Spaced Repetition Explained: How FSRS Schedules Cards

The idea in one paragraph

You study a card, wait, and study it again. If you wait too short a time, the second visit is wasted effort because you still remember the card. If you wait too long, the card has faded and you start close to the beginning. Spaced repetition tries to land in the middle: the gap grows after each successful recall, and shrinks after a miss. Wikipedia’s overview describes it as a technique usually done with flashcards, where new and harder cards appear more often and older, easier ones less often.

Software did not invent this. In the Leitner system, which uses paper boxes, a correctly answered card moves to a less frequent box and a missed card goes back to the first box. A scheduler like FSRS does the same job with a calculation per card, so one hard word does not have to share a box with a hundred easy ones.

If you are learning English vocabulary, our post on how many words each CEFR level holds shows how big a list is before you schedule it.

The forgetting curve and what a replication found

The forgetting curve is a graph of how much you lose over time. Hermann Ebbinghaus described the first one in the 1880s, according to Wikipedia. In 2015, Jaap Murre and Joeri Dros published a replication in PLOS ONE. One learner studied lists of nonsense syllables and relearned them after intervals from 20 minutes to 31 days. The savings, meaning how much easier relearning was, came out close to Ebbinghaus’s original curve, with a possible small upward step around the 24-hour point that the authors say does not establish a cause.

Treat this as a picture, not a law for your vocabulary. It is one learner and meaningless syllables, and real words connect to things you already know.

A second source speaks to spacing itself. A 2006 review in Psychological Bulletin pooled 839 assessments from 317 experiments, and its abstract reports that the gap between study sessions and the time until the final test work together: the best gap grows as the wait for the test grows. That is the argument for gaps that lengthen as a card gets safer.

Older schedulers and FSRS

Anki’s older scheduler is SM-2. The Anki manual calls the Free Spaced Repetition Scheduler “an alternative to Anki’s legacy SuperMemo 2 (SM-2) algorithm” and says it aims to estimate more accurately how much you are likely to forget. It adds that Anki 23.10 supports FSRS. That is all we claim about which one is better: the manual’s own wording.

What makes FSRS different is the model underneath. The project’s own explainer says each card has a memory state of three variables:

So the scheduler asks a practical question for every card: when will your chance of recalling it drop to the level you chose? It then sets the next visit for that moment.

What the answer buttons do

After you flip a card, you tell the scheduler how it went. Cardday offers four buttons: Again, Hard, Good and Easy.

• Again means you forgot or got it wrong. The card returns soon and is treated as harder.
• Hard means you recalled it, but with effort. The Anki manual says that when you press Hard, FSRS assumes you have recalled the information correctly. If you actually forgot, press Again, because Hard after a miss pushes the next visit too far away.
• Good is the normal answer: you recalled it as expected.
• Easy means it was effortless, so the gap jumps further.

Here is one card across three visits. The numbers are round figures for illustration only, not output from FSRS or any app.

The pattern matters more than the numbers: success stretches the gap and a miss pulls it back.

Desired retention: the one setting to understand

Desired retention is the chance that you will remember a card when it comes back. The Anki manual says it controls how likely you are to remember cards when they are scheduled, and calls it the most important FSRS setting. The trade-off is workload. The manual says that as desired retention approaches 100%, the workload increases drastically. It sets 90% as the default and describes it as a good balance of retention and workload, and it advises keeping the value lower than 97%. The FSRS project’s explainer calls values between 70% and 97% reasonable.

In plain terms, a lower value means fewer visits a day and more forgotten cards. A higher value means more visits and fewer forgotten cards. Neither is right for everybody. Decide how much daily time you have, start near the default and change one thing at a time.

Common mistakes

1. Adding too many new cards. Every new card creates future visits. Ten new cards a day means about 300 cards in circulation after a month and 600 after two. Choose the number you can keep up on your worst day, not your best one.
2. Skipping days and then catching up in one go. A pile of overdue cards makes the next session long and discouraging. A shorter session on a busy day beats no session.
3. Pressing Hard instead of Again. As above, Hard tells the scheduler you recalled the card. Answer honestly.
4. Judging by one session. A single day tells you little. Look at a week.

How Cardday uses FSRS

Cardday is our widget-first flashcard app for English words with Turkish meanings. When you study a card, you open it, tap the speaker to hear the word, turn it over for the Turkish meaning and an example, and tap Again, Hard, Good or Easy. FSRS sets the next visit, and each button shows when the card comes back, so you see the consequence of your answer before you tap. You can undo your last answer with one tap, which helps when a finger slips.

A day counts toward your streak when you answer five different cards, and a Stats tab shows what is due today. Cardday has no account, so the words and your progress stay on your device. The free starter pack is 300 words of A1, and the Word of the Day widget is free. If you are curious where the word list and the levels come from, where a vocabulary app’s word list comes from explains it.

Questions

Are flashcards spaced repetition?

Not by themselves. Flashcards are the cards, and spaced repetition is the schedule that decides when each card returns. A pile you shuffle and flip through is flashcard practice. A scheduler that gives easy cards longer gaps and hard cards shorter ones is spaced repetition.

Is FSRS better than the older scheduler?

The Anki manual calls FSRS an alternative to its older SM-2 algorithm and says it aims to estimate more accurately how much you are likely to forget, so you can remember more in the same study time. We stay with that wording and make no stronger claim.

What should desired retention be?

It is the chance of remembering a card when it comes back. The Anki manual sets 90% by default as a good balance and advises keeping it below 97%, and the FSRS project calls 70% to 97% reasonable. Higher values mean more visits every day.

How many new cards a day?

Pick the number you can keep up on a busy day. Every new card adds future visits, so ten new cards a day means about 300 cards in circulation after a month. Start small, watch how long your daily session takes, then adjust.

Do I need an account to use spaced repetition?

No. The method is a schedule, and a schedule can live on your own device. Whether an app asks you to sign up is the app’s choice. Cardday has no account, so your cards and progress stay on your phone.