“90% confident” is one of the most misread phrases in forecasting — it does not mean “this will happen, and we’re 90% sure,” and the difference matters once real money is involved.
A CI90 is a price range, not a single number: the band between the 5th percentile and the 95th percentile of every simulated outcome a forecasting model produces for a given horizon. The claim being made is about coverage: if the model is well-calibrated, about 90% of realized prices should land inside that band across many forecasts — not that any one forecast is 90% likely to hit its exact midpoint, and not that the direction (up or down) is 90% certain. The other roughly 10% of outcomes landing outside the band isn’t a failure — it’s exactly what “90%” means by construction.
If you mistake “90% confidence band” for “90% chance this exact price happens,” you’ll be surprised far more often than the number suggests — the band covering the outcome and the mean forecast being right are two different things, and a wide band that gets missed rarely is not necessarily better than a narrow band that gets missed more often, if the narrow one is honestly calibrated to be narrow. What actually matters is whether a tool’s stated confidence matches its real-world coverage.
Ask whether the tool publishes its own calibration — the actual share of past forecasts whose realized outcome landed inside the stated band, including the misses, not just an aggregate hit-rate. On Quantustik’s own committed TOP-20 backtest, measured against the model as served, 90% bands have covered roughly 90.6% of outcomes at the 3-month horizon, 90.9% at 6 months, and 90.2% at 1 year. All three sit within a point of the 90% target and inside the well-calibrated 85–95% range — and coverage still varies widely by individual stock, from the mid-30s to nearly 100%. A tool that won’t show you this kind of check, or that only shows a single flattering number, hasn’t earned the confidence it’s claiming.
This lesson is investor education, not personalized advice. A confidence interval describes coverage across many forecasts; it says nothing about a specific single outcome, and it is never a promise of profit.
Live example: as of the last scan, AAPL's 3-month CI90 band runs from $308 to $400. Neither end of that range is “the forecast” — the whole range is the forecast. See the full AAPL forecast for the mean estimate inside that range.
No. It means about 90% of realized outcomes are expected to land somewhere inside the band across many forecasts — it says nothing about where within the band, or about the direction of the move.
Not by itself. A well-calibrated 90% band is expected to miss about 10% of the time — that’s what the number means. What matters is whether the actual miss rate over many forecasts roughly matches the stated 90%, which is what a calibration check verifies.
Ask whether it publishes real calibration data — the actual coverage rate of its confidence bands against past outcomes, including the misses, not just a single flattering headline number.
Investor education only — not investment advice, and never a promise of profit. Every investment can lose value.