Current operational ensemble prediction systems (EPS) are mostly aimed specifically at the medium-range, beyond about 3 days, but there is also considerable interest in probabilistic information for short-range use, particularly for severe weather prediction. There is increasing evidence that for effective short-range ensembles it is important to include both quasi-random initial condition perturbations and also perturbed physics or multi-models. One relatively cheap approach to achieving both these requirements is to create an ensemble by combining the NWP output from a number of different operational centres. The Met Office is currently undertaking a project, with widespread international collaboration, to test such a "Poor Man's" EPS. Initial results from some low-resolution experiments will be presented, showing that this approach considerably out-performs the ECMWF EPS for probabilistic prediction at short range. Plans and progress for much more comprehensive tests will also be presented. It is hoped to include some preliminary results from higher resolution prediction of surface weather parameters.