For years, reservoir engineers have used analytical or semi-analytical models to obtain quick estimates of steam drive performance. Although such models, which consist of little more than an energy balance, can give reasonable predictions of cumulative oil-to-steam ratio, they can also yield very poor forecasts of oil production vs time. The reason for the latter is the lack of coupling between reservoir heating and fluid flow near wells. To improve upon this drawback, a new steam drive model for personal computers has been formulated. This model can predict the performance of pattern steam drives with accuracies comparable to those of more expensive full-blown simulations. The formulation of this model is based on the material and energy balance across a moving steam front. The steam zone shape is determined by a modified van Lookeren theory which incorporated the effect of ky/kh. In this paper, the physical and mathematical backgrounds of this model are presented. Capabilities of the model are demonstrated by comparisons with simulator predictions for a wide range of input parameters: reservoir thickness, kv/kh, steam vent rate, pump location, initial oil viscosity, and well pattern. Comparisons with field data are also highlighted.