Data and method

The source

Every percentile on this site is computed from the OpenPowerlifting database, a public-benefit archive of powerlifting competition results. The project contributes its data to the public domain, and you can download the same file we used and check every number here.

Our snapshot is dated 2026-07-25 and contains 3,991,443 results spanning 1964-11-08 to 2026-07-20, across 443 federations.

Rule 1 — one row per lifter, not one row per result

A lifter who competes thirty times appears thirty times in the raw file. Left alone, that pulls the whole distribution toward whoever competes most often, which is not a property of strength. We keep each lifter's best lift and the bodyweight they did it at, so 1,818,459 results become 580,559 people.

Rule 2 — raw only

Equipped lifting uses squat suits, bench shirts and briefs that add substantial weight to the bar. Mixing equipped lifts into a "standards" table inflates every number and tells nobody anything useful. We publish raw only, which here means the federation's raw category: belt and sleeves permitted, no supportive suit.

Rule 3 — successful lifts only

OpenPowerlifting encodes failed attempts as negative numbers. Anything at or below zero is dropped, as are results marked disqualified, doping-disqualified or no-show. A total is only counted when the lifter contested all three lifts in the same meet, because a "total" from a bench-only meet is not a total.

Rule 4 — no thin slices

A bodyweight class is only published when it has at least 200 lifters, and a page is only generated at 500. Below that the 95th and 99th percentiles are a handful of individuals rather than a distribution, and publishing them would be inventing precision.

Classes are 5 kg wide. Percentiles are interpolated between order statistics — the same definition R and NumPy use by default — and stored at every whole percentile from 1 to 99. Above the 99th and below the 1st we say the number is clamped rather than extrapolating a tail we have not measured.

What these numbers are not

They are not a sample of the general population, and they are not a sample of people who go to the gym. Everyone here entered a powerlifting competition, paid an entry fee, weighed in, and lifted in front of referees. That is a self-selected group and it is stronger than the gym-going public.

We could have quietly presented these as "average person" figures — the traffic would be the same and the numbers would look more impressive. We would rather say plainly what the population is. If you are at the 30th percentile here, you are still stronger than the overwhelming majority of people who have never competed.

The honest alternative does not exist: there is no random, verified sample of what people lift in commercial gyms anywhere in the world. What is usually published instead is data people typed into a training app about themselves, unchecked and unwitnessed.

Sex categories

The source records male, female and Mx (a gender-neutral category some federations offer). Mx entries are far too few to compute a trustworthy distribution from, so they are counted in the totals but not given their own percentile tables. We would rather publish nothing than a percentile built on a handful of people.

The formulas

1RM estimates use the published Epley, Brzycki, Lombardi, O'Conner and Mayhew equations, shown together because they disagree. DOTS, Wilks and IPF GL coefficients are the published ones, written out in full in the site's source. RPE percentages follow the widely used Reactive Training Systems chart, which is a convention of the sport rather than a measured constant, and the page says so.

Who publishes this

WhiteLights is built and maintained by Amit Sharma, who builds free, sourced reference sites and publishes the working rather than hiding it. Corrections are welcome and acted on — if a number here is wrong, say so and it gets fixed in the next build.