How to estimate your 1RM, and how far to trust the formulas

Lift 100 kg for 5 reps and a formula puts your max near 115 kg. Lift it for 20 and the formulas no longer agree on anything.

A one-rep max (1RM) is the heaviest load you can lift once with full range of motion. Most lifters never test it and estimate it from a harder set of several reps. This guide covers the four formulas behind most estimates, how far apart they land on the same set, what validation studies measured on real lifts, and when a real test is worth doing. To get a number for your own set, use the 1RM calculator.

How a 1RM estimate works

Every prediction equation rests on the same observation: the heavier the load, the fewer reps you can do before failure. Take a set to failure, count the reps, and the equation converts that count into a percentage of your max.

LeSuer et al. (1997) listed seven such equations. Four of them are still in common use:

• Epley (1985): 1RM = weight × (1 + reps / 30)
• Brzycki (1993): 1RM = weight / (1.0278 − 0.0278 × reps)
• Mayhew et al. (1992): 1RM = 100 × weight / (52.2 + 41.9 × e^{−0.055 × reps})
• Wathen (1994): 1RM = 100 × weight / (48.8 + 53.8 × e^{−0.075 × reps})

All four assume the set ended at or very close to failure. Stop a set with three reps left and every formula reads it as your limit, so the estimate comes out low.

Four formulas, one 100 kg set

The table applies each formula to the same 100 kg set at eight rep counts. The Overload column is the estimate the site's 1RM calculator returns: the weight itself for a single, the average of Epley and Brzycki from 2 to 6 reps, the average of Epley and Wathen from 7 to 10, Wathen alone from 11 to 15, and a gradual shift from Wathen to Mayhew from 16 to 30. The last column is the gap between the highest and the lowest of the four formulas.

Reps at 100 kg	Epley	Brzycki	Wathen	Mayhew	Overload estimate	Confidence	Gap between formulas
1	100	100	100	100	100	High	0 kg
3	110.0	105.9	109.0	114.0	107.9	High	8.1 kg
5	116.7	112.5	116.6	119.0	114.6	High	6.5 kg
8	126.7	124.2	127.7	126.3	127.2	Medium	3.5 kg
10	133.3	133.4	134.7	130.9	134.0	Medium	3.8 kg
12	140.0	144.1	141.5	135.4	141.5	Low	8.7 kg
15	150.0	163.7	150.9	141.7	150.9	Low	22.0 kg
20	166.7	212.0	164.5	151.2	160.0	Very low	60.8 kg

The formulas agree best between 8 and 10 reps, within 4 kg. At 3 and 5 reps the gap comes almost entirely from Mayhew, which runs 2 to 8 kg above the others. Epley and Brzycki, the pair the calculator averages there, stay within 4.2 kg of each other.

Past 10 reps the gap opens fast: 8.7 kg at 12 reps, 22 kg at 15, then 60.8 kg at 20, which is 38% of the estimate itself. Brzycki causes most of it. Its denominator shrinks with every rep and reaches zero at 37 reps, so its estimate climbs without limit. Mayhew and Wathen flatten out instead.

The single row has a trap too. Fed a single at 100 kg, the raw formulas return between 100 and 108.9 kg (Mayhew), a max above the weight you just lifted once. The calculator returns the weight itself for 1 rep for that reason.

What the validation studies measured

A formula can track the real max closely and still miss it by a fixed amount. In every study below, predicted and measured 1RMs were highly correlated (r above 0.95), and several equations were still off in the same direction every time.

Exercise	Study and lifters	What the equations did
Bench press	LeSuer et al. (1997), 67 untrained students, sets of 10 reps or fewer	Mayhew and Wathen within 0.8% of the measured max. The other five underestimated it by 1 to 6%.
Squat	Same study	Only Wathen did not differ significantly from the measured max. The others underestimated it by 2 to 8%.
Deadlift	Same study	All seven underestimated it, by 9 to 14%.
Lat pulldown and seated cable row	Pérez-Castilla et al. (2021), 12 men and 11 women, one set to failure at 80% of 1RM	Mayhew and Wathen underestimated the max by 2.1 to 6.7 kg. Estimates from bar velocity had no systematic error.

The deadlift result is the one to remember. If you estimate your deadlift from a set of 5, expect the number to sit below what you would lift in a real test.

Rep count changes accuracy more than any choice of formula. Reynolds et al. (2006) tested 70 men and women aged 18 to 69 at 1, 5, 10 and 20RM on the bench press and the leg press. The 5RM predicted the 1RM best (R² 0.993 on the bench, 0.974 on the leg press), and the authors advise using no more than 10 reps in linear equations. Mayhew et al. (2008) found the same in 103 young women, before and after 12 weeks of training: the equations were more accurate with fewer than 10 reps to failure.

The exercise matters more than the lifter. Nuzzo et al. (2024) pooled 952 tests to failure from 269 studies (7,289 people). Lifters did more reps on the leg press than on the bench press at every percentage of 1RM, while sex, age and training status changed the relationship very little. A single equation cannot fit both lifts equally well.

Estimating from RPE or reps in reserve

Most working sets stop before failure. The rating of perceived exertion scale based on reps in reserve (RIR) handles that: RPE 10 means no rep left, RPE 9 one rep left, and so on down the scale (Zourdos et al., 2016). Add the reps in reserve to the reps you did and you get the rep max the set stands for.

The site's RPE calculator works this way. A set of 100 kg for 5 at RPE 8 counts as a 7-rep max, for an estimated 1RM of 123.7 kg. If that set was an RPE 9, the estimate drops to 118.1 kg. If it was an RPE 7, it rises to 127.2 kg. One rep of misjudgment moves the result by 3.5 to 5.6 kg.

Lifters misjudge by about that much. In the meta-analysis by Halperin et al. (2022), 414 participants underestimated the reps they had left before failure by 0.95 reps on average. Predictions were slightly more accurate close to failure and in sets of 12 reps or fewer, and training experience did not improve them. Even at a true max, powerlifters rated the lift 9.6 to 9.7 on average rather than 10 (Helms et al., 2017).

Testing a real 1RM or estimating it

A real test is reliable. Grgic et al. (2020) reviewed 32 studies (1,595 participants): the median test-retest intraclass correlation was 0.97 and the median variation between two tests 4.2%, in trained and untrained lifters, men and women, young and older adults alike.

The risk sits mostly with beginners. Shaw et al. (1995) tested 83 adults aged 65.8 years on average on five exercises: 2 were injured (2.4%), both in the group with no weight-training experience, and 58% reported soreness afterwards. Gearhart et al. (2011) recorded no injuries in 49 adults in their sixties whose 1RM attempts were guided by an RPE scale.

Practical recommendations

• Estimate from a set of 3 to 10 reps. Around 5 reps is where Reynolds et al. (2006) measured the best prediction.
• Take the set to failure or rate it honestly with RPE. Expect to have about one rep more in reserve than you think (Halperin et al., 2022).
• Treat anything above 12 reps as a range. At 15 reps the formulas are 22 kg apart on a 100 kg set.
• Compare estimates on the same exercise and the same rep range. The trend over weeks means more than one absolute number.
• Add a margin on the deadlift, where all seven equations tested by LeSuer et al. (1997) came in low.
• To choose your working reps and loads from that estimate, see how many sets and reps per goal.

Conclusion

A 1RM formula is reliable in the range it was built on: sets of 10 reps or fewer, taken close to failure. On a 100 kg set, the four formulas stay within 4 kg of each other at 8 to 10 reps and end up 60.8 kg apart at 20. The bench press is predicted best, the deadlift is underestimated by 9 to 14%, and a real test repeats with a median variation of 4.2%.

This article presents the current scientific evidence. For a personalised program, consult a health professional.