Spaced repetition calculator

What spaced repetition is

Spaced repetition means reviewing something at growing gaps instead of cramming it in one sitting. A 2021 meta-analysis by Latimier, Peyre and Ramus (Educational Psychology Review) concluded that strong recommendations to teachers and students in favour of spaced retrieval practice are warranted. Research comparing expanding gaps with equal gaps has found them about as good as each other (Karpicke and Roediger, 2007), so the exact schedule matters less than reviewing at all and doing it consistently. This page turns a schedule into dates: pick a method, enter the day you learned the card and what happened at each review, and read off when to see it next.

SM-2: the SuperMemo algorithm

Piotr Wozniak described SM-2 in a text dated 1990 that is adapted from his master's thesis, and used it in SuperMemo versions 1.0 to 3.0 (December 1987 to March 1989). Every item starts with an easiness factor EF = 2.5. After you learn a card it is reviewed after I(1) = 1 day, then I(2) = 6 days, and from then on each gap is the previous gap × EF, with fractions rounded up. After each review you grade your answer q from 0 to 5, and the factor changes: EF' = EF + (0.1 − (5 − q) × (0.08 + (5 − q) × 0.02)), never below 1.3. If q is below 3, the card starts over with the 1-day and 6-day steps and EF stays as it was. Wozniak also advised repeating, in the same session, every card graded below 4 until all reach at least 4; the table marks those reviews.

The grade scale, with the change each grade makes to a starting EF of 2.5: 5 is perfect response (+0.10); 4 is correct response after a hesitation (+0.00); 3 is correct response recalled with serious difficulty (−0.14); 2 is incorrect response; the correct one seemed easy to recall (−0.32); 1 is incorrect response; the correct one remembered (−0.54); 0 is complete blackout (−0.80). A steady stream of 4s leaves EF at 2.5, so the gaps run 6, 15, 38, 95 days and so on; every 5 raises EF by 0.1 and every 3 lowers it by 0.14.

Leitner boxes

The Leitner system, from Sebastian Leitner's book So lernt man lernen, keeps cards in numbered boxes. A new card goes in box 1. When you recall it, it moves up one box; when you miss it, it goes back to box 1. Higher boxes are reviewed less often. How often is the part sources disagree about: Leitner's original box was divided by partitions 1, 2, 5, 8 and 14 cm wide, and modern guides use anything from three to five boxes with different day counts. For that reason the days between reviews are a field you control here (the example uses 1, 3 and 5 days, matching the three-box example on Wikipedia). Box rules are simple enough to run on paper, which is their main advantage over SM-2.

The forgetting curve and its caveat

Hermann Ebbinghaus measured forgetting on himself in the 1880s with lists of nonsense syllables and a "savings" score: how much less effort it took to relearn a list after a delay. A common textbook model of the curve is R = e^(−t/S), where R is the fraction remembered after t days and S is a memory strength in days. With S = 5 days the model gives 81.9% after one day, 24.7% after a week, and R falls to 90% after t = −S × ln(0.9) = 0.53 days, about 13 hours. S is not something this page can measure for you; it is a number you try.

The caveat is that real data are messier than one exponential. In 2015 Jaap Murre and Joeri Dros replicated Ebbinghaus's experiment (PLoS ONE) and found that the curve is not completely smooth: it most probably jumps upward starting at the 24-hour point. Their one subject's savings scores are shown below next to Ebbinghaus's. Note that savings is a relearning measure, not the probability of recalling the item, and that the two curves are far apart at 31 days, where the authors suggest the later subject's lists were learned early in the study, under different conditions.

Worked examples

SM-2, with the card learned on Sunday, November 1, 2026 and grades 5, 4, 4, 3, 5: the gaps after the five reviews are 6, 16, 42, 110, 271 days, the ease factor ends at 2.56, and the fifth review lands on 2027-04-25. The 3 lowers EF by 0.14 and is also flagged for a same-day repeat, because it is below 4. With grades 5, 5, 2, 4 the 2 sends the card back to a 1-day gap, EF stays at 2.70, and the next successful review gets the 6-day gap again: 6, 16, 1, 6 days.

Leitner with boxes of 1, 3 and 5 days and answers yes, yes, no, yes: the card is reviewed on 2026-11-02, 2026-11-05, 2026-11-10, 2026-11-11; at those four reviews it sits in boxes 1, 2, 3, 1, and the miss at the third review drops it back to box 1.

Limits of these schedules

The dates assume each review happens on the due day and that you grade yourself honestly; if you review late, the real gap is longer and the next one should start from the actual date. SM-2 dates from the late 1980s, and the apps that grew out of it, including Anki, use modified versions, so their dates will not match this page exactly. SM-2 itself is a rule of thumb that worked for its author; Wozniak reported 92% retention after excluding recently learned items in his first year (10,255 English vocabulary items), which describes one learner with one method, not a guarantee. And a schedule decides when to look at a card, not whether the card is good: small, clear, single-fact cards are what the method assumes.

What does the SM-2 grade 0 to 5 mean?

5 is a perfect response, 4 a correct response after a hesitation, 3 a correct response recalled with serious difficulty, 2 an incorrect response where the correct one seemed easy to recall, 1 an incorrect response where the correct one was remembered when you saw it, and 0 a complete blackout. Grades 3 to 5 count as success; below 3 the card starts over.

What is the ease factor?

A number attached to each card that says how fast its gaps should grow. It starts at 2.5, which means each gap is 2.5 times the last one (after the first two fixed gaps of 1 and 6 days). Easy answers raise it, hard ones lower it, and it never drops below 1.3.

How many days should the Leitner boxes be?

There is no single correct answer. Sources differ on the number of boxes and on the days between reviews; Leitner's original box used partitions of different sizes rather than days. Start with something simple such as 1, 3 and 5 days and adjust: if you miss most cards in a box, shorten its gap; if everything is easy, lengthen it.

Is the forgetting curve really R = e^(−t/S)?

It is a convenient model, not a law. Ebbinghaus's own data and the 2015 replication by Murre and Dros are better described by other shapes, such as a sum of exponentials or a power function, and the replication found a bump after about a day. The formula is useful for comparing "what if memory strength S were 5 days or 20 days", not for predicting any particular learner.

What if I miss a review day?

Review as soon as you can and enter the grade as usual. In this tool each review is assumed to happen on its due date, so for a late review count the new gap from the day you actually reviewed it.

Does spaced repetition really work?

The evidence for spacing out practice and for testing yourself is strong: a 2021 meta-analysis concluded that strong recommendations in favour of spaced retrieval practice are warranted. What is less settled is the best exact schedule, which is why the intervals here are yours to change.

Sources, checked October 27, 2026: P. A. Wozniak, "Application of a computer to improve the results obtained in working with the SuperMemo method", supermemo.com (text dated 1990); S. Leitner, So lernt man lernen, Herder; J. M. J. Murre and J. Dros, "Replication and analysis of Ebbinghaus' forgetting curve", PLoS ONE 10(7): e0120644 (2015, CC BY; savings table from the article); A. Latimier, H. Peyre and F. Ramus, Educational Psychology Review 33(3), 2021, 959–987; C. Karpicke and H. L. Roediger, Journal of Experimental Psychology: Learning, Memory, and Cognition 33(4), 2007, 704–719. Related: the study hours calculator, reading plan calculator, date calculator, countdown calculator and final grade calculator. How we check our calculators.