Read a model’s evidence label

A live market and a model forecast answer different questions. The market says what price was available at a particular time. A model adds an estimate that needs its own evidence, including how it was tested and whether its inputs were available before the event.

Work through it

1. Identify whether the display represents a live price, a research ranking or a calibrated prediction. Keep those categories distinct.
2. Review sample size, evaluation period and held-out results. Check whether the method was selected after seeing the same outcomes used to advertise it.
3. Look for missing or stale inputs and the conditions under which the system passes rather than making a selection. A clear no-result state is part of a usable research tool.

Check your result

• The label has a documented meaning.
• Testing uses information available at decision time.
• Unavailable evidence remains visible.