We describe a deterministic approach to conformational searching. The algorithm is a best-first, depth-first traversal of the conformer graph, i.e., a graph generated by a small set of deterministic deformation operators. No deformation operator is ever applied to the same starting structure more than once. Thus, this algorithm avoids the source of diminishing returns, common in Monte Carlo searches, that is caused by the rising probability that a given starting structure is subjected to the same or similar perturbations more than once. We apply this technique to the task of finding all locally optimal conformations of cycloheptadecane whose MM2 energies lie within 3 kcal/mol of its global optimum. This task, which is considered to be challenging for contemporary algorithms and computer hardware, was used by Saunders et al. (J. Am. Chem. Sec. 1990, 112, 1419-1427) as a benchmark for comparing existing techniques in terms of their efficiency and thoroughness. The new algorithm compares favorably, by both criteria, with all of the methods tested by Saunders et al. Moreover, the conformer graph can be used for an analysis of the potential-energy surface that is not directly made possible by other search methods. The particular set of deformation operators used is specialized for medium-sized ring molecules. Ways to design deformation operators for larger molecules such as proteins are suggested.