Deterministic codes need an a priori knowledge of the solution in order to obtain the convergence of the iterative process described previously (for
example, the calculation of mean cross-sections is a very delicate step which is generally system dependent). They have been developed in the past years, when computers were too slow. The predictions of such codes are limited but they can be useful to study perturbations to a given system. The main drawback is that the discretization of space, time and energy implies approx­imations. Moreover, they are very memory and computer time consuming if a realistic system is to be described, and, in practice, it is not possible to have a reasonably good description of a real 3D system.

Monte Carlo codes are very well adapted to the description of complex 3D systems. There are no approximations due to discretization. These codes allow very detailed representations of all physical data. With increasing computer speed, very precise results can be obtained for a system within a few hours. The same codes could be used to describe experimental set-ups and give precise predictions to which experimental results can be compared, improving confidence in the code, in the case of good agreement. Code reliability lies in the validity of cross-sections (which are directly taken from evaluations); if they are not correct, the results will also be wrong. Of course, this is also true of deterministic codes.