In a nutshell
In deck settings, set the New Interval under the Lapses tab to make the new interval of forgotten cards a percentage of their previous interval. Ideally, play with this percentage until you have an 80-90% success rate on re-learned cards.
Anki’s default behavior
I’m pretty sure that by default, what happens with Anki is that when you forget a review card, it goes into the relearning cue, and after it exits, it enters the review queue again with an interval of 1 days.
In my opinion, this isn’t very efficient.
Imagine that you’ve been studying a card for 10 years. You’ve gotten that card right every time and its last interval was large; maybe 3 years. You haven’t seen this card in 3 years, and you get it again today.
This time, you can’t quite remember it.
By default, the card would then start over again with an interval of only 1 day. Sure, it’s ease factor might be really high, so those intervals would grow faster than they originally did, but chances are, after being reminded of the answer, you probably could have remembered it for another 3 years or so.
Then again, there’s also a chance that you’ve mostly hit good or something and the ease factor may not have changed much at all. Either way…
Starting over from stage 1 is a waste of time.
Luckily, there’s an easy way to make those initial intervals after a lapse longer.
Open up the deck options for a deck and go to the Lapses tab.
Here you’ll see an option to set the New Interval after a lapse to a percentage of interval it had before the lapse.
Anki uses the term lapse for review cards that you forgot.
By default, this is set to 0%. After a lapse, instead of getting an interval of 0 days, cards get the minimum interval set here. By default, that’s a 1-day minimum interval.
What’s an ideal setting?
I think the ideal setting for a new interval percentage is whatever gives you an 80 to 90% success rate on that next review. And my hunch is that that’s actually a pretty high percentage; maybe 75%.
My reasoning behind the 80 to 90% success rate is that I think that’s where the most rewarding experience is; the range where reviews are not too hard, but not so easy that you don’t feel a sense of accomplishment.
Anki doesn’t provide access to this statistic out-of-the-box. There’s no simple view that tells you the success rate you have on that first review after relearning a card. But it is possible to use an SQLite browser to search the reviews table in the database for this.
What you have to do is come up with a search that finds all review reviews where the previous review was a relearning review. I actually have an addon code that does this and then adjusts the new interval percentage up or down for me. Alas, it’s tied up with a bunch of other code in a big addon I cobbled together a little at a time. I’m working on separating that beast into smaller components, and I’ll try to post some code that does just this soon.
Although the manual refers to only three queues (new, learning, and review), there are actually four queues in Anki: new, learning, review, and re-learning.
The re-learning queue is for cards that you forgot (lapsed cards) and are learning again. This is very similar to the learning queue (cards that were previously new), but the learning and re-learning queues each have their own steps.