Granularity of Probability Estimates

Probability granularity is the resolution at which probabilities are expressed. Finer values can preserve informative distinctions, while unsupported digits can obscure uncertainty; the appropriate reporting depends on the model and use.

Granularity is the spacing of probability values a forecaster uses, such as tenths, five-percentage-point steps, or finer values. Coarse rounding can discard useful distinctions, but additional digits do not create information. A computed probability can legitimately have several decimals while uncertainty about its model remains large.

Match reporting to the evidence and decision without declaring one universal precision level. If a choice changes near a threshold, a rough bucket may conceal a consequential difference; if uncertainty is broad, extra digits can distract. Evaluate whether finer distinctions improve predictions on new cases rather than assume numerical detail causes calibration.

When to use it

When analyzing data; when building predictive models; when evaluating statistical claims; when quantifying uncertainty.

How it can help

Define the event, inspect whether distinctions matter to the decision, and compare rounded versus finer forecasts on common resolved cases. Communicate model uncertainty separately from numerical rounding.