What is FSRS?

What problem does FSRS solve?

Review a card too soon and you waste time on something you still know. Review it too late and you have already forgotten, so you are back at the start. The useful window is the moment just before you would forget, and that window is different for every card and every person. Ebbinghaus first charted this decline in 1885 with his forgetting curve, showing that memory drops off sharply at first and then levels out. The job of a scheduler is to place each review near the bottom of that curve, where one act of recall does the most to flatten it.

Fixed intervals cannot do this well. A card you find trivial and a card you keep failing should not come back on the same timetable, yet a one-size schedule treats them alike. FSRS estimates the right moment per card so easy material drifts far into the future while stubborn material stays close.

How does the FSRS memory model work?

FSRS describes your memory of each card with three numbers, and updates them every time you review.

• Retrievability: how likely you are to recall the card right now. It decays with time since the last review and is the value the scheduler watches.
• Stability: how slowly that retrievability falls. Higher stability means the memory lasts longer before it needs another review, and each successful recall raises it.
• Difficulty: how hard this particular card is for you. Harder cards gain stability more slowly, so they return more often.

After you rate a card, FSRS recomputes stability and difficulty, then schedules the next review for when retrievability is predicted to fall to your target, commonly around ninety percent. Because the model was fit to large amounts of real review data by the open-spaced-repetition project on GitHub, those predictions track how memory actually behaves rather than a fixed rule of thumb.

What does an FSRS review look like in practice?

Say you add the card front fastidious, back paying careful attention to detail. You see it, try to recall the meaning, then rate yourself. Rate it Good and FSRS might schedule the next review in three days. Recall it again easily then and stability climbs, so the gap stretches to a week or two, then a month, then longer. Each clean recall buys a longer rest. The first time you press Again, FSRS reads that as low stability for this card, drops the interval back to a day or so, and nudges difficulty up so it watches the card more closely from then on. You never pick the numbers; you just say how the recall went and the schedule reshapes itself.

How is FSRS different from SM-2?

SM-2 is the algorithm Anki used for years. It multiplies an interval by an ease factor and adjusts that factor up or down based on your rating. It is sound and simple, but it assumes a fixed shape for forgetting and leans heavily on one ease number per card. FSRS instead models retrievability, stability and difficulty separately and was tuned against real review histories, so it adapts more precisely to you and to each card. In practice that usually means similar retention for fewer total reviews, because time is not spent on cards you would still remember.

What are common mistakes with FSRS?

• Rating dishonestly. Pressing Good when you actually guessed teaches the model the wrong stability, and the card comes back too late. Rate what really happened.
• Cramming ahead of the due date. Reviewing early, before retrievability has dropped, gives the model little signal and wastes the review. Trust the queue.
• Chasing a very high retention target. Asking for ninety-eight percent recall buries you in reviews for a small gain; a target near ninety percent is the usual balance.
• Burying difficult cards in a giant card. If a card keeps failing, the fix is often to split it into smaller cards, not to fight the scheduler.

Do I need to configure FSRS in Recense?

No. FSRS is on by default and the defaults are sensible. You rate how each card went and the scheduler handles the timing, which means fewer wasted reviews, steadier long-term retention, and a daily queue that reflects what you are genuinely about to forget rather than an arbitrary list.

Bottom line: FSRS turns a guess about timing into a per-card prediction. It models how each memory fades and books the next review at the point where recall does the most good, so you keep more with less total study.