Every practice attempt across every subject feeds one engine, not several separate ones. Here's exactly what it does today: • Tracks real per-topic mastery using an Elo-style rating, not just a running percentage - so "good at fractions" actually reflects how they've done specifically on fractions questions, weighted by how hard each one was. • Brings back topics before they fade. Each mastered topic gets a review date based on a forgetting-curve model, so it resurfaces right when your child's about to start forgetting it, not months later, and not never. • Explains itself. Your dashboard shows the actual reason a question was chosen: because it's due for review, because it's a genuine weak spot, or because it's a confidence-building win they've already earned. Not just a score. • Flags patterns, not just wrong answers. If the same kind of mistake keeps showing up (like confusing perimeter and area), it's tracked as a misconception in its own right, not just another wrong tally. • Reaches into practice too. Arcade games and quick-practice modes read from the same mastery data, so a wobbly multiplication fact shows up more often in Honey Catch, not just in the next worksheet. This page gets updated as the engine changes - if a claim here stops being accurate, we'll fix it rather than quietly leave it.