If you have ever crammed the night before an exam, done fine on the test, and then realized two weeks later you could not explain the material to someone else, you have already experienced the problem spaced repetition solves. Forgetting is predictable, and FSRS, the algorithm behind most modern flashcard scheduling, uses that to decide when to show you a card again.
Why forgetting is predictable, and why that matters
Memory researchers have known for over a century that forgetting follows a fairly predictable curve: sharp right after you learn something, then leveling off. The useful part is not the curve itself, it is what happens when you review something right before you would have forgotten it. Each successful review at that point makes the next gap before you would forget it longer. Review too early and you waste time on something you still remember fine. Review too late and you have to relearn it almost from scratch, which is close to as slow as learning it the first time.
Spaced repetition is this idea turned into a system: instead of guessing when to review something, an algorithm tracks each individual fact and tells you when it is close to the edge of being forgotten, so you review right at that point instead of too early or too late.
What FSRS is
FSRS (Free Spaced Repetition Scheduler) is an open, published algorithm that decides when each flashcard should come back to you. It became the default scheduler in Anki, one of the most widely used flashcard apps, replacing an older method called SM-2 that had been the standard since the 1980s.
The difference matters in practice. SM-2 gave every card one number, called an "ease factor," that adjusted up or down a little each time you reviewed it. It treated a well-known fact you rate "Easy" and a hard concept you keep struggling with using almost the same math, with a different multiplier. FSRS tracks two separate things for every card instead of one:
- Stability, roughly "how many days can pass before I am likely to forget this."
- Difficulty, roughly "how hard is this specific fact for me, independent of how long ago I learned it."
Because these are tracked separately, FSRS can tell the difference between a fact that is inherently hard but you have reviewed enough times to remember for a while, and a fact that is easy but you have only seen once so it is still fragile. SM-2 could not make that distinction.
What "90 percent" means and why it is the target
FSRS aims each review at roughly the point where you have about a 90 percent chance of still remembering the card. That number is a deliberate trade-off, not an accident. Scheduling for 99 percent retention would mean reviewing everything constantly, which wastes time on facts you already know cold. Scheduling for 50 percent would save review time but mean you forget half of what you are reviewing, which defeats the purpose. Ninety percent is close to the sweet spot researchers have found between total time spent reviewing and how much you retain.
This is also why occasionally forgetting a card and rating it "Again" is not a sign anything is broken. At a 90 percent target, a small fraction of reviews are expected to be misses. That is the system working as designed, not failing.
How your ratings change the schedule
Every flashcard review normally has four buttons: Again, Hard, Good, and Easy. What they do under FSRS:
- Again means you did not remember it. The card's stability drops, and it comes back to you soon, often within a day or two, so you can relearn it before it fades further.
- Hard means you got it, but it took real effort or you were not fully confident. The next gap grows, but less than a normal "Good" would grow it.
- Good means you remembered it without much strain. This is the default expected rating and produces a normal increase in the next gap.
- Easy means it was trivial, no effort. The next gap grows more than "Good" would, on the logic that a fact this solid does not need to come back as soon.
The one thing that hurts your schedule is rating dishonestly, specifically marking something "Easy" or "Good" when you did not know it and recognized the answer once you saw it. FSRS trusts your rating to update its model of how well you know that fact. A dishonest "Good" tells it your memory is stronger than it is, which pushes the next review further out than it should be, right when the card needed to come back sooner.
How to use flashcards with it
A few practical habits follow directly from how the algorithm works:
- Rate honestly, every time. If you had to think hard or got lucky, say so with "Hard," not "Good." The schedule only works if your input is accurate.
- Review on the days the app tells you to, not in one big batch. Doing all your overdue cards in one sitting once a week works, but doing a smaller number daily, as they come due, spreads the effortful recall out the way the algorithm intends and tends to feel less exhausting.
- Do not delete a card because you keep forgetting it. A card you keep marking "Again" is telling you it is hard for you. That is the card most worth keeping in rotation, and often the one worth understanding more deeply, not memorizing harder.
- New cards need patience. A brand new card has no history yet, so its first few intervals are short almost no matter what. It takes a handful of successful reviews before FSRS has enough data to space it out properly.
If you build flashcards in Scholarwire's Study tab, this is the scheduler running underneath every deck: each card carries its own stability and difficulty, and the app decides what is due today based on the math above, not a fixed daily quota. See how the whole thing fits together, or start free.
Spaced repetition assumes you have time to space things out. When you do not, the compressed version of the same principle is in how to study for an exam in five days, and getting the review blocks onto a real week is how to turn Canvas assignments into a weekly study schedule.