Data Quality & Methodology
How confident are we in this data?
This database holds 110,447 athletes and 349,823 competition results, gathered from many sources spanning nearly a century of bodybuilding.
Not every record is equally certain. A recent Mr. Olympia result is far better documented than an obscure regional show from the 1980s.
Instead of hiding that, we show it. Every athlete profile and contest result carries a confidence badge that tells you how well sourced and verified the record is.
The rest of this page explains exactly how those badges are calculated and where the underlying data comes from.
The four confidence tiers
Each record falls into one of four tiers. The badge and a short tooltip appear next to the data so you can judge it at a glance.
| Tier | Athlete records | Share |
|---|---|---|
| Verified | 0 | 0.0% |
| High confidence | 695 | 0.6% |
| Medium confidence | 105,971 | 95.9% |
| Needs review | 3,781 | 3.4% |
Verified sits at zero on purpose. It is awarded by a person confirming a record against a primary source, one record at a time, and that work has not started.
Nothing is promoted into it automatically, so a rising number there will mean somebody checked.
Medium holds nearly all of it, and that is the honest reading. For most athletes we know the contests they entered and where they placed, and little else: no birth year, no height, no written biography, and no record of which source the profile came from. Every one of those absences lowers the score, which is what the tier is for.
These figures are the badges themselves, counted. The same function that draws the badge on an athlete page produces this table, so the two are not separate opinions about the same record. Twenty-four profiles are a known exception: their division is worked out from their results when the page loads, which this count reads from a stored field that is still empty, so they show as Medium and are counted here under Needs review. Filling that field is the fix.
Sourcing is recorded per result, not per page: 334,866 of 349,823 results carry at least one named source row, and 179,121 of those are confirmed by two or more independent sources, 53.5% of the sourced set.
-
Verified
Awarded when a person has checked the record against a primary source, one of the governing-body or event-organiser sources listed below, and the record names it. It is the highest level of certainty we assign, and the only tier a score cannot reach on its own.
-
High confidence
Reached two ways: the record names a source we rate as trusted, or the career behind it is documented deeply and completely enough to stand without one. Depth is evidence about the record, not about where it came from, so a page can read High confidence and still say the source is unidentified. The tooltip tells you which of the two earned it. Nothing in this tier has been individually human-verified.
-
Medium confidence
A single medium-trust source, a missing structural field such as a classified division, or a profile we hold results for and not much else. The core facts are probably right and the detail around them is thin. Most of this database sits here.
-
Needs review
Almost nothing on file beyond a name, or one or more open reader reports. Treat these records with caution. A record can never read as Verified while a reader report against it is still open.
How confidence is computed
The tier is calculated from signals that already exist in the data. No scraping or external calls happen when you load a page; it is a transparent function of the inputs below, and every tier comes with human-readable reasons.
Source trust
Each source is rated by how authoritative it is. Higher-trust sources push a record toward High confidence; an unidentified source leaves it lower.
- Primary (highest trust): NPC News Online, IFBB Pro League official results, and Arnold Sports official results. These are the primary sources for the 2011-present era.
- Well-verified: Center Podium, plus Wikipedia and Wikidata biographical data.
- Historical / community: MuscleMemory, GreatestPhysiques and other community archives. MuscleMemory is the historical backbone for results predating reliable online official records.
- Unidentified: either no source was recorded, or the tag on the record names who touched it rather than where the fact came from. A profile marked as manually curated, with nothing named as checked, still counts as unidentified, because that is what it is.
Cross-source agreement
When two or more independent sources hold the same record and agree, confidence rises. A single source is fine; multiple sources that match is better.
Completeness
We check that the structural fields a record should have are actually present, for example whether the division is classified and whether the placement is known.
A result with a missing division or placement scores lower than a complete one. For athlete profiles it counts seven things: country, gender, primary division, a career start and end year, a birth date or year, a height or weight, and a written biography. How much of a career we hold counts too, under its own name, because twenty recorded contests is better evidence than one.
Human verification
A person checking a record against a primary source is the only thing that awards the Verified badge. However high a record scores, the score alone leaves it at High confidence.
Machine work gets a machine label. Where a field was worked out from an athlete's own contest records rather than read off a source, the tooltip says so, and that derivation earns the record no extra confidence.
Reader flags
Open reports from readers lower a record's confidence. The more open reports against a record, the lower it scores, and any open report prevents a Verified badge. This is how the community keeps us honest between formal reviews.
Data sources & priority
When sources disagree, we resolve conflicts by source priority. For modern results the official sources win; for older history we lean on the archives that have the records at all, enriched with biographical detail where it exists.
Table by table, with the field names, the ingestion rules and the identity-resolution process: About the database.
| Source | Role | Primary coverage |
|---|---|---|
| NPC News Online | Primary results source | 2011 to present |
| IFBB Pro League (official) | Primary results source | Pro-level results, modern era |
| Arnold Sports (official) | Primary results source | Arnold Classic results |
| MuscleMemory | Historical backbone | Pre-2011 history and deep archives |
| Wikipedia / Wikidata | Enrichment | Biographical detail, nationality, dates |
Spotted something wrong?
Every athlete profile and contest page has a Report an issue button. If a winner, division, country, year, placement, name, or image looks wrong, use it. You can tell us what is currently shown and what you believe the correct value is.
Reports go straight into a moderation queue. They are never applied automatically. A person reviews each one before any change is made.
In the meantime, an open report lowers that record's confidence badge, so other readers see that it is under question.
This is how the data keeps improving after launch. We would rather publish what we have, clearly labelled, and correct it openly than hold everything back until it is perfect.
To be candid: some obscure competitors may have little more than a name and a result.
Photos and biographical detail are missing for them because that information was never published anywhere, not because we overlooked it. Where a reliable source exists, we will add it.

