Computer-aided crash reconstruction has become common-place in the automotive safety profession, primarily because of widespread distribution of software under public auspices. The SMAC (Simulation Model of Automobile Collisions) program, for instance, is available through NHTSA at nominal cost.This paper exhibits some of the limitations and strengths of accident reconstruction simulations, with illustrations and emphasis drawn from the SMAC program. In particular, some coarse physical approximations used and some coding errors incurred in the formulation of SMAC are discussed, together with their respective effects on the accuracy of prediction. Revisions of the basic SMAC coding have been developed at BYU to overcome these shortcomings. Results of uncorrected and revised SMAC simulations are demonstrated by comparison with the physical theory. Comments regarding a new SMAC program just completed under U.S. Government contract are presented where appropriate. Conclusions and recommendations regarding the use of SMAC and other similar programs are presented.