Sharpness
Sharpness describes how concentrated your forecasts are away from 50%. Higher sharpness means you often make confident calls, but it is only good if you stay well calibrated.
Definition
Sharpness describes how “bold” or informative your probability forecasts are, independent of outcomes. Forecasts clustered near 50% are low sharpness. Forecasts that often move toward 0% or 100% are higher sharpness.
Why it matters
In prediction, being calibrated is not enough. A forecaster who always says 50% can be calibrated in a trivial sense but provides little value. Sharpness captures whether you make differentiated calls.
Sharpness and scoring
Sharpness improves your score only when it comes with good calibration. If you are overconfident, sharpness will increase error and worsen Brier score.
How to measure it
Simple proxies include:
• the distribution (histogram) of your predicted probabilities
• average distance from 0.5 (for binary events)
Related
Sharpness is often discussed alongside resolution and reliability. For incentives to report true beliefs, see proper scoring rule.