Adaptive spacing algorithms use a student's study history to decide which item or skill to present next; well-known applications include Anki, SuperMemo, and Mnemosyne. Prior approaches to designing such schedulers fall into two categories: (1) approaches that statistically model human memory, recommending the item whose predicted memory strength is closest to a predefined target value, and (2) approaches that do not rely on an explicit memory model, instead training deep learning algorithms and evaluating them on simulated students. Traditional adaptive spacing algorithms rely on spacing/retrieval strategies but do not adapt to the memorization of skills (as opposed to individual items); extending adaptive spacing to skill-level memorization is the gap this related work identifies.