Learn
What does model confidence mean?
What does model confidence actually mean?
Short answer
Confidence describes how strongly and consistently a model's inputs agree with each other under current conditions. It is not the probability that an investment will be profitable, and a high-confidence estimate can still be wrong.
- Author:
- NeoInvest team
- Reviewer:
- Review pending
- Published:
- 2026-07-19
- Last reviewed:
- 2026-07-19
This article has not yet been reviewed by a named subject reviewer. It is published for transparency but excluded from search indexing until that review is complete.
What confidence is measuring
A forecasting model draws on several inputs at once. Confidence summarises how much those inputs agree. When they point the same way and the recent data is complete and stable, confidence is higher. When they conflict, or the data is thin or unusually volatile, confidence falls. It is a statement about the state of the evidence, not about the future.
Why it is not a probability of profit
This is the most common misreading, and it is an easy one to make because both are expressed as percentages. A forecast can carry high confidence and still describe a small, uncertain movement. Confidence tells you how much internal agreement sits behind the estimate; the estimated range and the direction probability tell you what the estimate actually says. Reading confidence as a chance of gain silently converts an analytical measure into a promise.
Confidence and range move together
A useful habit is never to read confidence on its own. A high-confidence estimate with a wide range is saying something quite different from a high-confidence estimate with a narrow one. The range describes the spread of plausible outcomes; confidence describes how settled the inputs behind that spread are. Both are needed before the number means anything.
Why confidence changes
Confidence is recalculated as inputs change. New data can bring inputs into agreement or pull them apart, and a stretch of unusual volatility will typically lower it even when the direction of the estimate has not changed. A confidence figure is therefore tied to the moment it was generated, which is why every NeoInvest forecast carries a generation time and an expiry.
Common misunderstandings
Not quite: 80% confidence means an 80% chance of making money.
It means the model's inputs are strongly consistent. Profitability depends on the estimated movement, your entry point, costs and what actually happens next — none of which confidence measures.
Not quite: High confidence means low risk.
Confidence says nothing about the size of a potential loss. A confident estimate of a volatile instrument is still an estimate of something volatile.
Not quite: Confidence should stay stable if the price does.
Confidence can fall while the price is flat, if the underlying inputs start disagreeing or the data quality drops.
Limitations
- Confidence is calculated from the inputs the model can see, and cannot account for events absent from that data.
- It is specific to the horizon it was generated for; one-day and ten-trading-day estimates carry separate confidence figures.
- It expires. An old confidence figure describes conditions that may no longer hold.
Sources and further reading
Educational information only. This article does not constitute investment, tax or legal advice, and does not take your circumstances into account.