AbstractPrecipitation forecasts from five global numerical weather prediction (NWP) models are verified against rain gauge observations using the new stable equitable error in probability space (SEEPS) score. It is based on a 3 × 3 contingency table and measures the ability of a forecast to discriminate between “dry,” “light precipitation,” and “heavy precipitation.” In SEEPS, the threshold defining the boundary between the light and heavy categories varies systematically with precipitation climate. Results obtained for SEEPS are compared to those of more well-known scores, and are broken down with regard to individual contributions from the contingency table. It is found that differences in skill between the models are consistent for different scores, but are small compared to seasonal and geographical variations, which themselves can be largely ascribed to the varying prevalence of deep convection. Differences between the tropics and extratropics are quite pronounced. SEEPS scores at forecast day 1 in the tropics are similar to those at day 6 in the extratropics. It is found that the model ranking is robust with respect to choices in the score computation. The issue of observation representativeness is addressed using a “quasi-perfect model” approach. Results suggest that just under one-half of the current forecast error at day 1 in the extratropics can be attributed to the fact that gridbox values are verified against point observations.