Brier Scores

Brier scores are used to assess the precision of probability predictions. For an event that can only occur in a set of mutually exclusive categories, the Brier score is the sum of the squared differences between the predicted probabilities that the event will occur in a specific category (numbers in the interval from 0 to 1) and the observed outcomes (1 if the event occurs in a specific category, 0 otherwise). Brier scores were originally proposed as a means to describe the precision of probabilistic weather forecasts (e.g., categories “rain,” “no rain”). Here, they were appreciated because they allow for a finer assessment of a forecaster's ability to generate accurate predictions than mere counts of numbers of correct predictions. For the same reason, Brier scores ...