SOAR SCAN · calibration record
Every tool in this space tells you to verify what you are told. So here is ours, pointed at itself. These numbers are generated from the same database that serves the signals — not written by hand, not curated, and counted against us as readily as for us.
A signal fires when a pool's SOL reserve falls further than the price move can explain. Anyone who sees one can tell us it was wrong, from inside the terminal, in one click.
Those reports get read. If we investigate and agree the signal was wrong, it counts here as a false positive — the numerator is the cases we conceded, not the cases we dismissed. Reports about layout, wording or anything other than a computed verdict are not counted, because they are not what this measures.
We do not publish a rate until there are at least 20 signals and 5 settled disputes. A 0% false-positive rate across four signals is not a record; it is an absence of evidence wearing a percentage sign, and publishing it would be the exact kind of flattering statistic this page exists to refuse.
It measures signals we fired and someone disputed. It cannot measure drains we missed — nobody files a report about a warning that never arrived. Treat this as a ceiling on our precision, never as a claim about recall.
The drain detector is in beta. It shipped before it was calibrated, deliberately, and this page is what makes that choice honest rather than reckless.