Many existing steel bridges are constructed by using hot-driven steel rivets as fasteners. It is known that the fatigue strength of these joints depends on their geometry, since it affects the load transfer mechanism and, therefore, the severity of the stress concentration. Attempts have been taken in the past to provide the characteristic fatigue strength of riveted joints based on test results, i.e. to provide a detail classification. However, the limited number of experimental results per geometry and their large scatter jeopardises the accuracy of the classification, which has resulted into different classifications by different researchers. This paper presents a theoretical fatigue strength prediction model for hot riveted double covered butt joints. The model results agree very well with fatigue test data and the model provides a fundamental understanding of the empirical observations. The model reveals that the classification of joints as proposed in the past is inaccurate for many cases. As an example, the plate width over rivet diameter ratio and the surface finish have a much larger influence on the strength than what has always been considered, whereas the ultimate tensile strength of the plate material has limited influence.