Advanced forecasting of space weather requires prediction of near-Earth solar-wind conditions on the basis of remote solar observations. This is typically achieved using numerical magnetohydrodynamic models initiated by photospheric magnetic field observations. The accuracy of such forecasts is being continually improved through better numerics, better determination of the boundary conditions and better representation of the underlying physical processes. Thus it is not unreasonable to conclude that simple, empirical solar-wind forecasts have been rendered obsolete. However, empirical models arguably have more to contribute now than ever before. In addition to providing quick, cheap, independent forecasts, simple empirical models aid in numerical model validation and verification, and add value to numerical model forecasts through parameterization, uncertainty estimation and ‘downscaling’ of sub-grid processes.