The Wang-Sheeley model is an empirical model that can predict the background solar wind speed and interplanetary magnetic field (IMF) polarity. We make a number of modifications to the basic technique that greatly improve the performance and reliability of the model. First, we establish a continuous empirical function that relates magnetic expansion factor to solar wind velocity at the source surface. Second, we propagate the wind from the source surface to the Earth using the assumption of radial streams and a simple scheme to account for their interactions. Third, we develop and apply a method for identifying and removing problematic magnetograms from the Wilcox Solar Observatory (WSO). Fourth, we correct WSO line-of-sight magnetograms for polar field strength modulation effects that result from the annual variation in the solar b angle. Fifth, we explore a number of techniques to optimize construction of daily updated synoptic maps from the WSO magnetograms. We report on a comprehensive statistical analysis comparing Wang-Sheeley model predictions with the WIND satellite data set during a 3-year period centered about the May 1996 solar minimum. The predicted and observed solar wind speeds have a statistically significant correlation (~0.4) and an average fractional deviation of 0.15. When a single (6-month) period with large data gaps is excluded from the comparison, the solar wind speed is correctly predicted to within 10--15%. The IMF polarity is correctly predicted ~75% of the time. The solar wind prediction technique presented here has direct applications to space weather research and forecasting. ¿ 2000 American Geophysical Union