A fast automated method is developed to estimate particle importance in the Los Alamos Monte Carlo code MCNP. It provides an automated and efficient way of predicting and setting up an important map for the weight windows technique. A short analog simulation is first performed to obtain effective group parameters based on the input description of the problem. A solution of the multigroup time-dependent adjoint diffusion equation is then used to estimate particle importance. At any point in space, time, and energy, the particle importance is determined, based on the calculated parameters, and used as the lower limit of the weight window. The method has been tested for neutron, photon, and coupled neutron-photon problems. Significant improvement in the simulation efficiency is obtained using this technique at no additional computer time and with no prior knowledge of the nature of the problem. Moreover, time and angular importance that are not available yet in MCNP are easily implemented in this method.