I’m vaguely working on a WaniKani app, and I want to make the reviews feel “just right”.
By “review queue algorithm”, I mean how WaniKani orders reviews when you’re doing them. It feels like it’s doing something behind the scenes to loosely “group” readings/meanings together, since otherwise in massive review queue’s you’d spend ages feeling like you aren’t making any progress.
I’ve poked around the JS files and it isn’t immediately jumping out at me (or I’m not looking in the right place).
First if all in the queue meaning and reading are just one item, and WK chooses which to show when the item comes up. Secondly WK has an “active queue” of 10 items. This means that you can have at most 10 partially answered items at one time. When you finish all questions for an item it is removed from the active queue and another item takes its place
I don’t think there’s an intentional mechanism. I suspect it’s a byproduct of introducing kanji with a common radical once the radical has been learned.
There’s that, but I more often manage to confuse things from completely different levels, like 那 and 郡 to pick a recent example (levels 51 and 55).
I think it’s just because I miss both halves of a confusion pair often enough that they both end up getting scheduled in the same review session eventually.