Home
Editorial transparency

Accuracy

Ref Geek publishes officiating coverage on a deliberate pipeline: an AI writer drafts every piece from grounded stats, validators check for hallucination and banned phrasing, and reviewers approve high-stakes pieces before publication. Here's the public scoreboard for that pipeline.

Articles published
4
In reviewer queue
22
Awaiting human signoff
Rejected by reviewer
0
Pulled from the public surfaces
Reviewer signoff rate
100%
Of published, share signed off by a human

By content type

TypePublishedReviewedAwaiting reviewIn queueRejectedLast published
Pre-game scouting report44020050d ago
Post-game officiating report00000
Pre-series playoff breakdown00000
Post-series playoff review00000
The Week in Stripes00020

Recent corrections

No corrections logged. When an article ships with a factual error and a reviewer corrects it after the fact, that record shows up here — including which article and what changed.

How the pipeline works

  • Grounded.Every article is drafted from a structured Data Pack pulled from Ref Geek's aggregate tables — no free-form opinion.
  • Validated. Each draft runs through five checks: banned phrases, length band, tone, hallucination (every cited number must trace to the Data Pack), and name consistency.
  • Routed. Pre-game and post-game articles in the AUTO and STAGED tiers auto-publish; HIGH-importance pre/post-game and all weekly columns / playoff-series pieces require human reviewer signoff before publication.
  • Auditable. Articles that auto-publish without reviewer signoff carry a “Not yet reviewed” banner. Corrections and retractions log to a public audit trail.

Prediction-vs-outcome tracking (pre-game O/U calls vs actual totals, weekly column predictions vs follow-up data) lands when structured predictions are extracted from article bodies. For now this page is the transparency foundation: the pipeline is public.