A first version of nationally covering hydrological statistics for Sweden based on the S-HYPE hydrological model for the period 1961–2010 is described. A key feature of the proposed method is that observed data are used as input wherever such data are available, and the model is used for interpolation in between stations. Short observation records are automatically extended by the use of the model. High flow statistics typically differed by about ±10% from observations. The corresponding number for low flow was about ±30%. High flow peaks were usually simulated slightly too low whereas low flows were too high. In a relative sense low flows were more uncertain than high flows. The mean flow was relatively certain. The annual maximum values were fitted to a Gumbel distribution, by the method of moments, for each subbasin. Flood statistics were then calculated up to a return period of 50 years. According to a Kolmogorov–Smirnov test, less than 1% of the fitted distributions were rejected. Most rejections occurred in regulated systems, due to difficulties in simulating regulation strategies, but also due to uncertainties in the precipitation input in the mountainous region. Results at small scale are very uncertain. The proposed method is a cost-effective way of calculating hydrological statistics with high spatial resolution.