FSRS Explained: How the Spaced Repetition Algorithm Decides When You Review

If you've used Anki recently, or any study app that "brings back cards right before you forget them", there's a good chance the scheduling behind it is FSRS — the Free Spaced Repetition Scheduler.

You don't need to understand the maths to benefit from it. But knowing roughly how it thinks will help you use it well: how to rate your answers, why some cards come back tomorrow and others in three months, and why skipping reviews hurts less than you'd fear.

The problem spaced repetition solves

In the 1880s, Hermann Ebbinghaus memorised lists of nonsense syllables and tested himself at different delays. He found that memory falls away quickly at first and then more slowly — the shape now known as the forgetting curve. He also found that each time you relearn something, it fades more slowly than before.

Put those two facts together and you get the core idea of spaced repetition:

1. Review something just as you're starting to forget it.
2. Each successful review makes the memory more durable.
3. So the gap before the next review can be longer.

Spacing isn't a niche trick. A review of more than 300 experiments by Cepeda and colleagues (2006) found that spreading study out over time reliably beats cramming the same amount into one session, and a later study (2008) showed that the ideal gap between sessions grows with how long you need to remember the material. And when Dunlosky and colleagues (2013) rated ten popular study techniques, distributed practice was one of only two rated "high utility" — the other being practice testing.

The hard part is the scheduling. When exactly is "just before you forget"? That differs for every fact and every person, and that's the job of the algorithm.

The old way: SM-2

For decades the standard answer was SM-2, an algorithm Piotr Woźniak created for SuperMemo in 1987. Anki was built on a modified version of it.

SM-2 is simple. Each card has an ease factor, starting at 2.5. After you first learn a card, it comes back after 1 day, then 6 days, and from then on each interval is the previous one multiplied by the ease factor. Answer well and the ease nudges up; answer badly and it drops, and the card starts over.

It works — millions of people have learned with it. But it has well-known weaknesses:

• The same rules for everyone. The starting intervals and multipliers are fixed, whether your memory for a topic is excellent or poor.
• "Ease hell". Cards you fail a few times end up with a low ease factor and come back far too often, long after you've learned them.
• It doesn't really model forgetting. It never estimates how likely you are to remember a card on a given day, so it can't aim for a target.

How FSRS works

FSRS came out of research by Jarrett Ye and colleagues on optimising review schedules — published at the KDD conference in 2022 — and was developed as an open-source project. Anki added it as a built-in option in version 23.10, in late 2023.

Instead of a fixed multiplier, FSRS keeps a small model of your memory for every item, built on three numbers:

Each time you review, FSRS updates these numbers based on how long it had been and how well you did:

• Remembering when R was low (you'd nearly forgotten) boosts stability a lot. This is the "desirable difficulty" that makes spacing work.
• Remembering when R was still high (you reviewed too early) boosts it only a little.
• Forgetting knocks stability down, though not back to zero — relearning is faster than learning from scratch.
• Hard items gain stability more slowly than easy ones.

Then it schedules your next review for the day your retrievability is predicted to drop to your desired retention — commonly 90%. With a 90% target, the next interval is simply the item's stability.

Why that's better in practice

Because FSRS models memory directly, it can be fitted to real review history. The default parameters were trained on a very large set of anonymised reviews, and with enough of your own reviews they can be tuned to you. In public benchmarks on real review data, FSRS predicts whether a card will be remembered more accurately than SM-2.

In day-to-day terms, students who switch tend to notice:

• Fewer reviews for the same recall, especially of material they already know well.
• No ease hell — a few failures don't condemn a card to daily reviews forever.
• Graceful catch-up after a break. Because FSRS knows how overdue each card is, it can prioritise what's genuinely fading.

How to rate your answers

FSRS learns from your ratings, so rating honestly matters more than anything else. Most apps use four buttons:

• Again — you couldn't recall it, or got it wrong. Use this freely; it's not a punishment.
• Hard — you got it, but it took real effort or you weren't sure.
• Good — you recalled it with normal effort. This should be your most common answer.
• Easy — instant, effortless recall.

Two common mistakes:

1. Pressing Hard when you actually failed. If you didn't really know it, press Again. Pressing Hard on a failure tells the model you remembered, and the card comes back too late.
2. Overusing Easy. It pushes intervals out a long way. Save it for items that are genuinely trivial for you.

Getting the most out of FSRS

• Review a little every day. FSRS can cope with missed days, but a short daily queue beats a huge weekly one.
• Don't cram reviews early. Reviewing long before an item is due barely raises stability, so you spend time for little gain.
• Fix bad items instead of failing them forever. If a card keeps coming back as Again, the problem is usually the card (too vague, too much on it) or the understanding behind it — not your memory. Rewrite it or go back to the source.
• Use questions, not just flashcards. Being tested is what strengthens memory. Any question you answer — flashcard, multiple choice or fill in the blank — can be scheduled the same way.

FSRS in DeepRevision

DeepRevision schedules every question you answer with FSRS — multiple choice, true or false, fill in the blanks, matching, flashcards and image labels — not just flashcards. The questions are written from your own notes, slides, PDFs and lectures, so what's being scheduled is exactly what's on your syllabus.

After each answer you say how well you knew it, and the next review is booked. Each day, Today's Review shows only what's fading, so clearing it means you're done for the day.

Wrong answers also go into your mistake analysis, so FSRS brings back the exact questions you got wrong, right before the correction would fade.