One common problem of many atmospheric circulation models is the overestimation of the mean westerly winds in the mid-latitude North Pacific. This westerly wind bias is also a prominent feature in a recent version of the ECMWF model. Here we use the ECMWF model to investigate whether the use of stochastic parameterizations help reduce this error, using the concept of weather regimes. The focus is on the winter season when the atmospheric regime structure is most pronounced. It is shown that the operational version of the ECMWF stochastic physics scheme has little impact on the frequency of occurrence of North Pacific weather regimes. A recently developed scheme, however, which is based on combining a cellular automaton with a stochastic backscatter component, leads to substantial improvements in the simulation of the frequency of occurrence of North Pacific weather regimes and therefore a reduction of the westerly wind bias.