Planting cover crops after harvest of the main crop has become a key practice in temperate agriculture to reduce N leaching and increase soil organic matter. However, the growth of cover crops can be affected by adverse weather. Growing mixtures is thought to increase yield and reduce variation in productivity, but quantitative information on this subject is limited. Moreover, uncertainty remains on the optimal choice of species and mixture composition for cover cropping to obtain high cover crop yields and resilient performance under different conditions. Here we tested a broad selection of pure stands and mixtures of cover crop species in two years (2017–2018) at four sites: three sites in the Netherlands (Wageningen, Neer and Scheemda) and one site in Germany (Grundhof). All pure stands and mixtures were grown for a period of 11–13 weeks between August and November in each year. Aboveground biomass and N yield were determined. Yields in different treatments (unique pure stands or mixture compositions) in each site-year were regressed on the mean yields in each site-year to assess differences in responsiveness between treatments. Mixed effects models were used to estimate and compare yield variability in pure stands and mixtures at three levels: 1) between site-years, 2) between treatments and 3) between plots. This analysis was performed for biomass and N yield. Across all pure stands and mixtures tested, average biomass was greater in mixtures than in pure stands, but average biomass was similar when this comparison was made between the five highest yielding pure stands and the five highest yielding mixtures across all site-years. Thus, the lower mean productivity in pure stands was mostly due to some low yielding species. The five best mixtures had 9% higher N yield than the five best pure stands. The response of treatment yields to mean site-year yield was similar for mixtures and pure stands. Variation in cover crop yield over site-years was large in both pure stands and mixtures. On the other hand, mixing species significantly reduced the variability in biomass between treatments and between plots. However, when pure stands with low productivity were excluded, this difference in yield variability disappeared. This implies that the risk of choosing a sub-optimal cover crop is lower when a species mixture is used instead of a pure stand, unless the highest yielding species are known in advance. The results indicate that the positive effects of diversity on productivity and yield variability in cover cropping are restricted to reducing variability within the field and do not provide insurance against adverse conditions related to variability in growing conditions amongst sites and years.