Why we will never show a fake approval probability

The failure mode that breaks the number: a drug can have two clean, statistically significant Phase 3 trials and still get a Complete Response Letter; for a manufacturing (CMC) problem, a third-party facility inspection, or a labeling dispute that has nothing to do with whether the drug works. A “99% approval” tile cannot see that. The history is full of strong-data drugs that were delayed or rejected on issues no efficacy model captures.

So instead of inventing a probability, we show you the things that are actually true and verifiable:

• The verified facts, the FDA-set or company-guided date, the drug, the indication, each tagged by how we know it and linked to a primary FDA, SEC, or ClinicalTrials.gov source.
• Historical base rates: how this kind of decision has resolved (by therapeutic area, by cap size, after a prior CRL), with the sample size shown, not a confident point estimate dressed up as one.
• The run-up path; what the market is actually pricing into the date, not what we guess the FDA will do.
• Our own misses, we label price-only data, flag unverified items, and post corrections. A black box can't do that.

A fake 82% asks you to trust a model you can't inspect. Verified history and sourced facts ask you to think for yourself, and give you what you need to. That is the entire difference, and it is the one thing none of the “AI %” sites can copy without rebuilding their pipeline around the truth.