Allan H. Murphy introduces a new vector partition of the probability score (PS), which decomposes it into three components: a measure of uncertainty, a measure of reliability, and a new measure of resolution. The original partition of the PS, formulated by Sanders, also includes measures of reliability and resolution but is not linearly related to the new measures. The new partition offers advantages in evaluating probability forecasts by providing a more comprehensive assessment of forecast attributes. The new partition is compared with the original in terms of their geometric interpretation, the attributes they measure, and their use in evaluating forecast performance. The new partition is shown to be more appropriate than the original, as it provides a better description of forecast resolution and allows for more meaningful evaluation of probability forecasts. The new partition is also represented geometrically within a simplex framework, which facilitates the interpretation of forecast accuracy and resolution. The paper concludes that the new vector partition of the PS is more suitable for evaluating probability forecasts than the original partition.Allan H. Murphy introduces a new vector partition of the probability score (PS), which decomposes it into three components: a measure of uncertainty, a measure of reliability, and a new measure of resolution. The original partition of the PS, formulated by Sanders, also includes measures of reliability and resolution but is not linearly related to the new measures. The new partition offers advantages in evaluating probability forecasts by providing a more comprehensive assessment of forecast attributes. The new partition is compared with the original in terms of their geometric interpretation, the attributes they measure, and their use in evaluating forecast performance. The new partition is shown to be more appropriate than the original, as it provides a better description of forecast resolution and allows for more meaningful evaluation of probability forecasts. The new partition is also represented geometrically within a simplex framework, which facilitates the interpretation of forecast accuracy and resolution. The paper concludes that the new vector partition of the PS is more suitable for evaluating probability forecasts than the original partition.