The ability to self-regulate learning (SRL) is a skill theorized to transfer across learning environments. Students with this ability can consider a learning task, identify a goal, develop a plan to achieve it, execute that plan, and monitor and adapt learning until the goal is met. This paper examines the educational implications of developing the SRL expertise of high and typical-ability students, as operationalized by high school performance, who entered college and struggled with mathematics in their 1st year. Students who initially failed a 6-week intensive college math course completed a 3-h SRL training mid-semester and re-engaged in math learning with an adaptive problem-solving program and resources hosted on a course website. Students trained to evaluate tasks, plan, employ cognitive strategies, and monitor learning behaved distinctly from those who completed a math refresher course. Non-parametric, comparative analyses revealed that SRL-trained students more efficiently mastered math topics during digital problem-solving, demonstrating superior learning efficiency. Under a sequential explanatory mixed methods design, a follow-up multiple case study approach aligned to the Situated Model of SRL traced adaptive learning processes employed by multiple high-ability self-regulators and contrasted them with learning processes of exemplar learners from the untrained group.