Coverage and confidence are not probabilities
Why a complete-looking score can still have thin evidence, and how to read the two labels.
Three questions that should stay separate
Availability asks whether enough scoring dimensions exist to publish a result. Coverage asks how much of the underlying input set is present. Confidence is a formula derived from that coverage. None of these establishes that the data is correct, that a prediction has been calibrated, or that a return is likely.
A dimension can be calculable while some of its inputs are absent. For example, a trend component can exist without a 200-day moving average. Looking only at the component’s presence would overstate the depth of the evidence.
Read the formula literally
In the current implementation, confidence = round(40 + 0.6 × coverage), clamped to 0–100, when a score exists. Coverage of 50 therefore produces confidence of 70. That transformation does not create an independent statistical test: it expresses the same availability information on a different scale.
Do not read “confidence 70” as “70% probability that this stock will rise.” The system does not attach a calibrated win rate, holding period or expected return to this number. Confidence is absent when no score can be produced.
Completeness and freshness are different
A fully populated financial record may refer to an older reporting period. A current price alongside it does not refresh the company’s accounts. Check the financial period, calculation time and price timestamp separately. Missing news also cannot prove there have been no important events.
When two observations use different input coverage, a change in score may reflect a change in evidence rather than a change in the business. Keep the component breakdown next to the headline score if you record a research history.
Use coverage to decide what to verify next
Low coverage should produce a narrower conclusion and a more specific next step. If volume inputs are missing, verify trading activity. If fundamentals are thin, open the company filing. A filled field is a starting point for verification, not a substitute for it.
- Record the unavailable inputs rather than treating them as zero.
- Distinguish an older fact from an unavailable fact.
- Compare observations with similar data coverage.
- Do not use confidence as position sizing advice.
Sources and evidence
The editorial policy explains AI assistance and the limits of these studies.
Continue your research
What the FINVERCO Score measures
A weighted evidence summary, with missing inputs, risk caps and a worked example.
A FINVERCO research checklist: price, score and value
Keep three different questions separate, then connect evidence into a company-specific thesis.