The Stanford Watershed Model uses a hydrologic budget to model the land phase of the hydrologic cycle and thereby simulate streamflow on a continuous basis from climatological data and watershed parameters. Values for key parameters are estimated by trial-and-error. Subjective estimating differences have made it difficult to correlate parameter values with known physical characteristics of the watershed. This study sought to computerize a parameter optimization procedure based on the FORTRAN version of the Stanford Watershed Model known as the Kentucky Watershed Model. The resultant self-calibrating watershed model is named OPSET because it determines an optimum set of parameter values by matching synthesized flows with recorded flows. The first step in program development used sensitivity studies to determine which watershed parameter values are critical in simulating flow and are difficult to measure directly. The second step was to adjust estimates of the selected parameters in a way systematically improving flow simulation until the best possible matching is achieved. Independent adjustment schemes were used for parameters associated with runoff volumes, recession flows, and flood hydrographs. The third step was to empirically improve the model by applying it to a number of watersheds in Kentucky. OPSET estimates watershed parameters on a one water year basis, and values best describing watershed characteristics should be averaged from several OPSET-selected one-year-based values. (WRSIC abstract)