Background: The potential of adaptive feedback in digital educational games remains largely unexplored. Fractions are a suitable topic for investigating the effectiveness of adaptive feedback, as the complexity of this domain highlights the need for adequate feedback. Objectives: This study examines the effectiveness of explanatory adaptive feedback in a digital educational game to address two particular misconceptions regarding fractions (i.e., Natural Number Bias and Unit of Reference). Methods: A total of 197 4th graders were randomly assigned to two different conditions, each playing a different version of a digital educational game--one with corrective feedback and one with explanatory adaptive feedback. During gameplay, we collected log data of students' item-wise correctness and misconception errors. Results: Explanatory item response analyses indicated that correctness improved in both game versions, with a more pronounced increase for the game with explanatory adaptive feedback compared to the game with corrective feedback. However, no decrease in misconception errors was observed in either game version. Moreover, neither the type of misconception nor students' prior fraction knowledge were moderating factors. These results suggest that adaptive feedback can support students in learning fractions; however, to reduce misconception errors concrete feedback should be optimised.