Vorhersagegrad

Vorhersagegrad is a German term used to describe the degree to which a forecast corresponds to observed outcomes. It serves as a general indicator of predictive power or forecast accuracy and can be applied in various disciplines, including statistics, data science, meteorology and economics.

In practice, the Vorhersagegrad is often expressed as a percentage or index that summarizes how well a set of forecasts matches what actually occurred. For binary forecasts (yes/no events), a simple interpretation is that the Vorhersagegrad equals the proportion of correct predictions. For probabilistic forecasts, the concept may be linked to calibration, discrimination or other performance metrics, sometimes by converting a probabilistic forecast into a thresholded decision and then measuring the resulting hit rate.

Calculation methods vary by context. Common approaches include accuracy, precision, recall, and the F1 score for binary outcomes; the Brier score or log loss for probabilistic forecasts; and calibration or ROC-based measures for assessing reliability and discrimination. Because the term is not rigidly defined in all fields, different communities may emphasize different aspects of forecast quality when referring to the Vorhersagegrad.

Limitations include sensitivity to base rates, the chosen threshold, and the specific dataset used for evaluation. A high Vorhersagegrad on one dataset does not guarantee similar performance elsewhere, and it may not capture all dimensions of forecast usefulness, such as lead time or economic value.

See also: forecast accuracy, predictive power, calibration, Brier score.