This contribution addresses the issue of predictive policing in the framework of a common trend in many sectors towards algorithmic prediction and data-driven decision making. The setting of its definition, its categorizations and its genealogy will be followed by the analysis of the ideological postulates that are implicit in the implementation of predictive policing systems, with a special focus on the delimitation of the material scope of intervention resulting from the operational characteristics of these techniques. The reflection will be supported by the in-depth study of a practical case, the Crime Anticipation System in the Netherlands, which indicates the existence of direct tensions with several guarantee principles of the legal system. The final section will deal with the topic of consistency between predictive policing practices and the right to human intervention in automated decision-making recognised by the new General Data Protection Regulation (EU) No. 2016/679, highlighting at the same time the necessity of rethinking the role of the concept of predictive accuracy as a sufficient condition for limiting basic guaranties.