SM-2 vs FSRS: Which Spaced Repetition Algorithm Should You Use?

Published October 12, 2026 · 7 min read

Short answer: SM-2 is the classic spaced repetition algorithm from 1987: one simple formula, the same for everyone, that works from your very first review. FSRS is a modern, open-source scheduler that learns from your own review history and predicts when you're about to forget each card. With a long review history, FSRS usually schedules more efficiently, so you need fewer reviews for the same retention. For a new learner or a vocabulary app, SM-2 is simpler, predictable and still very effective. The bigger factor in both is reviewing every day.

How SM-2 works

After each review you rate how well you remembered the card, from 0 to 5. A rating below 3 counts as a failure, and the card starts over with a 1-day interval. Otherwise the card moves on: 1 day after the first review, 6 days after the second, and after that the previous interval multiplied by the card's ease factor. Every card starts with an ease factor of 2.5. Easy answers nudge it up, hard ones down, but never below 1.3. A card you keep rating "good" goes roughly 1, 6, 15, 38, 94 days.

That's the whole algorithm. Its strengths are that it's predictable, needs no data about you, and you can check any interval yourself. Try it in the SM-2 calculator, or read the full SM-2 explanation. Its weakness is that the formula is the same for everyone: it doesn't know whether you, specifically, forget faster or slower than average.

How FSRS works

FSRS (Free Spaced Repetition Scheduler) describes each card with three numbers:

• Difficulty: how hard this card is for you.
• Stability: how many days it takes for your chance of recalling it to fall from 100% to 90%. Each successful review increases it.
• Retrievability: your estimated chance of recalling the card right now. It falls over time since the last review.

Instead of multiplying intervals, FSRS schedules the next review for the moment your predicted recall drops to your desired retention, for example 90%. Set it higher and you review more often but forget less; set it lower and you review less but forget more. The model's parameters can be optimized on your own review log, so the schedule adapts to how your memory actually behaves.

Which one is better?

In the open-source benchmark published by the FSRS project, which tests algorithms on large sets of real review logs, FSRS predicts whether a card will be remembered more accurately than SM-2. Better predictions mean reviews land closer to the moment you'd forget, so for the same retention you typically do fewer reviews. That matters most when you have thousands of cards and a long history.

The gap matters less when you're just starting: FSRS has little of your data to learn from, and SM-2's fixed early steps (1 day, then 6) are already sensible. And neither algorithm can save a habit of skipping days. Missing reviews hurts retention far more than the choice of scheduler.

Which should you use?

Choose FSRS if

• you're a long-time Anki user with a big collection and months or years of review history;
• your daily review load feels too heavy and you want the fewest reviews for a target retention;
• you like tuning settings such as desired retention.

SM-2 is a fine choice if

• you're new to spaced repetition or starting a new language;
• you want a schedule you can understand and predict;
• you'd rather spend your effort on adding words and reviewing than on settings.

Why WeaveWords uses SM-2

WeaveWords is built for vocabulary learners who want to start reviewing on day one. SM-2 works well without any history, behaves the same on web and Android (including offline on your phone), and every interval can be explained. In WeaveWords you rate Flashcards as Still Learning, Know It or Mastered, which map to SM-2 quality scores of 1, 3 and 5. Quiz answers are scored automatically. The part we automate is the one SM-2 doesn't cover: building the card, with an AI translation, example sentence and pronunciation for every word.

Coming from Anki with FSRS? Your words import from an .apkg file, but review history doesn't carry over, so cards start fresh on SM-2. See WeaveWords vs Anki.

Frequently asked questions

Is FSRS better than SM-2?

For predicting when you'll forget a card, yes: in the FSRS project's open benchmark on real review data, FSRS is more accurate than SM-2, which usually means fewer reviews for the same retention. The advantage is largest with a long review history; for beginners SM-2 works well too.

Does Anki use SM-2 or FSRS?

Both. Anki's original scheduler is based on SM-2, and FSRS has been available as an option in deck settings since Anki 23.10. When you turn it on, Anki can optimize FSRS's parameters on your existing review history.

What is desired retention in FSRS?

It's the probability of recall you want at the moment a card is due, for example 90%. A higher value means more frequent reviews and less forgetting; a lower value means fewer reviews but more lapses.

Does WeaveWords support FSRS?

No. WeaveWords schedules every word with SM-2, which works from the first review without any history and is easy to predict. You can see exactly how it calculates intervals in the free SM-2 calculator.