Yield loss analyses are critical to inform tactical and strategic decisions for crop health management. Yield loss measurement requires the quantification of (1) levels of injury caused by disease or pest, (2) actual (injured) yield, and (3) attainable (un-injured) yield. Reverse modelling allows reconstructing an object or a process from limited information combined with a mathematical model. We reconstructed attainable yields and derived yield losses caused by diseases in winter wheat using a simulation model (WHEATPEST) combined with field data generated by a network of experiments across France, where disease injuries and actual yields, but not attainable yields, were measured. The analysis covered 70 [Year x Region x Variety x Crop Management] combinations, from 2004 to 2008. In each combination, three simulation steps were taken in order to model: (1) actual yields, (2) attainable yields, and (3) yield losses associated with individual diseases. Simulated overall yield losses from combined diseases ranged from 0 to 4.2 t ha–1, and averaged 0.80 t ha–1. Septoria tritici blotch caused the highest mean yield loss of 0.66 t ha–1. Other diseases causing losses, in decreasing order, were FHB, leaf rust, stripe rust, and powdery mildew. The results highlight the contribution of varietal improvement to agricultural sustainability. Reverse modelling can be applied to other crops and diseases or pests, in order to estimate individual and combined diseases yield losses.