The development and application of Markov chain Monte Carlo methods based on the celebrated Metropolis-Hastings algorithm has played a major role in the recent advancement of statistical procedures. However, the approach has made relatively little impact on stochastic population modelling in spite of this field being central to many processes studied in biology, physics and chemistry. Attention is almost exclusively focussed on sampling and the construction of distributions and parameter estimates, with virtually no consideration being given to the structure of the process trajectories. This paper therefore tackles the inverse problem of determining whether there is an optimal method of selecting a proposal matrix corresponding to a given set of equilibrium probabilities, both in terms of trajectory characteristics and the coverage of the range of support. Using the discrete Cauchy distribution as a benchmark, simple and general random walk schemes are compared with the ‘small world’ scenario. It is then shown that these are all outperformed by a direct P-matrix approach based on a mass annihilation and immigration process.