Epley vs Brzycki: how accurate is an estimated 1RM?

Epley and Brzycki give the same answer at 10 reps and drift apart on either side. Below 10, Epley's estimate is higher, by up to about 4%; above 10, Brzycki's is. Both are accurate enough from a hard set of 10 reps or fewer, and validation studies find the rep count matters more than which formula you pick. Above 10 reps, every equation gets unreliable, and if you mostly run, it gets worse.

The two formulas#

Epley's formula started life in 1985 as a poundage chart, not a study. Brzycki's was published in a physical education journal in 1993. Written out, as a recent validation paper lists them:¹

• Epley: 1RM = weight × (1 + reps ÷ 30)
• Brzycki: 1RM = weight ÷ (1.0278 − 0.0278 × reps)

Epley's line is straight, so every extra rep adds the same 3.3% of the weight. Brzycki's curve steepens as the reps climb. Here's what each one assumes, as a percentage of your 1RM, for a set taken to the last rep.

What each formula assumes you can lift for a given number of reps

reps to failure	Epley (% of 1RM)	Brzycki (% of 1RM)
1	96.8	100
2	93.8	97.2
3	90.9	94.4
4	88.2	91.7
5	85.7	88.9
6	83.3	86.1
7	81.1	83.3
8	78.9	80.5
9	76.9	77.8
10	75	75
11	73.2	72.2
12	71.4	69.4

Computed from the Epley and Brzycki equations as written in Beia 2024, Table 1. The lines cross at 10 reps.

The gap is never large in this range. At 5 reps Epley puts you at 85.7% of your max and Brzycki at 88.9%, so 100 kg for 5 becomes 116.7 kg or 112.5 kg. At 10 reps both say 75%. At 12, Epley says 71.4% and Brzycki 69.4%. One quirk: raw Epley says a single rep is 96.8% of your max, which makes no sense, so hybridAF just returns the weight you lifted.

What the validation studies found#

The most useful comparison is Mayhew's 1995 study. 220 men across seven groups, from untrained students to football players and middle-aged lifters, did a real 1RM bench and then a set to failure with a lighter weight. Six equations correlated with the true max at r = 0.86 to 0.98. For sets of 10 reps or fewer, three equations overpredicted and two underpredicted, and Brzycki was the most accurate. Above 10 reps, three overpredicted and three underpredicted.²

Correlation is the number people quote, and it flatters the equations. An r above 0.9 says strong people get bigger estimates. It doesn't say your estimate is within a few kilos. The other studies are more useful on that.

Reynolds and colleagues tested 70 adults, aged 18 to 69, at 1, 5, 10 and 20 rep maxes on the bench and leg press. The load didn't fall in a straight line as reps went up. The 5RM predicted the true max best, and the authors' advice was blunt: don't use more than 10 reps in a linear equation.³ Mayhew found the same in 103 women: predictions were more accurate from sets of fewer than 10.⁴ In the squat, Brechue and Mayhew got estimates within 5% for 95% of 58 football players, using sets at 80% of 1RM.⁵

When you fit an equation to a specific group, it gets tighter. Macht built a bench equation for 60 moderately trained men that adds shoulder width to the weight and reps, and its standard error was 3.5 kg.⁶ That's a fair picture of the best case: a few kilos either way, from a hard set of low reps, on the lift the equation was built for.

Runners do more reps at the same weight#

None of the studies above tested distance runners, and that matters if you run. Richens and Cleather had eight weightlifters and eight endurance runners do leg press sets to failure at fixed percentages of their own 1RM.⁷

At 70% the runners managed more than twice as many reps. The difference was significant at 80 and 70%, though not at 90%. The authors concluded that standard rep tables predicted poorly, especially for the runners.⁷

Run that backwards. If you're a runner and you get 15 reps with a weight, Epley assumes you were lifting about 67% of your max. The runners in this study were doing about 20 reps at 80%, so for someone like them, 15 reps means a heavier share of their max than the formula thinks. The real max is smaller, and the estimate comes out too high. It's one small study on the leg press, with only men, and the authors note the runners may have under-performed on their 1RM because heavy loads were unfamiliar. Still, the direction is the useful part: the more endurance-trained you are, the less you should trust a high-rep estimate.

Why hybridAF stops at 12 reps#

The app and the Hybrid Score calculator use Epley, and only on sets of 1 to 12 reps. A set of 15 or 20 is ignored for the estimate, so a high-rep finisher can't fake a strength PR. If a lift has no set in that range at all, the app falls back to the heaviest weight you lifted.

Twelve is our choice, and it's a little more generous than the research. The studies point at 10 or fewer. Between 10 and 12 the two formulas still sit within about 3% of each other, so the formula isn't the risk there. The risk is that a 12-rep set says less about your max than a 5-rep set does, especially if you run a lot. If you want the most honest estimate, give it a heavy set, not your last set of the day.

My recent bench sets, run through both#

Here are four bench sets I've logged since September, run through both formulas.

Eight estimates, from 104.6 to 112.3 kg, for the same lifter over seven weeks. That's a spread of about 7%. Some of it is the formula and some is the sets: both formulas assume the set went to the last possible rep, and I didn't log whether these did. The triple on 17 October says my max is above 102.1 kg. By how much is exactly what the formulas disagree on, and I haven't tested a true 1RM since, so I can't tell you which line is right for me.

The app grades my bench from my best estimate across all my logged history, not my latest. So these four sets don't move my score at all. That's how it works for everyone: the strength half keeps your best lift with no time limit.

Getting an estimate you can trust#

The practical version is short:

1. Use a set of about 3 to 6 reps, taken close to failure, on the lift you want to estimate.
2. Don't use a set above 10 reps, and if you mostly run, be wary of anything above 6.
3. Compare estimates over time on the same lift and rep range, not across rep ranges.
4. If two estimates disagree, averaging two equations is a reasonable fix; Mayhew suggested it.²

None of this replaces an actual max, but you don't need one often.

The way I put it to runners: a 1RM is your time trial under a bar. Testing your ceiling and building it are two different jobs, so test rarely and build constantly.

If you want the lift's whole history and the heaviest you've done at each rep count, looking up a lift shows where the app keeps it. And once you have a number, how strong should a runner be puts it against bodyweight standards. If you're wondering how close to failure those sets should be in the first place, that's RPE, RIR and failure.