One Lapse Resets the Interval, and the Ease Factor Never Fully Recovers
Published 10/23/2026 · 3 min read · Everyday calculators
Lena Hoffmann — Science & education writer at OneKitly
Mathematics · Physics
Checked against 2 sources
The scheduler keeps two numbers per card: how many times it has been recalled in a row, and an ease factor that starts at 2.50. A successful review multiplies the interval by the ease — a card at six days with a rating of four goes to fifteen — and the sequence of first intervals is fixed at one day, then six, then multiplied thereafter. A failure does two things at once. It sends the interval back to a single day and the repetition count back to zero, and it also subtracts from the ease factor: a card rated two after a thirty-day interval comes back tomorrow with an ease of 2.18 instead of 2.50. The interval recovers; the ease factor does not, except very slowly.
A card recalled well goes from a 6-day interval to 15. A card failed after 30 days drops to 1 day and its ease factor falls from 2.50 to 2.18 — permanently.
Getting it right is not one thing
The three passing grades all produce the same next interval and three different futures, because they move the ease factor apart: a card answered perfectly gains a tenth, one answered correctly holds still, and one answered correctly but with effort loses fourteen hundredths. After a dozen reviews those small differences compound into intervals that differ by months, which is the entire mechanism by which the algorithm learns which cards are hard for you. Rating everything as merely correct therefore throws away the only signal it has, and the deck degenerates into a fixed schedule that ignores the material.
The multiplication is the whole point
Intervals that multiply rather than add are what make the method scale: one day, six days, fifteen, thirty-eight, ninety-five, and a card reviewed eight or nine times in total is remembered for years. The cost of a deck therefore grows far more slowly than its size, since old cards ask for almost nothing while new ones take most of the daily time. The corollary is that the method rewards patience and punishes cramming — a card added the night before an exam has had one review and is at the bottom of the ladder, and nothing about the schedule can be accelerated by wanting it more.
| Rating | Next interval | New ease |
|---|---|---|
| 5 — perfect | 15 days | 2.60 |
| 4 — correct | 15 days | 2.50 |
| 3 — correct with effort | 15 days | 2.36 |
| 2 — failed | 1 day | 2.36 |
Frequently asked questions
- Why does the failed card keep the reduced ease?
- Because the failure is evidence about the card, and the algorithm keeps evidence. A card you have forgotten once is more likely to be forgotten again, so the scheduler permanently slows the rate at which its interval grows — the card climbs the same ladder afterwards, but with shorter rungs. Ease can recover, since perfect recalls add to it, but only by a tenth at a time against a fifth or more lost per lapse, so a card that has lapsed several times settles at a low ease and stays in frequent rotation. That is usually a signal to rewrite the card rather than to keep drilling it.
- Is this algorithm still the state of the art?
- No, and it does not need to be. The scheme here dates from the late 1980s and has been superseded by models that fit a forgetting curve to a user's own review history rather than adjusting a single ease number, and those measurably schedule better. What the old algorithm has is transparency: every step can be done by hand, the state per card is two numbers, and anyone can see exactly why a card came back today. For understanding how spaced repetition works — as opposed to squeezing the last few per cent out of a schedule — that is worth more than the accuracy.
Articles you may find interesting
All guides →Related tools
Sources
Spotted a mistake in this article?