The final piece of the card is the honesty check: a calibration badge, and a link to the live track record. A confidence number is only worth anything if the tool can show its past confidence matched reality — so this is the lesson that teaches you to hold any tool to account.
When the model draws a 90% confidence band around a forecast, calibration asks one question: how often does the real outcome actually land inside that band? The honest target is 90%. On Quantustik’s committed TOP-20 backtest (checked in, and measured against the model as served), 90% bands have covered 90.63% of realized prices at the 3-month horizon, 90.91% at 6 months, and 90.20% at 1 year. All three sit within a point of the 90% the bands claim, and inside the well-calibrated 85–95% range — which is what a coverage figure should look like when the stated uncertainty matches reality.
This is the single most important thing to understand. Coverage tells you whether the width of the uncertainty band is honest — not whether the forecast pointed the right way. A model can be perfectly calibrated and still be a coin-flip on direction: if it says “the price will land somewhere in this wide range” and it does, coverage is fine even though the call gave no directional edge. Never read good coverage as “the forecasts are 95% accurate.” They are two separate questions.
Honest calibration means showing the failures too, and the aggregate improving does not mean the individual names did. Coverage varies widely by stock: on the committed backtest, AVGO’s 1-year band covered only about 36.5% of outcomes, far below the 90% target; MSFT’s 6-month band covered 47.6%, and at 3 months MA sat at 55.6%, V at 57.1% and JNJ at 65.1%. The tighter 50% band is reported the same way: at 3 months it has covered about 52% of outcomes, close to its 50% target. A tool that only ever shows its best number hasn’t earned your trust; one that shows AVGO’s 36.5% alongside the aggregate has.
The backtest is history; the live track record is the forward test. It is a live, auto-updating aggregate of closed signal outcomes with both hits and misses shown — not a curated highlight reel. Because honest performance reporting requires a real sample before any claim is made, Quantustik publishes the outcomes as they accumulate rather than headlining a flattering early number. Weigh the committed calibration and the accumulating live record together — and treat any tool that won’t show you both with appropriate skepticism.
This lesson is investor education, not personalized advice. Calibration is band coverage, not directional accuracy; past calibration is not a guarantee of future results, and none of it is a promise of profit.
No. Coverage measures whether the 90% band is the right width — how often the outcome lands inside it — not whether the forecast pointed the right way. Band calibration and directional accuracy are two separate questions.
No, and that is the most important caveat on the page. The cross-ticker average lands near 90% at every horizon, but individual names on the same backtest run from the mid-30s to nearly 100%. Read the per-ticker figure for the name you care about rather than assuming the average applies to it.
No. Coverage varies widely by stock, and the weak cases are shown: on the committed backtest, AVGO’s 1-year band covered only about 36.5% of outcomes, MSFT’s 6-month band 47.6%, and JNJ’s 3-month band 65.1% — well short of target, and disclosed rather than hidden.
Investor education only — not investment advice, and never a promise of profit. Every investment can lose value.