Snow over the middle-high latitudes of Eurasia has been shown as an important forcing mechanism of climate variability, but observation stations in this region are sparsely distributed and recorded only until 1996. Based on three current widely used reanalysis datasets—the twentieth century reanalysis of the National Center for Atmospheric Research (NCAR-20th century reanalysis), supplied by the U.S. National Oceanic and Atmospheric Administration, the interim reanalysis of the European Centre for Medium-Range Weather Forecasts (ERA-Interim), and the Japanese 55-year reanalysis (JRA-55), supplied by the Japan Meteorological Agency—we evaluate the applicability of these three reanalysis datasets using historical soviet daily snow depth. Results show that the three reanalysis datasets can describe the spatiotemporal features of winter snow depth in the middle-high latitudes of Eurasia, especially the JRA-55 dataset, which shows the best consistency with station observations. There is 90% consistency between JRA-55 and station observations (174 stations), while the NCAR-20th century reanalysis shows 76% and ERA-Interim only 50%. Even on regional scales JRA-55 exhibits good consistency with stations in Europe and southern Siberia during 1961-1990; their positive correlation coefficients are 0.91 and 0.87, while for the NCAR-20th century reanalysis they are 0.77 and 0.32. JRA-55 exhibits interdecadal variation of snow depth over the middle-high latitudes of Eurasia: Snow depth is less in the 1960s but more in the 1970s; from the 1980s, it shows a decreasing trend that continues until the 20th century, and this decrease in snow depth is caused by changes in the high latitudes.