The article aims to depict the political framing of three grading reforms in Swedish compulsory school, in terms of the political problem they are supposed to solve and what kind of attention is given to the lowest performing pupils. Discourse analysis is employed, focusing on statement producers. The empirical material consists of policy documents from the late 1930s to 2010. The analyses show that three cases of the same type of policy change, a new grading system, rely on very different problem representations. The changes were launched as an equality tool, an accountability measure and a remedy for declining results, respectively. The discourse about the least successful pupils differs. Reasonable demands and a ranking scale without a failing grade characterize the introduction of a norm-referenced system; a first criterion-referenced system rests on a belief that virtually all pupils will meet the formulated levels for passing, an expectation not met, and a changed focus behind the second criterion-referenced system normalizes that some pupils will fail compulsory school. The article also illustrates the merits of studying educational policy change through the theoretical lens of problem representations and directs attention to how reforms can have discursive effects as well as unintended side effects that matter substantially for some people.