How this register measures

Every figure on this site is a direct reading from a public Roblox API, stamped with when it was taken. Nothing is self-reported, nothing is estimated, and nothing is interpolated between readings.

The census

Roblox group ids are sequential integers, which makes a full census possible: the register walks the entire id space in fixed segments and records, for every id, whether a group answered and what it said. The census frontier — the highest id region where live groups appear — is re-probed daily as new groups are created.

A group only becomes a page here once it has been successfully measured. The census queue behind the register is never presented as coverage.

Insert-on-change history

A new snapshot row is written only when a measured value actually changed. A re-reading that agrees with the previous one is recorded as a confirmation of that reading, not as a new point. This is why charts here are sparse: a dormant group is one honest point, not a flat line drawn through days it was never re-measured — and a chart is only drawn at all once there are at least two measured points.

Deletions, terminations, and what “gone” means

A group that stops answering is first marked as a candidate. It is recorded as gone only after a second independent read, at least 24 hours later, also returns Roblox’s explicit “invalid or does not exist” answer. Transport failures, rate limits and server errors never count as evidence of deletion. The same two-read rule applies to user accounts.

For terminated accounts, Roblox reports the status but not the date. This register records the date we detected the termination and says so; it never claims to know when the termination happened.

Ownership and name changes

Ownership transfers and renames are detected by comparing consecutive sweeps. The date shown is the detection date, which bounds the change between two measurements — the register never asserts the instant a transfer happened.

What search engines are offered

A group page is offered for indexing only once it has earned it: at least three measurements spanning at least fourteen days, or a measured size large enough that the page answers a question people actually ask. Pages that have not earned it still exist and still link onward — they are just not offered.

What is never published

No estimated numbers. No trend drawn through a single point. No page for an unmeasured entry. No accusation: nothing on this site calls a group a scam — where a trust observation is ever published, it will carry its evidence inline or not appear at all. Group descriptions and shouts are quoted as group-authored content, attributed, and are not statements by this register.

Corrections

Every reading carries its timestamp, so any figure can be checked against the public API it came from. If you believe a record here misstates what the API returned at the stated time, contact the operator and the reading will be re-taken and, if wrong, corrected with the correction noted.