Six methods for smoothing double-entry expectancy tables (tables that relate two predictor variables to probability of attaining a selected level of success on a criterion) were compared using data for entering students at 85 colleges and universities. ACT composite scores and self-reported high school grade averages were used to construct expectancy tables based on data for students entering each institution in 1969-1970. Tables were constructed using two levels of success--"C or better" and "B or better" first semester grade point averages. The tables were smoothed using each method and evaluated according to how closely the smoothed tables corresponded to observed data at the same institutions in 1971-72 and in 1972-73. The smoothed tables were more accurate than those based on 1969-70 observed relative frequencies. A linear regression of observed relative frequency on predictor value was most accurate; two extensions of an isotonic regression method were nearly as accurate. A commonly used regression method was found to be less accurate than most other methods. (Author)