FSRS Explained Without the Maths

FSRS (Free Spaced Repetition Scheduler) is an open-source algorithm that decides when to show you a flashcard again by modeling two things about that card: how long the memory is likely to last (stability) and how hard the card tends to be for you (difficulty). It fits that model to your own review history instead of applying one formula to everyone.

Short answer

• FSRS tracks stability and difficulty separately for every card, not one combined number.
• Stability estimates how long you can go before recall probability drops to your target.
• Difficulty estimates how hard this card is for you, independent of timing.
• Every review updates both numbers based on how you rated it and how late or early the review was.
• You set a desired retention target, and FSRS turns that target plus the two numbers into the next interval.

What FSRS actually is

FSRS is a scheduling algorithm, not a piece of software you install. It sits inside a flashcard app and answers one question for every card: given everything the app knows about how you've reviewed this card, when should it come back? It's maintained as an open-source project at open-spaced-repetition on GitHub, and it has been adopted by several flashcard tools because it schedules reviews more precisely than older formulas without asking you to configure anything by hand.

The name breaks down simply: Free because it's open source and free to use, Spaced Repetition because it's built around the spacing effect described in how spaced repetition works, and Scheduler because scheduling — deciding when, not what — is its only job. It doesn't pick which cards you study or write your flashcards; it only decides the gap between reviews.

The two numbers that drive everything: stability and difficulty

Older algorithms track one number per card, usually an "ease" value that grows or shrinks based on your last answer. FSRS splits that single number into two, because a card can be inherently hard and still hold well in memory once learned, or inherently easy and still fade fast if you haven't reviewed it enough times.

Stability is the number of days until your recall probability for that card is expected to drop to your desired retention target. A stability of 40 means: if you don't review this card, you have roughly your target chance of recalling it correctly 40 days from now. Stability generally grows every time you recall a card successfully, and grows faster the more times you've already reviewed it correctly.

Difficulty is a separate score for how hard a specific card tends to be for you, on its own scale, independent of how long it's been since you saw it. Two cards can have the same stability today and completely different difficulty: one is easy but new, so its stability will grow quickly with future reviews; the other is hard and long-studied, so its stability grows more slowly even with a correct answer.

Keeping the two separate is what lets FSRS react differently to "I got this right, but it was hard" versus "I got this right, and it was easy" — a distinction a single ease number can't represent well.

How a single review updates the model

Every time you review a card, FSRS updates both numbers using three inputs: the grade you gave (Again, Hard, Good, or Easy), the previous stability and difficulty, and how much time actually passed since the last review compared to what the algorithm predicted.

• A correct review on time or early increases stability, more so if the interval was already long, because a successful recall after a long gap is stronger evidence of a durable memory than a successful recall after a short one.
• A correct review rated Easy increases stability further and can lower difficulty slightly, since an easy correct answer suggests the card needs less caution going forward.
• A review rated Hard still counts as a pass, but increases stability less and can raise difficulty, since it signals the card is more fragile than the schedule assumed.
• A failed review (Again) resets stability closer to a short starting value and raises difficulty, because a lapse is the strongest signal the model gets that the previous schedule was too generous.

This is why two cards you rate identically today can leave the review session with very different next intervals: the model is reacting to the whole history behind that grade, not just the grade itself.

Why FSRS needs your review history to work well

A brand-new card has no history, so FSRS starts it with default stability and difficulty estimates and refines them as you review it. This is sometimes called the cold-start problem: early predictions on a card you've reviewed once or twice are necessarily rougher than predictions on a card with twenty reviews behind it.

The practical effect is that FSRS gets more accurate the longer you use it and the more reviews accumulate. It isn't tuned to a hypothetical average learner; it's tuned to you, card by card, using your own pass and fail pattern. That's also why importing years of review history from another tool can immediately sharpen your schedule — the algorithm has more signal to work with from day one instead of starting cold on every card.

FSRS vs. SM-2, in plain terms

Neither approach is "wrong." SM-2 is simple, predictable, and has scheduled flashcards for decades. FSRS trades that simplicity for a closer fit between what you actually forget and when the algorithm assumes you will.

What desired retention actually controls

FSRS doesn't decide an interval in isolation — it solves for the interval that keeps your recall probability at a target you choose, called desired retention. Set it to 90% and the algorithm schedules reviews so you'd correctly recall roughly 9 in 10 due cards if tested at that exact moment. Set it to 80% and reviews space out further, trading some forgetting for fewer total reviews.

There's no universally correct value. A higher target means more reviews and less forgetting; a lower target means fewer reviews and more forgetting along the way. The right number depends on how costly a forgotten card is for you — exam material warrants a higher target than a language deck you're keeping warm for a future trip.

How StudyDaily shows you this instead of hiding it

StudyDaily runs FSRS-6, the current version of the algorithm, on every card. Rather than compute the next interval behind the scenes, each rating button — Again, Hard, Good, Easy — shows the interval it will produce before you tap it. You see the consequence of a grade before you commit to it, which makes the stability and difficulty model visible instead of theoretical.

Reviews, statistics, and scheduling all run on the device, so nothing about your daily study depends on a network connection. If you're moving from another FSRS-based tool, StudyDaily's importer reads existing stability and difficulty values from a compatible export rather than resetting every card to a cold start — see what a migration actually preserves for the full picture of what an import can and can't carry.

What should you do next?

Open a deck with a mix of easy and hard cards and watch the interval preview change as you rate them differently. That five-minute test tells you more about how FSRS behaves than any description of the formula.

• Review a card you know well and note the interval Easy produces.
• Review a card you find hard and compare the interval Hard produces.
• Check your desired retention setting and adjust it if reviews feel too frequent or too sparse.
• Keep reviewing consistently — the model gets sharper with more history, not with a perfect first guess.

Questions people ask

What does FSRS stand for?

Free Spaced Repetition Scheduler. It's an open-source scheduling algorithm, not a standalone app — flashcard tools implement it to decide when each card comes due.

Is FSRS better than SM-2?

FSRS generally schedules more precisely because it separates stability and difficulty instead of using one ease factor, and it can fit its predictions to your own review history. SM-2 is simpler and has a long track record, but it reacts to your last answer alone rather than modeling memory directly.

What is "stability" in FSRS?

Stability is the number of days until your recall probability for a card is expected to fall to your desired retention target. A card with high stability can go a long time between reviews and still be likely to come back correct.

Does FSRS need a lot of reviews to work well?

It works from the first review, using default estimates, and gets more accurate as reviews accumulate for a given card. Importing existing review history from another FSRS-compatible tool gives the model a head start instead of starting every card cold.

Sources used for this guide

Algorithm mechanics described here come from the open-source FSRS documentation and wiki, maintained by the open-spaced-repetition project. StudyDaily's implementation claims come from the product's own FSRS-6 scheduler and importer, checked against the source documentation for this guide's 15 June 2026 publication date.