Selective literature survey and analysis of runoff statistics. Single brief records are inaccurate. Improved estimates are possible if single records supplemented with spatial correlations or regional consistencies. Recently developed complete rainfall-runoff models enable record to be extended, but results not always consistent. A comprehensive approach is suggested to replace infiltration theory with general retention theory, whence an improved rainfall-runoff model is developed. This improved model is demonstrated and tested with data from a sixty acre watershed.