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 and options-implied move — 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.