A measurement of the skill of a forecast. The Priestley skill score, P, applies to value forecasts (such as forecasts of temperature). Let fi denote a forecast, and let oi denote the subsequent observed value, with ō denoting the mean of a set of n observed values. Then P is given by The Brier score was introduced in 1950 to assess the accuracy of probabilistic weather forecasts (e.g. ‘There is a probability pi of rain tomorrow’). Denoting the actual weather outcome by oi (equals 1 if the event occurs, and 0 if it does not) the Brier score, B, is given (for n forecasts) by The smaller the value of B, the more skilful the forecasting procedure. If Bs denotes the Brier score for a standard forecasting system, then S, given by is called the Brier skill score.
Subjects: Probability and Statistics.