Aefing tries to bring each question back just before you would forget it. Too soon and you spend time on what you already know. Too late and you have to learn it again.
What happens after you answer
- A new question, and every miss, comes back the next morning at your cue time.
- After a right answer, Aefing estimates how well you still know the question and picks the day it should return.
- The better you know a question, the longer the gap grows, up to a year. A miss starts the climb again.
- An answer that is half right or more, such as two of three blanks or an essay marked Half, counts as a hard recall. It comes back sooner than a fully right answer would, and it is never counted as a miss.
Three right answers on three separate days mark a question learned. That rule is separate from the schedule, and it does not change.
The model behind it
The estimate comes from FSRS, the Free Spaced Repetition Scheduler. It is an open model of memory, also used by Anki. For every question you answer it keeps three numbers.
| Number | What it means |
|---|---|
| Stability | How many days until your chance of recalling it falls to 90 percent |
| Difficulty | How hard this question is for you, from 1 to 10 |
| Retrievability | Your chance of recalling it right now |
Each answer updates stability and difficulty. A right answer after a long gap raises stability the most, because remembering something you nearly forgot strengthens it most. A miss lowers it. A second answer on the same day counts for less, because the question was still fresh. Days are counted on your own calendar, in your timezone.
Aefing uses FSRS version 6 with its published starting values. Its math is checked against the reference implementation by the FSRS authors.
Your recall target
The question comes back when its retrievability has fallen to your recall target. The target is 90 percent unless you change it under Daily in Settings, anywhere from 80 to 97.
- A lower target spaces reviews further apart. You review less and forget a little more.
- A higher target brings reviews sooner. You review more and forget less.
Changing it reschedules every question you have.
What the research shows
Spacing out practice works. A 2006 review of 317 experiments by Cepeda and colleagues found that spaced study beats cramming, and that the best gap grows with how long you need to remember. Aefing is built on that finding.
FSRS was created by Jarrett Ye, then at the language app MaiMemo, and published with colleagues at KDD 2022 and in IEEE Transactions on Knowledge and Data Engineering in 2023. Its strength is predicting when you will forget. On an open benchmark of about 350 million reviews from 10,000 Anki users, it predicts recall more accurately than the schedulers it was compared with.
Better prediction should mean fewer wasted reviews for the same recall. We know of no controlled trial yet that shows FSRS users score higher on exams than users of a simpler spaced schedule. Most of the benefit comes from spacing and from testing yourself, which Aefing does either way.
How you can check it
Your retention checks measure what you kept, apart from practice. If the schedule works, your checks should land near your recall target.
Sources
- Cepeda, Pashler, Vul, Wixted and Rohrer, Distributed practice in verbal recall tasks, Psychological Bulletin, 2006.
- Ye, Su and Cao, A stochastic shortest path algorithm for optimizing spaced repetition scheduling, KDD 2022.
- Su, Ye, Nie, Cao and Chen, Optimizing spaced repetition schedule by capturing the dynamics of memory, IEEE TKDE, 2023.
- The open spaced repetition benchmark.