How reviews are scheduled

Every flashcard in Lingara carries three numbers and a date. The numbers are an ease factor, an interval in days, and a count of repetitions. The date is when the card is next due. Answering the card recomputes all four. That is the whole scheduler, and this page is what those four values do.

The deck you open is the due cards

A review session does not walk your whole collection. The app asks for the cards whose next-review date has already passed — literally, the ones where next_review_at <= now() — and those are what you see. Cards scheduled for next week are not in today’s deck, are not counted in today’s due count, and cannot be brought forward by studying harder.

This is the single most consequential thing about spaced repetition and the easiest to misread as a limitation. It is the feature. A deck of a thousand words you “should” review is a deck nobody opens; a deck of the eleven cards that are actually at risk of being forgotten today is one you finish before your coffee cools. The count grows on the days you have been away and shrinks on the days you have kept up, which is the correct signal in both directions.

The four answers

When a card is shown you grade it, and the grade is what the scheduler acts on. Internally the grades run on a 0–5 quality scale inherited from the SM-2 family, but only four outcomes are distinct:

Again. Any grade below 3. The interval is reset to 1 day, the repetition count is reset to zero, and the ease factor drops by 0.2. The card comes back tomorrow, and it comes back as though you had never seen it before — because as far as recall is concerned, you had not.

Hard. The interval is multiplied by 1.2 and rounded up; a card with no interval yet gets 1 day. The ease factor drops by 0.15. The card advances, but barely, and it gets harder to advance next time.

Good. The standard answer. The first repetition schedules the card for 1 day, the second for 3 days, and every one after that multiplies the current interval by the card’s own ease factor. The ease factor is left alone.

Easy. The interval jumps: 2 days on the first repetition, 4 days on the second, and afterwards the current interval multiplied by the ease factor and then by a further 1.3. The ease factor rises by 0.15.

A new card starts with an ease factor of 2.5. So a card you answer Good every time runs 1 day, 3 days, then 8, 20, 50, 125 — each step two and a half times the last, rounded up. Six correct answers spread over four months is the entire cost of a word you actually know.

What a miss costs

An Again is not just “tomorrow”. It resets the repetition count as well as the interval, so the card re-enters the 1-day, 3-day ladder from the bottom, and it takes the ease factor down with it. A card you have missed three times sits at an ease factor of 1.9 rather than 2.5, which means its intervals grow by 90 % rather than 150 % each time. Miss it often enough and the ease factor hits its floor of 1.3 and stays there: the scheduler refuses to go lower, because below that the card would effectively never leave the daily deck.

That floor is the scheduler saying something useful. A card stuck at 1.3 is not a card you need to see more often. It is a word you have not got a hook for — the sound, the character, the false friend in your own language — and no scheduling will fix that. Open it in the tutor and ask what it is doing there, or look at the example sentence it came from. The card will move once the word means something.

Why the intervals are the shape they are

Two design choices in the numbers above are worth naming, because they differ from the textbook algorithm.

The first is that grades 0, 1 and 2 are collapsed into a single Again. The classic algorithm distinguishes “complete blackout” from “wrong but familiar”, and then does almost nothing different with them. Collapsing them removes a decision you would have to make honestly, under time pressure, about a word you have just failed — and if you cannot make it honestly, the distinction is noise entering the schedule.

The second is that Hard has its own multiplier rather than being a small ease penalty on a normal interval. A card answered Hard genuinely did come back; it should not reset. But a 1.2× step is deliberately close to standing still, so a habit of pressing Hard to be safe produces a card that keeps appearing and never matures — which is the correct outcome for a word you keep half-remembering.

What this means for how you study

Most days beats long days. The scheduler hands you the cards that are due, and cards become due on a clock, not on a session count. Five minutes on six days surfaces six separate due sets. An hour on Sunday surfaces one, and every card you were supposed to see on Wednesday has been sitting expired since Wednesday.

Do not grade generously. The ease factor is the scheduler’s model of how well you know this particular word, and pressing Good on a word you dredged up after eight seconds tells it something false. The cost is not immediate — it arrives in three weeks, as a card you do not recognise at all and a schedule that had stopped watching it.

A backlog is not a debt. Coming back after a fortnight away produces a large due count, and none of it is punishment. Work through it in the order the app gives you and grade honestly; the Agains will reset the words you have lost and the Goods will push the rest further out than they were before. Two or three sessions later the count is back to normal, because that is what the interval growth is for.

Deleting your account deletes all of this. Your flashcards, your saved words and your progress are erased with the account — see how your data is handled. The scheduling state above is yours alone and is not pooled or shared with anyone.