How to read a stock forecast — a plain-language guide
A Quantustik forecast is several numbers meant to be read together, not any one of them in isolation. Reading order matters: start with the range, not the headline number.
1. Start with the CI90 band, not the mean
The CI90 band (the 90% confidence interval) is the honest starting point — it tells you the range of outcomes the model actually expects, not just its single best guess. A narrow band means conviction; a wide one means genuine uncertainty. Read expected growth and target price as the center of that range, never as a promise.
2. Check whether the band has historically held
A confidence band is only useful if it is honest. Per-ticker backtest CI90 accuracy tells you whether this specific name's past 90% bands actually contained 90% of outcomes — a ticker with poor historical coverage deserves less trust in its current band, even if the band itself looks reasonable. MAPE, separately, tells you how far off the point forecast has typically been.
3. Weigh the signal by its stated confidence
Model confidence tells you how strong the model's current view is — distinct from whether that view has historically been correct. A BUY signal with low confidence is a weak signal, not a strong one dressed up as BUY.
4. Size the position, don't just take the direction
Kelly size and recommended position size translate the edge and confidence into how much capital to actually risk — direction without sizing discipline is how a correct call still loses money. Beta and daily volatility (σ) explain why two tickers with the same expected return can warrant very different position sizes and stop-loss distances.
5. Read every number as probabilistic, never a promise
Every metric above describes a probability distribution or a historical average — none of them is a guarantee, and Quantustik is educational research, not investment advice. See are AI stock predictions accurate? for the honest, self-published answer.
The eight terms, in reading order
- confidence interval
- volatility
- beta
- expected growth
- mape
- model confidence
- kelly position size
- backtest accuracy
Frequently asked questions
What's the single most important number on a forecast?
There isn't one — the CI90 band, per-ticker backtest accuracy, model confidence, and position sizing are designed to be read together, not cherry-picked individually.
Should I trade on expected growth alone?
No. Expected growth is the mean of a probability distribution; trading on it alone ignores the spread (CI90 band) and how reliable that band has historically been for the ticker.
Where do I check if the model's confidence bands are honest?
The live calibration page publishes real CI90 coverage per horizon against a committed backtest, including the misses.
Educational research only — not investment advice.