How spaced repetition works in Decks

Why does spacing out reviews work?

Memory fades on a curve: fast at first, then slower. Hermann Ebbinghaus measured this on himself in the 1880s by learning lists of nonsense syllables and timing how quickly he forgot them. He also found that spreading repetitions over several days took far fewer of them than piling them into one sitting.^[1]

The effect has held up for over a century. A 2006 review pooled 839 comparisons from 317 experiments and found spaced study beat massed study, with the best gap growing the longer you need to remember something.^[2] That last finding is why a spaced repetition scheduler stretches the gap each time instead of using a fixed one.

It applies to flashcards directly. In Kornell's 2009 experiments, studying one big stack of cards (so each card came back after a long gap) beat studying four small stacks, and beat cramming the day before the test, for 90% of participants. Yet after the first session, 72% of them believed the small stacks had worked better.^[3] Spacing feels harder, which is why people avoid it without a scheduler doing it for them.

Why flashcards instead of rereading notes?

Because pulling an answer out of memory strengthens it far more than looking at it again. In a study published in Science, students who kept testing themselves on foreign vocabulary remembered far more a week later; students who kept restudying words they had already recalled gained nothing from the extra study.^[4]

A 2013 review of ten common study techniques gave only two its highest rating, practice testing and distributed practice, and rated rereading, highlighting and summarising low.^[5] A flashcard with a scheduler is both of the top two at once: every review is a test, and the scheduler spreads them out.

What do Again, Hard, Good and Easy do?

Each card has an interval (days until it's due) and an ease (how fast that interval grows, starting at 2.5). Your rating changes both. For a card that's currently due every 10 days:

A new card first goes through two short learning steps, 1 minute and 10 minutes, before it graduates to a 1-day interval. Easy on a new card skips the steps and graduates it straight away. Ease never drops below 1.3, so even a card you've missed many times keeps getting longer gaps once you know it.

Rate one card Good every time and it goes 10 minutes → 1 day → 2 → 5 → 12 → 30 → 75 days: seven reviews to carry it from new to two and a half months.

How is this different from the original SM-2?

SM-2 was written by Piotr Woźniak for SuperMemo in 1987.^[6] The original grades each answer 0–5 and adjusts ease with a formula. Decks follows Anki's version instead: four ratings, fixed ease changes (−0.20, −0.15, 0, +0.15) and short learning steps for new cards. Two rules matter beyond that:

• Good and Easy always add at least one day. Decks rounds intervals down to whole days, which on its own can trap a card: at 1 day with an ease of 1.9, 1 × 1.9 rounds back down to 1, so a card you keep getting right would come back every day forever. A few lapses are enough to put a card there. Anki applies the same floor; a correct answer should always buy more time.
• Hard doesn't get that floor. Below 5 days, Hard can leave the interval where it is (4 × 1.2 rounds to 4). That's on purpose: “I barely got this” shouldn't have to earn a longer gap.

Why SM-2 and not FSRS?

FSRS is a newer scheduler, available in Anki, that fits a model to your own review history. It's a real improvement for people with large review histories. SM-2's known weakness is “ease hell”: ease that ratchets down after lapses and never recovers. When we measured ease on a real collection, the drift was mild, with the average close to the 2.5 starting value and many cards above it. If that changes, FSRS is the fix we'd reach for, rather than tuning SM-2's constants by hand.

References

1. Ebbinghaus, H. (1885). Über das Gedächtnis. English translation: Memory: A Contribution to Experimental Psychology (1913). psychclassics.yorku.ca/Ebbinghaus/index.htm
2. 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.org/10.1037/0033-2909.132.3.354
3. Kornell, N. (2009). Optimising learning using flashcards: Spacing is more effective than cramming. Applied Cognitive Psychology, 23(9), 1297–1317. doi.org/10.1002/acp.1537
4. Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319(5865), 966–968. doi.org/10.1126/science.1152408
5. Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students' learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. doi.org/10.1177/1529100612453266
6. Woźniak, P. A. (1990). Optimization of learning (Master's thesis, Poznań University of Technology). SM-2 algorithm description. super-memory.com/english/ol/sm2.htm

How Decks compares with Anki