Dynamical model skill in forecasting precipitation is often limited beyond a two-week lead time, but such models tend to have more skill when predicting atmospheric variables. Previous research has attempted to exploit this by “bridging”, where statistical relationships are used to estimate future precipitation from dynamical forecasts of some atmospheric variable. One such method is to link precipitation with weather patterns. Weather patterns are generally a product of clustering daily atmospheric fields into a number of discrete states that collectively represent the most typical patterns of atmospheric circulation. Precipitation can be empirically estimated from forecasts of these weather patterns. Despite showing some promise for sub-seasonal forecasting applications in Europe, this technique is not particularly skilful over southern Australia. This can be attributed to large variability in how representative the weather patterns are of the actual circulation. However, when focusing on forecast periods for which the model is particularly confident about the upcoming weather pattern, there is greater skill in forecasting precipitation compared to periods for which the model is not confident about the upcoming weather pattern. This implies that such methods might be useful as a tool for identifying periods in which we have more confidence in the forecast.