Grade each flashcard by how the recall actually went, not by whether the answer was eventually right. A spaced repetition scheduler can only plan well if its input is honest, and inflated grades quietly produce a schedule that is too optimistic.
Key takeaways
- FSRS-based apps, including Cirello, use four grades: Again, Hard, Good and Easy.
- Again means you did not recall the answer. Good is the normal grade for a correct recall with reasonable effort.
- Use Hard when you got it but it was a struggle, and Easy only when the answer was effortless.
- Do not grade up because you "almost" had it or because you recognised the answer once you saw it.
- Honest grades make reviews harder in the short run and lighter in the long run.
Why grades matter
Grading is the step where you tell the scheduler how a recall went. For example, you recall a word but only after a long pause, so you choose Hard rather than Good.
In FSRS, each review is recorded as one of four grades (1 Again, 2 Hard, 3 Good, 4 Easy), and the review updates the model's estimate of the card: how difficult it is, how stable your memory has become, and how likely you are to recall it right now. We explain these in what FSRS is. The grade you give is the main signal. Mark a card Good when you actually failed it and the model schedules the next review too far away, so you meet the card again at a moment when you have forgotten it.
The research on practice testing explains why the recall itself matters: retrieving an answer is what strengthens memory, and feeling that you know something is a poor guide to whether you do (why rereading fails and practice testing works). Grading is how you report the result of that retrieval honestly.
The four grades, in practice
| Grade | What it means | Effect on the schedule |
|---|---|---|
| Again | You could not produce the answer, or produced a wrong one | The card comes back soon; in Cirello a card you blank on is offered again in the same session |
| Hard | You recalled it correctly, but with doubt or after a long effort (Anki's manual describes Hard this way) | A shorter gap than Good would give |
| Good | You recalled it correctly with normal effort | The standard gap; most correct answers should be Good |
| Easy | You recalled it instantly and it felt trivial | The longest gap of the four |
If you use Easy often, you may be making cards that are too simple, or grading too generously.
Common mistakes
- "I knew it, I just forgot for a second." That is still a failed recall at that moment. Grade it Again or Hard.
- Recognition instead of recall. (Recognition is knowing the answer when you see it, for example nodding at a revealed word.) Seeing the answer and thinking "yes, that is it" is not the same as producing it. Decide before you reveal.
- Grading to keep the pile small. Choosing easier grades to shrink tomorrow's queue only moves the work to the future, where it is larger.
- Ambiguous cards. If you cannot tell whether you were right, the card is the problem. See how to write good flashcards.
A quick habit
Before you reveal, commit to an answer in your head or out loud. Then compare. If the answer you committed to was fully correct, Good. If not, Again. Use Hard and Easy for the clear cases at the edges. This habit is a practical suggestion, not a research finding.
Grades and your workload
How often cards come back is also shaped by the retention target you set. Desired retention in FSRS explains the trade-off, and how many flashcards to review per day helps you size your daily session. If you are preparing for a test, combine this with the plan in how to study for an exam with spaced repetition. For the scheduler itself, see the FSRS app page.
Written by the Cirello team; see about Cirello and how we write for our editorial approach. Questions or corrections? Contact us at [email protected] or through the support page.
Sources
- FSRS project wiki: The Algorithm (grade values 1 to 4) and ABC of FSRS (difficulty, stability, retrievability)
- Anki manual: Studying (what Again and Hard mean)
- Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319(5865), 966-968. PubMed