Feature Idea: SRS-stage Ordering

HanziHero Feedback
1

This is not so much of a request as a “would be interesting.”

If reviews were ordered by item proficiency (srs-stage).

For example, if I have reviews set up in order of {components, chars, words, dicts}, then, while I’m reviewing components, I’d like to review the ones of the lowest srs-stage first then work my way to my more skilled components (so review components in the order of novice i, ii, apprentice i, ii, journeyman i, ii, etc.), then once I get to characters, have those reviews also ordered like that, and so on.

If I have reviews set up completely randomly, then have all of the items ordered in that way (all novice i items first and randomised, then all novice ii items randomised, and so on).

Like I said, it’s not a need at all, but my reasoning is because, let’s say I have a queue of 300 items for review one day, maybe I just want to review the ones of the lowest ranks first just in case I don’t feel like finishing my queue for the day. That way, I’m reviewing the most urgent items, whereas pushing off the more skilled items by a day is not as big of a loss.

QUESTION:

Does limiting a one’s review queue for the day automatically prioritize lower SRS-stage items? Like for example, if I would’ve had 300 reviews for one day, but I set my limit to 100, will the system choose the 100 items of the lowest SRS stage for me to review, or is just random at that point given the review date? Because I think that would be beneficial for someone limiting their reviews for the day.

1 Like
2

It currently does not. I think we should do something though.

One issue with prioritizing lower SRS-stage items is, let’s say you have a review limit of 100 a day and you do 20 lessons a day, you may end up always focusing on lower SRS without ever reaching items that are higher up, thus permanently “hidden” from one’s review queue since the 100 reviews are always taken up by lower SRS items.

Still, in terms of a forgetting curve, I think it makes sense to focus on the lower SRS items first. I’m open to suggestions on this though, but maybe a default or setting on getting lower SRS stage items first sounds like a reasonable idea :slight_smile:

1 Like
3

Maybe a balance between lower SRS stage, and post-due-date percentage.

For example:

  • ITEM 1 is novice i, and its due date is today, it should be up for review
  • ITEM 2 is master i, and its due date was yesterday, it’s not a priority, throw it toward the back of the line
  • ITEM 3 is master ii, and its due date was 5 days ago, which is now exceeding 10% of original delay time (i’m making numbers up btw), so this should be prioritized (at least in this case, prioritized over ITEM 2)
4

The FSRS team did some really deep analyses on the best order to review cards in (data is here along with discussion for those interested in details) and it actually turned up the opposite of expected. I assumed you do the hard cards first and the easy cards later, since they’re easier they can wait right? (and if I’m reading the above correctly this is what you’re basically asking for with young cards first).

Stats wise, apparently it’s significantly better to do the easy cards first, to get them out of the way so they never get the chance to become hard/forgotten cards you might get wrong. They’re quick to answer and then don’t bother you again for a long time so not doing them doesn’t save much. So instead of trying to grind the hard/forgotten cards more, you try to minimise the chance a card BECOMES a forgotten card in the first place. If one of those easy mature cards does become a leech going to cost you WAAAY more time in the long run than just answering it on time would have.

Which makes sense once you think about it leeches are always the worst cards. Would you rather answer a card on time and get it right once? Or skip answering it to save 5 seconds now but have to review it an extra 5 times later because it’s late and you forgot it?

1 Like
5

Hmm that makes sense. I know that admin talked about maybe(?) implementing FSRS (idk if that’s false memory), but that’d be cool to see in real time.