← RectoVerso

Known
01

Recalling beats rereading.

Trying to recall an answer strengthens memory more than reading it again. In the classic experiments, students who were tested remembered more a week later than students who studied the material again.12 Every card is a test: the answer stays hidden until you ask for it.

02

Spacing beats cramming.

The same study, spread over days, is remembered longer than in one sitting. A synthesis of over three hundred experiments on recalling words and facts found spaced study ahead of massed study, and the best gap growing with how long you need to remember.3

03

Review just before you'd forget.

FSRS, an open scheduler, models each card's memory with three numbers: how hard it is, how stable it is, and how likely you are to recall it today. It sets the next review for when that likelihood falls to the retention you choose, 90 % by default.45 It's tested in the open on some 727 million reviews from 10,000 Anki users,6 and Anki has offered it since version 23.10. RectoVerso runs FSRS-6, checked against the reference implementation on 1,746 reviews, and refits it to your own reviews once there are enough.

100% 90% 80% 70% 60% 010203040 days
Fig. 1 The chance of recalling a card between reviews, as FSRS-6 models it. Each review comes when it falls to 90 %, the retention you choose (the red line), and each memory lasts longer than the one before. Dotted: without that review. Illustrative stabilities of 2.3, 7.5, 21 and 55 days.
Ours
04

Measured, not judged.

Most apps ask you to grade your own answer: Again, Hard, Good or Easy. RectoVerso doesn't. You open the phone, or pull the answer up, only as far as you need, then say Right or Wrong. The most of the answer you saw before saying it is the effort, and it's recorded with the answer.

05

From letters to ratings.

FSRS still takes one of its four ratings, and the effort picks it. Wrong is Again, at any depth. Right before a single letter is Easy, Right part-way is Good, and Right after 60 % or more of the answer is Hard. From there, FSRS schedules the card as it would any review.

nothing 60 % all Easy Good Hard Wrong, at any depth Again
Fig. 2 The rating a Right gets from the share of the answer seen first. In la bibliothèque, 60 % falls between the eighth letter and the ninth. A card still in its first steps gets Good instead of Hard, so it can finish them.
06

Not proven yet.

Letter cues are old: memory researchers have taught words by taking letters away as they're learned.7 Grading by the letters you needed is ours, and we haven't tested it in a study. So every review keeps its effort, how the answer was revealed and how long it took; FSRS's parameters are refitted to your reviews and kept only when they predict your answers better. If the mapping turns out wrong, it can be fitted again from what's recorded.

Sources
  1. Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255. doi:10.1111/j.1467-9280.2006.01693.x
  2. Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319(5865), 966–968. doi:10.1126/science.1152408
  3. Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380. doi:10.1037/0033-2909.132.3.354
  4. Ye, J., Su, J., & Cao, Y. (2022). A stochastic shortest path algorithm for optimizing spaced repetition scheduling. Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. doi:10.1145/3534678.3539081
  5. Su, J., Ye, J., Nie, L., Cao, Y., & Chen, Y. (2023). Optimizing spaced repetition schedule by capturing the dynamics of memory. IEEE Transactions on Knowledge and Data Engineering, 35(10), 10085–10097. doi:10.1109/TKDE.2023.3251721
  6. Open Spaced Repetition (read 2026-09-27). SRS Benchmark. about 727 million reviews from 10,000 Anki users. github.com/open-spaced-repetition/srs-benchmark
  7. Glisky, E. L., Schacter, D. L., & Tulving, E. (1986). Learning and retention of computer-related vocabulary in memory-impaired patients: Method of vanishing cues. Journal of Clinical and Experimental Neuropsychology, 8(3), 292–312. doi:10.1080/01688638608401320