Various methods for tuning virtual population analyses (VPAs) using effort data have been tested on simulated data representing several sorts of fish stocks and fisheries.When there are no systematic changes in catchability, all the methods tested work reasonably well even in the presence of substantial random noise (coefficients of variation of the order of 50 %) in both the catch numbers and effort data. However, systematic trends in catchability may cause gross errors or failure of all current methods except the Rho method, and even this may fail if the trends are sufficiently strong. A new method is proposed which has a low bias and variability even in the presence of strong trends and noise, and which should be fairly widely applicable.