Reep Register

Sponsorship

Sponsorship funds upkeep. Here is the upkeep.

The Reep Register is open infrastructure: complete, dated snapshots of the provider-ID crosswalk, free to everyone. Keeping it true is continuous work — cutting releases, repairing identity errors, adding coverage and measuring accuracy. Sponsors fund that work, and every claim on this page is a number the register produced about itself.

releases published6
entities maintained945,771
competitions covered443
corrections on record8,099

What support pays for

Four kinds of continuous work.

The register currently maintains 945,771 entities and 2,960,290 provider bridges across 53 provider surfaces and 443 competitions. None of it stays correct on its own.

Publication

Cutting and publishing releases.

Every release is generated, gate-checked, measured and promoted on a weekly rhythm. The dated snapshot is the artefact everyone downstream depends on, and it only exists because someone cuts it.

Repair

Finding and fixing identity errors.

Duplicate records get merged, wrong merges get split, and retired identifiers keep forwarding so nothing that ever pointed at the register breaks. Corrections are the ongoing cost of a register that stays true.

Coverage

Adding competitions, sources and history.

New competitions, deeper season history and additional independent sources widen what the register can corroborate. Coverage is grown by finding corroboration, never by lowering the evidence bar.

Assurance

Measuring accuracy, and hosting the result.

Precision and recall are estimated from independently verified random samples and gate each release. Bulk files and the API are hosted and served free to everyone.

Assurance

Accuracy is measured, not asserted.

Random samples are drawn from the release and independently verified. The interval is the honest range given the sample size — a small sample cannot prove perfection, and we do not present it as though it does.

Precision

100% 95% CI 96.3%–100.0% · n=100

The share of sampled entities that are correct: no two real people fused into one record. Drawn 19 Jul 2026.

Recall proxy

89% 95% CI 81.4%–93.7% · n=100

The share of sampled source records that reached the register rather than waiting for corroboration. Drawn 18 Jul 2026.

Corrections

2,164 plus 5,935 retired

Records merged into a survivor, with the retired identifier permanently forwarding. History is superseded, never deleted, so every repair stays countable.

The upkeep record

Every published release, on the record.

6 releases published since the publication ledger opened on 13 Jul 2026. This is the receipt for what sponsorship funds.

Published releases, most recent first.
ReleasePublished
2026-07-20 23:52Z 21 Jul 2026
2026-07-20 12:38Z 20 Jul 2026
2026-07-15 00:34Z 15 Jul 2026
2026-07-14 21:52Z 15 Jul 2026
2026-07-13 17:50Z 13 Jul 2026
2026-07-12 08:56Z 13 Jul 2026

Current release 2026-07-20 23:52Z, built from 27 sources across 407 competition slices. Figures on this page are generated from that release, not entered by hand.

How to sponsor

Three ways in.

GitHub Sponsors

Individuals and small teams

Recurring or one-off, self-serve, from a few pounds a month. The simplest way to support the register if you use it in your own work.

Invoiced agreement

Organisations

Clubs, federations, agencies, platforms and data vendors sponsor by agreement, with an invoice, a named contact and public acknowledgement. Get in touch and we will scope it properly.

Source access, allocation or compute

In kind

A licensed source, an API allocation or infrastructure directly buys coverage, and is acknowledged exactly like cash. Contributed identifiers are published only with explicit permission, and are held to the same evidence rules as everything else.

Independence

What sponsorship never buys.

Sponsors receive acknowledgement and project updates. Register decisions stay governed by the published evidence rules regardless of who funds the work.