The evaluation and improvement of crop yield forecasting systems is important because many strategies respond to the expected levels of crop production. With a possible increasing frequency of abnormal climate conditions in the coming decades, it is also important that such forecasts accurately predict extreme yield losses. In their study, Ben-Ari et al. (2016), provide a transparent framework to (i) compare the ability of climatic and bio-climatic indicators; and model simulations to accurately predict extreme yield losses (ii) calculate probabilities of yield loss from indicator threshold values that can be used as decision tools and (iii) analyze the sensitivity of the results to a set of necessary methodological choices. In particular, this paper investigates extreme wheat and maize yield loss in France and in Spain. Best-identified indicators are monthly averaged maximum temperatures and precipitation in France and phenology based drought indices in Spain. Notably, single climate variables perform as well as – or even outperform – complex model and phenological indices. The results suggest that an explicit statistical methodology to assess extreme yield loss could be a useful contribution to improve the JRC-MARS crop yield forecasts. The relevance of accurate meteorological measurements is also pointed out, as single climate variables may be the most efficient and least costly alternative to anticipate severe yield losses. The framework may be applied to any other crop/country combinations.