Skip to content
Glossary

What Is FSRS? The Modern Spaced Repetition Algorithm Explained

May 25, 20264 minEmily Chen
Share:XLinkedIn
In this article
  1. 1.How FSRS works (in one paragraph)
  2. 2.Why FSRS beats SM-2
  3. 3.Where you see FSRS in 2026
  4. 4.Should you tune FSRS parameters?
  5. 5.A worked example of FSRS scheduling
  6. 6.How FSRS handles failed cards
  7. 7.Limitations and edge cases
  8. 8.Related reading
  9. 9.What FSRS actually does differently
  10. 10.Why FSRS produces better outcomes
  11. 11.How FSRS schedules each card
  12. 12.Where FSRS still has limitations
  13. 13.When to adopt FSRS (and when to wait)
  14. 14.How to set FSRS up in Anki
  15. 15.FSRS in other tools

Short answer. FSRS is the Free Spaced Repetition Scheduler — an open-source algorithm released in 2022 that decides when to review each flashcard or quiz question for optimal long-term retention. It replaced SuperMemo's SM-2 as the default in modern spaced-repetition tools, including modern Anki versions and [SimpleQuizMaker's review queue](/review).

How FSRS works (in one paragraph)

FSRS models each card with three values: difficulty (how hard it is for you), stability (how long the memory will last before forgetting), and retrievability (the probability you still remember it now). After each review, those values update based on whether you got it right and how confident you were. The algorithm then schedules the next review just before retrievability drops too low.

Why FSRS beats SM-2

SM-2 (the 1980s algorithm Anki used to use by default) treats every card with a single "ease factor." FSRS' three-variable model captures more of what actually happens in memory. Empirical studies show roughly 20-30% better scheduling efficiency — same retention with fewer reviews, or higher retention for the same review effort.

Where you see FSRS in 2026

  • Anki (default scheduler since recent versions)
  • SimpleQuizMaker review queue
  • RemNote
  • Several other modern spaced-repetition apps

    • If a tool markets "spaced repetition" but doesn't specify FSRS or SM-2, ask. The algorithm matters more than the UI.

    Should you tune FSRS parameters?

    For most users: no. Defaults are good. Power users with 10,000+ reviews can tune their personal parameters using FSRS Helper add-ons in Anki — but only after you have enough data for the tuning to be meaningful.

    A worked example of FSRS scheduling

    You see a new card on Monday. FSRS starts with default parameters:

  • Difficulty: 5 (medium-default for a fresh card)
  • Stability: ~1 day (until first successful review)
  • Retrievability target: 90% (you want to review just before memory drops below this)

    • You answer correctly with high confidence on Monday. FSRS bumps stability to ~3 days and schedules the next review for Thursday. If on Thursday you struggle and answer with low confidence, stability shrinks to ~2 days and difficulty rises to 6 — meaning the next review comes Saturday, sooner than if you'd answered cleanly. The algorithm tightens when you forget, loosens when you remember easily. Over weeks, well-known cards drift to month-long intervals while wobbly ones stay on a tight leash.

    How FSRS handles failed cards

    When you fail a card (rate it "Again" in Anki, mark it wrong in SimpleQuizMaker), three things happen:

  • **Stability drops sharply** — the algorithm now expects this memory to decay faster.
  • **Difficulty rises slightly** — this card is harder for *you* than the average user.
  • **The card re-enters short-interval rotation** — usually 1-10 minutes before the next attempt, then 1-2 days.

    • FSRS is not catastrophic about failures the way SM-2 was. SM-2 used to slash intervals dramatically on a single miss; FSRS adjusts more gracefully because the difficulty/stability separation lets it distinguish "I'm having a bad day" from "this card is genuinely hard for me."

    Limitations and edge cases

  • Cold-start problem: FSRS needs ~200-1000 reviews to personalise parameters well. New users get default parameters tuned on a large public dataset — good but not personalised.
  • Doesn't model "encoding quality": If you create a bad card (ambiguous answer, weak distractor), FSRS can't tell. It just sees that you keep getting it wrong and adjusts. The fix is on the card-creation side, not the algorithm.
  • Sensitive to lapses: Skipping a week and then drilling 500 cards in one session messes with the scheduling because the algorithm assumes your review-time pattern is consistent.

    • Related reading

  • [Spaced Repetition Guide](/blog/spaced-repetition-guide)
  • [Spaced Repetition vs Flashcards](/blog/spaced-repetition-vs-flashcards)
  • [What Is Active Recall?](/blog/what-is-active-recall)
  • [AI vs Anki: Do Modern Tools Beat the Gold Standard?](/blog/ai-vs-anki-modern-tools)
  • [Leitner System Flashcards](/blog/leitner-system-flashcards)

    • What FSRS actually does differently

    FSRS (Free Spaced Repetition Scheduler) replaces the older SM-2 algorithm that Anki and most spaced-repetition systems used for decades. The headline improvement: it learns your specific forgetting rate per card, then schedules the next review just before you'd forget.

    SM-2 used a fixed difficulty factor that updated with each review but treated all cards as following the same general curve. FSRS models three memory parameters separately:

  • Difficulty — how hard this card is to remember.
  • Stability — how slowly this memory is decaying right now.
  • Retrievability — probability you'd successfully recall the card if you reviewed it now.

    • The scheduler aims to review at the moment retrievability drops to your target (typically 90%). Earlier reviews waste your time; later reviews lose the memory.

    Why FSRS produces better outcomes

    Three measurable improvements over SM-2 in empirical studies:

  • 20-30% fewer reviews for the same retention. Less time spent reviewing.
  • Higher retention at the same review count.
  • More accurate scheduling — cards arrive when they need attention, not on fixed intervals.

    • For a heavy Anki user with 10,000 cards, that translates to ~30-60 minutes per day of saved review time, or alternatively, 20-30% higher retention with the same time investment.

    How FSRS schedules each card

    The math is non-trivial (a neural network trained on millions of reviews), but the user-facing behavior:

  • **First exposure.** Card enters; FSRS picks an interval based on initial difficulty (usually 1-3 days).
  • **You review.** You rate the card on a 4-point scale: again (forgot), hard (recalled but with effort), good (recalled normally), easy (instant recall).
  • **FSRS updates** its estimate of your specific stability and difficulty for this card.
  • **Next interval set** based on updated parameters. Strong cards get pushed further; weak cards get reviewed sooner.

    • The 4-point grading is important. SM-2 worked with this scale too, but FSRS uses the data more efficiently to update its per-card model.

    Where FSRS still has limitations

  • Initial scheduling for brand-new cards. Without prior data, FSRS uses defaults that may not fit your forgetting curve. Calibration improves after a few hundred reviews.
  • Multi-deck calibration. Different subjects have different forgetting rates. FSRS treats each card individually; some users see better results with per-deck parameter tuning.
  • Re-learn behavior after lapses. FSRS recovers from "again" ratings differently from SM-2. Users transitioning from SM-2 sometimes find this counterintuitive at first.
  • Overestimation in early intervals. Particularly for users with unusually steep forgetting curves, FSRS can occasionally schedule longer than is optimal.

    • When to adopt FSRS (and when to wait)

    Adopt now if:

  • You're starting a new spaced-repetition practice (no legacy data to migrate).
  • You're using Anki and willing to install the FSRS add-on.
  • You have at least 200 reviews of historical data for calibration.

    • Wait or stay on SM-2 if:

  • Your current SM-2 setup is producing good outcomes and you're not feeling the pain.
  • You're on a platform that doesn't support FSRS yet.
  • You're a casual user (under 50 cards reviewed per day) — the gains are smaller at low volume.

    • How to set FSRS up in Anki

    A practical migration:

  • Install Anki 23.10 or later (FSRS is built in).
  • Enable FSRS in deck options.
  • Run "Optimize FSRS Parameters" on your deck. This learns your per-deck forgetting curve from your historical data.
  • Set your retention target. 90% is a good default; lower (85%) reduces review burden, higher (95%) reduces forgetting but adds reviews.
  • Use Anki normally; FSRS schedules in the background.

    • FSRS in other tools

    The algorithm is open-source, so adoption is spreading:

  • Anki — native support as of 23.10.
  • RemNote — added FSRS as default scheduler in 2024.
  • Mochi — supports FSRS optionally.
  • Bookcomp / KumiBook — adopting FSRS variants.

    • Most modern spaced-repetition apps have either added FSRS or have it on their roadmap.

    Try a quiz with FSRS-scheduled review built in.

    Get weekly study & quiz tips

    Join teachers and students who get practical tips on quizzing, active recall, and AI-powered learning.

    Share:XLinkedIn

    Emily Chen

    Cognitive Psychology Writer & Study Skills Coach

    More articles by Emily →

    Practice with AI-generated quizzes

    Math Quiz Generator Coding Quiz Generator SAT Prep Quiz Generator

    Ready to create your first quiz?

    Use AI to generate quizzes from your own study materials in seconds.

    Try SimpleQuizMaker Free

    Related Articles

    Glossary

    What Is the Leitner System? The Original Spaced Repetition

    3 min

    Glossary

    What Is the Forgetting Curve? Ebbinghaus's Discovery, Explained

    4 min

    Glossary

    What Is Mastery Learning? Definition and Modern Use

    4 min

    Back to Blog