max calculator and you get four answers:

Five kilos separate the lowest estimate from the highest. On a squat in that range, five kilos is roughly one small plate per side, and it is the difference between a top single that grinds and one that fails.

The averaged figure is the one worth programming from, and not because averaging is mathematically principled. It is not. It is worth using because no single equation has been shown to dominate the others across lifts, populations and rep ranges, so committing to one of them expresses a confidence the evidence does not support. Averaging spreads the error rather than betting on a particular direction of it.

The spread widens as the set gets longer

Repeat the exercise at higher repetitions and the pattern becomes obvious. A lifter moving 100 kg for ten repetitions gets estimates of 133.3 kg from Epley, 133.4 kg from Brzycki, 125.9 kg from Lombardi and 134.7 kg from Wathen, averaging 131.8 kg. The gap has grown to nearly nine kilos, with Lombardi now sitting well below the other three.

At fifteen repetitions the equations stop being useful. The same 100 kg produces 150.0 kg from Epley, 163.7 kg from Brzycki, 131.1 kg from Lombardi and 150.9 kg from Wathen. That is a spread of more than thirty kilos, and Brzycki's inflation is exactly the denominator problem showing itself.

This is the practical rule: accuracy degrades above roughly ten repetitions, and past twelve the estimate is closer to a guess than a measurement. Long sets are also limited by things a maximum attempt is not, including local muscular endurance, breathing and pain tolerance. A set of twenty tells you about a different quality than a single does, and no equation can convert between the two.

The same logic argues for keeping test sets short. A set of three to five is the sweet spot: long enough to be safe and repeatable, short enough that all four curves still agree.

It also depends which lift you are doing

Estimation works best on the squat, bench press and deadlift. These are the movements the equations were built around, and they share the features that make estimation viable: a large amount of muscle involved, a stable groove, a load path limited by force production rather than by balance or grip.

Accuracy falls away elsewhere. On isolation work, single-joint movements and machines with unusual leverage curves, the relationship between repetitions and percentage of maximum shifts. Lifts with a substantial technical or stability demand, the overhead press and the front squat among them, tend to produce optimistic estimates because form degrades before force production does. Grip-limited pulls fail for a related reason: the hands give out while the back and legs still have capacity.

So treat an estimate as lift-specific. Your bench press estimate may be reliable while your overhead press estimate runs consistently high, and there is no correction factor for that beyond your own logbook.

When a true maximum is worth attempting

There are legitimate reasons to test. Powerlifters need to know their openers. Some strength standards and job-related tests require a measured single. And once or twice a year, a genuine attempt calibrates every estimate you have been working from.

Outside those cases the return is thin. A true maximum takes a long warm-up, leaves meaningful fatigue behind, and interrupts training for days. Because a single is a skill as much as a display of strength, an untrained lifter often tests below their actual capacity simply because they have never braced under that load. The ACSM Position Stand on progression models in resistance training treats maximal testing as one option among several rather than a routine requirement, and for good reason: most programming decisions need a load, not a certified number.

If you do test

A maximum attempt without a competent spotter and a graded warm-up progression is genuinely dangerous, and that is not a formality. On a bench press, a failed rep with no spotter can pin a loaded bar across the chest or throat. On a back squat, it can collapse under the load. Use a rack with pins set correctly, or a platform where the bar can be dumped safely, and have someone present who knows how to handle the lift.

Beyond that, work up in decreasing jumps with low repetitions, take three to five minutes between the heavy attempts, and stop after two failures. Grinding through a third rarely produces a number and reliably produces a bad session.

Testing is a clinician's or qualified coach's question if you have an existing injury, a joint or spinal condition, uncontrolled hypertension, or are pregnant or postpartum. The Valsalva manoeuvre that heavy lifting depends on sharply raises intrathoracic and blood pressure. That is a conversation to have in person with someone who knows your history, and no article can substitute for it.

Turning a maximum into training loads

The reason to have the number at all is that it converts into loads. Our percentage table runs from 100% down to 65%, which covers the range where percentage prescriptions are meaningful. For the lifter above, with an averaged estimate of 115.8 kg:

The repetition figures are typical capacities, not promises. Individual variation at a given percentage is wide: two lifters with identical maxima can differ by several repetitions at 80%, and that difference is stable enough to be a personal characteristic rather than a bad day. Percentages set the neighbourhood. Your own experience sets the address, which is why RPE and reps in reserve exist as a parallel system rather than a replacement for one.

How to use the number honestly

Read the estimate as a range, not a point. If the four equations put you between 112.5 and 117.5 kg, your working maximum is somewhere in there, and pretending otherwise adds precision that does not exist.

Re-estimate rather than re-test. A hard set of five, logged the same way every few weeks, tracks your strength perfectly well and costs you nothing. If the estimate is climbing, the training is working, which is the entire question progressive overload asks you to answer.

Recalculate after any long layoff. Estimates built on last spring's training do not describe this autumn's capacity, and the first heavy session back is a poor place to discover that.

Finally, keep the number in proportion. A one-rep max is a snapshot of force production on one movement on one day, sensitive to sleep, food, stress and how well you happened to brace. It is a useful instrument for choosing loads and a poor instrument for measuring your progress as an athlete. That work happens in the accumulated sets underneath it, governed by how many sets and reps you actually complete and recover from, and supported by enough food and protein to make the adaptation possible. If you are unsure whether your intake supports the training, the TDEE calculator and macro calculator are a reasonable place to start.