A point forecast asks the wrong question
A single number for a single date compresses an enormous distribution of outcomes into one value, then discards the uncertainty that made the exercise interesting. Whatever the market does, the number is almost certainly wrong, and being wrong tells you nothing about whether the reasoning was sound.
Worse, point forecasts are nearly impossible to score fairly. Close enough is a matter of opinion, and the path — the drawdown along the way — is invisible in the final figure.
Classification is testable in a way targets are not
A regime call is a falsifiable statement: this market is transitioning out of a bull regime, within a defined horizon. It either resolves as stated or it does not, and every call can be dated, logged and scored after the fact.
That makes an honest track record possible. Each signal fire carries a horizon, a base rate and a resolution — the same structure every time, so the record can be audited rather than narrated.
What walk-forward testing changes
Any model can be tuned until it explains history. Walk-forward testing removes that luxury by refitting on an expanding window and judging each call against only what was knowable at the time — no peeking, no in-sample tuning of out-of-sample windows.
It is a deliberately unflattering protocol, which is why it is the right one. Robustness checks — parameter perturbation, sub-period sampling, feature ablation — do the same job from a different angle.
Where a range still belongs
Ranges and probabilities are not the same thing as point targets. A distribution with a stated confidence level communicates uncertainty instead of hiding it, and it can be evaluated on whether outcomes land where the distribution said they should.
The distinction is simple: publish what the model can defend, disclose how it was measured, and do not dress a guess as a number.
Frequently asked
Why are stock market price targets usually wrong?
A point target collapses a wide distribution of outcomes into one number for one date, discards uncertainty, and cannot be scored fairly afterwards. The format, not the analyst, is the main problem.
What makes regime classification more testable?
A regime call is a falsifiable statement with a defined horizon. Every fire can be dated, logged and resolved, which makes the record auditable rather than anecdotal.
What is walk-forward validation?
Walk-forward validation refits a model on an expanding history and evaluates each call using only information available at that time, preventing hindsight from inflating results.
Does RegimeSignal publish price targets?
RegimeSignal™ publishes regime classifications and, where a range is disclosed, the confidence level and measurement basis behind it. See the Disclosures page for the full statistical caveats.