Using multiple high-resolution nested grids in the Advanced Regional Prediction System (ARPS), we apply the scaled lagged average forecasting (SLAF) ensemble technique to the explicit prediction of tornadic storms that occurred over Fort Worth, TX on 28-29 March 2000. The broad goal is to evaluate the viability of ensemble forecasting, particularly methodologies used operationally, at high (few km) versus moderate (order 10 km) resolution. We evaluate the sensitivity of forecast results to variations in model background field, type and amount of data assimilated (especially from WSR-88D radars), forecast start time and duration, and lateral boundary conditions. Direct comparisons are made of individual and mean forecasts against observed storm-scale fields (e.g., reflectivity), including an analysis of ensemble spread versus forecast error as well as objective forecast skill. We close by commenting on the potential of storm-scale ensembles to provide meaningful operational forecast guidance for convective weather.