The LiftGauge dataset
Every published standards download on this page traces to an OpenPowerlifting source dump dated 2026-05-16. Its latest included meet is dated 2026-05-10. This registry separates source rows, contributing rows, lift observations, and full-power observations.
Why this page exists
Most strength-standards sites publish multipliers with no source. LiftGauge publishes three CC0 JSON downloads from one OpenPowerlifting source dump: the percentile cube, its companion metadata, and the strength-age curve. They are the files the calculators read. The dump date (2026-05-16) identifies the source file; the maximum included meet and dataset date (2026-05-10) identifies the newest competition result inside it.
1 · Competition percentiles
Squat, bench press, deadlift and total percentiles (p10–p99) by sex, bodyweight class, exact age or all ages, and equipment. The trailing 12-month window starts at 2025-05-10 and ends at the maximum included meet date 2026-05-10. After place, sex, equipment, bodyweight and date filters, 186,303 source rows are eligible; 186,218 contain at least one usable lift or total and contribute 615,721 row-weighted lift observations to percentile cells. Cross-lift cells use 141,828 eligible full-power rows. These are not unique-person counts.
- Source — OpenPowerlifting, CC0 public dataset. Source dump dated .
- Maximum included meet / dataset date — .
- Source CSV SHA-256 — e97d116bb707065ecb9e867cee5d99649baea36aa33dd023a745efb433742eac
- Filters — Wraps equipment and DQ/DD places are excluded. Approximate .5 ages stay in all-ages percentile cells but are excluded from exact one-year age bins. Raw, single-ply and multi-ply cells remain separate.
- Caveat — sanctioned-meet rows and lift observations only, not population prevalence. A person may contribute more than one meet row and more than one lift observation. Sparse cells widen only through the documented sequence, and the widening is disclosed.
- Download — /data/cohorts.json · build metadata /data/cohorts.meta.json
2 · Strength-age curve
Raw median DOTS is computed at every exact integer age from 18 to 80, per sex, from 441,789 row-weighted raw full-power SBD observations in the all-time OpenPowerlifting source. The published calculator curve is a centered arithmetic mean of each age's raw median and its adjacent-age raw medians; ages 18 and 80 use the two available ages. Approximate .5 ages, Wraps, DQ and DD rows are excluded. Smoothed peak: age 28 for men, 23 for women. Powers the Strength Age calculator.
- Source — OpenPowerlifting, CC0. Source dump dated ; maximum included meet date ; curve ETL date ; source CSV SHA-256 e97d116bb707065ecb9e867cee5d99649baea36aa33dd023a745efb433742eac.
- Curve method — Median DOTS within each exact integer age and sex; published in raw_medians. Centered three-age arithmetic mean of adjacent raw medians, computed separately by sex. The youngest and oldest published ages use the arithmetic mean of that age and its sole adjacent age; a single-age input remains unsmoothed.
- Caveat — the young peak is partly a population effect: raw competition has grown fast since 2015, so younger cohorts carry modern programming and larger samples. Age bins with fewer than 50 meet entries are flagged as thin inside the file; the 70+ tail is directional, not precise.
- Download — /data/strength-age-curve.json
Licensing
- All three downloads — cohorts.json, cohorts.meta.json, and strength-age-curve.json are released under CC0 1.0, matching the OpenPowerlifting source license.
- Free access and reuse — the Dataset schema marks the dataset as accessible without payment and declares CC0 at both dataset and download level.
Use the files as an API
The three downloads live at stable URLs and double as a versioned read-only JSON API — the same files the calculators on this site load. URLs never change between releases; the contents do, and cohorts.meta.json carries the exact release identity (dataset_id, source SHA-256, dates) so a consumer can pin or detect a release.
- GET /data/cohorts.json — percentile cells keyed sex|lift|class|age|equipment (p10–p99, mean, sd, histogram, n) plus trajectory and cross-lift blocks.
- GET /data/cohorts.meta.json — release identity, filters, counts, and integrity hashes for the cube.
- GET /data/strength-age-curve.json — raw exact-age DOTS medians, sample sizes, thin-bin flags, and the published smoothed curve.
- Access — fetch server-side or download; the files are not served with cross-origin headers, so a browser on another origin should proxy through its own backend. For a drop-in visual instead, use the embed widget.
- Reproduce — the ETL that builds these files is open source at github.com/arvidfcjarfalla-prog/liftgauge-strength-data, with a Zenodo release at DOI 10.5281/zenodo.21462463.
- Machine index — /llms.txt lists these endpoints and the most-cited pages for AI crawlers and agents.
Cite this dataset
For an article, a forum post, or a methods section:
LiftGauge Strength Standards Dataset, OpenPowerlifting source dump dated 2026-05-16, maximum included meet date 2026-05-10, released 2026-07-18, CC0 1.0. https://liftgauge.com/datasetFAQ
- Can I use this data in my own project?
- Yes. All three OpenPowerlifting-derived downloads are released under CC0 1.0. Cite both the source dump date and maximum included meet date, and keep the observation-count caveats attached.
- How often is the dataset updated?
- When LiftGauge processes a fresh OpenPowerlifting source dump. The current dump is dated 2026-05-16; its maximum included meet and dataset date is 2026-05-10; the LiftGauge registry release date is 2026-07-18. Those dates describe three different events.
- Why are there three downloads?
- The percentile cube contains the standards, the companion metadata records source-file identity, build-policy identity, dates, counts, filters and integrity hashes, and the strength-age file contains raw exact-age medians plus the published three-age-smoothed curve. All three derive from the OpenPowerlifting source dump dated 2026-05-16.
- What does n mean on LiftGauge pages?
- For percentile cells, n is the number of contributing row-weighted lift observations in that cell, not source rows or unique people. Cross-lift n counts eligible full-power meet rows. Strength-standards pages require at least 200 observations per bodyweight class; the percentile calculator widens sparse cells and says so.
- Where is the ETL code that builds the cubes?
- The full pipeline is open source at github.com/arvidfcjarfalla-prog/liftgauge-strength-data — scripts/build-cohorts.mjs and build-strength-age-curve.mjs together with these JSON files, under MIT (code) and CC0 (data). The scripts read the OpenPowerlifting source and emit the JSON published here; the source CSV's SHA-256 is recorded so a rebuild is verifiable.
Related: strength standards built from these cubes · percentile calculator for a live lookup · method for every formula · embed widget to put the tables on your own site.