We publish our accuracy.
Every analysis carries a confidence number, and every confidence number is a claim: setups scored 70–80 should resolve favourably about three-quarters of the time. This page checks that claim against what actually happened. Nothing here is selected, smoothed or held back.
Target under 5%.
of 6 tracked.
Nothing to publish yet. n = 0.
Until setups have been graded and their outcomes have settled, there is no curve — and showing one anyway is precisely what makes every other tool's accuracy claim worthless. Confidence numbers are published as the model's own uncalibrated estimate, labelled as such, until this table can replace them with a record.
- It is this device's record, not a fleet average. ORIN stores everything locally, so there is no pooled dataset to aggregate. A hosted version would publish the platform curve here and the personal one inside the product.
- A resolved outcome means the market reached the stop or the first target. Ambiguous bars that span both resolve as the stop — assuming the good fill is how backtests flatter themselves.
- Misses stay on the page. A band that reads worse than it claimed is the reason the page is worth reading.
ORIN grades setups; it never tells you to buy or sell. Decision support, not advice. Past resolution rates are a record of what happened, not a promise about what will.