FSRS stands for Free Spaced Repetition Scheduler. It is an open, free algorithm that decides when you should see each flashcard again. It has replaced older systems such as SM-2 in a growing number of flashcard apps, not because it is new, but because it schedules by probability instead of by a fixed rule.
This article is about the algorithm itself — what it models, how your answers change it, and what it does and does not promise. It is not a product tour; it explains the machinery underneath the review buttons.
Why older schedulers fall short
The classic approach, SM-2, dated from 1987. It gave each card an “ease factor” and grew review intervals by a fixed multiplier. It works, and millions of learners have used it. But it treats every card through the same lens: it has no way to tell how close you are to forgetting a specific word, so it cannot time the review precisely. The result is a schedule that is roughly right for the average card and wrong for both the easy ones and the hard ones.
The three numbers behind every card
FSRS models each card with three properties, and nearly everything the algorithm does follows from them:
| Property | What it means |
|---|---|
| Difficulty (D) | How hard this item is for you, not in general. Higher difficulty tends to mean shorter intervals. |
| Stability (S) | How long the memory will hold before your recall probability decays to a set level. Measured in days. |
| Retrievability (R) | The probability you would recall the card right now. It falls as time passes since the last review. |
The relationship is intuitive once you see it: R depends on how much time has elapsed and on how stable the memory is. A high-stability card stays easy to recall for a long time; a low-stability card slides toward forgetting quickly.
Target retention: the dial that sets everything
Because FSRS works in probabilities, it can schedule a review for the exact moment your recall probability drops to a target — commonly around 90%. That number is a choice, and it trades off reviews against forgetting:
- A higher target (say 95%) means more reviews and very few lapses.
- A lower target (say 85%) means fewer reviews and more forgetting.
Around 90% is a common default because it sits near the point where each extra review buys relatively little. You rarely set this yourself; the app picks a sensible value.
How your ratings tune the model
Every time you review, you grade the card — in EnjoyCards, on Again, Hard, Good or Easy. That single press updates the card’s difficulty and stability:
- Again (a blank or a wrong answer) lowers stability sharply, so the card returns within minutes.
- Hard raises stability slowly, keeping the next interval short.
- Good raises it normally.
- Easy raises it quickly, pushing the card out weeks.
FSRS is also adaptive over time. As it observes your review history, it can fit its parameters to your own patterns — how quickly you forget, and how much a failed recall costs you — so the schedule becomes personal rather than generic.
Why it saves reviews
Because it schedules to a probability rather than a fixed multiplier, FSRS avoids both failure modes of older systems: it stops over-reviewing cards you already know and stops under-reviewing the ones you keep forgetting. Benchmarks across large review datasets put FSRS at roughly 20–30% fewer reviews than SM-2 for the same retention. The saving is not in any single session; it accumulates as your deck grows.
What FSRS is not
It helps to be precise about the limits:
- It is not a card format. FSRS only decides timing. It says nothing about what belongs on a card.
- It is not a memory pill. The algorithm cannot rescue cards that test five things at once, passive re-reading, or dishonest grading. It schedules whatever you give it.
- It is not locked to one app. FSRS is free and open, which is why multiple flashcard tools have adopted it.
How you meet it in practice
You almost never touch D, S or R directly. You create a card, answer it, press a rating, and the schedule updates. That is the whole interface. In EnjoyCards, FSRS runs automatically and each rating button previews its next interval (10m, 2d, 15d) before you commit, so the algorithm’s decision is visible instead of hidden.
If you want the learner-facing version of this — intervals, the spacing effect, and a study routine — start with our spaced repetition guide. To see how FSRS sits inside a full app, read the walkthrough of FSRS in EnjoyCards. And if you have wondered whether daily streaks matter as much as spacing does, the comparison in streaks versus spaced repetition is worth a look.
The algorithm is only half of the story; the other half is showing up. Download EnjoyCards for free and let FSRS handle the timing while you handle the ten minutes.