Computer-simulated ligand binding or docking is a useful technique when studying intermolecular interactions or designing new pharmaceutical products. In general, the purpose of a docking experiment is twofold: (1) to find the most probable translational, rotational, and conformational juxtaposition of a given ligand-receptor pair, and (2) to evaluate the relative goodness-of-fit for different computed complexes. From a computational point of view, these are extremely difficult tasks and a satisfactory general solution to the docking problem has not yet been found. To explain this, let us consider the naïve approach in which a ligand is systematically moved relative to a given receptor. Here, the term “moved” must be understood as the combination of all possible translational, rotational, and conformational changes of the ligand. These operations define the so-called “docking box,” i.e., the a priori accessible phase space of the ligand, having dimensions 3+3+N (three translational, three rotational and N conformational degrees of freedom). While the six topological dimensions already seriously impede a docking simulation, the a priori conformational flexibility of the ligand and the receptor certainly poses the hardest (and least studied) problem. A large part of this chapter is devoted to a possible solution to this problem.