The SM-2 Algorithm Explained: How Anki Decides When to Show You a Card

Every time you review a flashcard in Anki, an algorithm decides how many days to wait before showing it again. That algorithm is SM-2, developed by Polish researcher Piotr Wozniak in 1990 and published in Acta Neurobiologiae Experimentalis in 1994. Understanding how it works helps you use it more effectively — and explains why the choices you make when rating cards matter more than most users realise.

The two variables: interval and ease factor

SM-2 tracks two numbers for every card:

Interval (I): The number of days until the card should next be reviewed. A new card starts with an interval of 1.

Ease Factor (EF): A multiplier representing how easy this card is for you specifically. Default: 2.5. Minimum: 1.3 (enforced floor). Maximum: in standard Anki, no ceiling.

After each review, the algorithm calculates the next interval:

• If you answered correctly (rating ≥ 3): I_next = I_current × EF
• If you answered incorrectly (rating 0–2): I_next resets to 1, and EF decreases by 0.2

And the ease factor adjusts:

• Rating 5 (perfect): EF increases by 0.1
• Rating 4 (correct with effort): EF unchanged
• Rating 3 (barely correct): EF decreases by 0.14
• Rating 2 (incorrect, easy to recall on second look): EF decreases by 0.2
• Rating 1–0 (complete failure): EF decreases by 0.2

A worked example: what the schedule actually looks like

For a new card with default EF of 2.5:

After 6 reviews over ~8 months, the card appears roughly once every 5 months. You have reviewed it 6 times total, and it is now in near-durable long-term memory.

If you rated this card "Easy" on every review (EF increases by 0.1 each time), the intervals grow faster — you review less frequently, which is correct if you genuinely find the card easy.

If you rate it "Hard" twice (EF drops to 2.1), the schedule becomes:

The card is reviewed more frequently because your recall performance suggests it needs more reinforcement.

What "Again" actually does

When you rate a card "Again" in Anki (complete failure to recall), two things happen:

1. The interval resets to 1 day
2. The ease factor decreases by 0.2

If you fail the same card on its next review: another decrease. A card failed five times has an EF of 2.5 − (5 × 0.2) = 1.5 — still above the 1.3 floor, but producing much shorter intervals. A card that reaches the 1.3 floor and keeps failing is a "leech."

Leeches are always a card design problem. A card that you fail repeatedly despite many reviews is either:

• Containing too many facts (split it)
• Phrased unclearly (rewrite the front)
• Testing something you do not actually understand (go back to source material)
• Requiring a mnemonic you haven't built yet (add one to the back)

Do not keep reviewing a leech card as-is. Suspend it, redesign it, and re-add it.

The rating scale in practice

Many users misuse the rating scale. In Anki's current interface, the options are Again / Hard / Good / Easy.

• Again: You could not recall the answer at all
• Hard: You recalled it but with significant difficulty, hesitation, or partial failure
• Good: You recalled it correctly with moderate effort (this is the normal rating for most correctly-answered cards)
• Easy: You recalled it instantly with no effort — consider rating Easy sparingly, as it increases the interval aggressively

A common mistake is rating "Easy" too often. This rapidly increases intervals to weeks or months for cards that are not yet consolidated. The result: a long interval is scheduled before the memory is actually durable, causing failures at the next review that reset the card to short intervals — wasted effort.

The honest calibration: if you would not have produced this answer in an exam without seeing the card first, do not rate "Easy."

SM-2 vs FSRS: should you switch?

FSRS (Free Spaced Repetition Scheduler) is a newer algorithm available in Anki as an optional setting. It uses a machine-learning model with four memory parameters to predict retrievability more accurately than SM-2's simpler ease factor approach.

Published comparisons show FSRS produces slightly better retention (roughly 90–93% vs 85–90%) at equivalent review loads. For most learners, the practical difference is modest. FSRS is the better choice for long-term, high-volume decks (medical school, language learning). SM-2 is fine for medium-term study decks.

To enable FSRS in Anki: Tools → Settings → FSRS → Enable.

A simpler alternative: the WarpRead flashcard tool

Anki's power comes with complexity. If you want to explore spaced repetition without learning Anki's interface, the WarpRead Flashcard Tool provides a stripped-back experience: build atomic cards, enter focus mode for random-order review, and export as a standalone HTML file. It does not automate scheduling (that is Anki's job) but it is the right tool for building and previewing a deck before committing to a full SRS workflow.

For the theoretical foundation of SM-2 and how Wozniak arrived at it, the free Spaced Repetition course covers Lesson 4 with a detailed breakdown of the algorithm's derivation from the forgetting curve research.

References

• Wozniak, P.A., & Gorzelanczyk, E.J. (1994). Optimization of repetition spacing in the practice of learning. Acta Neurobiologiae Experimentalis, 54(1), 59–62.
• Murre, J.M.J., & Dros, J. (2015). Replication and analysis of Ebbinghaus' forgetting curve. PLOS ONE, 10(7), e0120644.
• Luo, J., et al. (2022). Optimizing spaced repetition schedule by capturing the dynamics of memory. IEEE Transactions on Knowledge and Data Engineering. (FSRS foundational paper)