What is FSRS? The spaced repetition scheduler, explained for NEET students

FSRS, the Free Spaced Repetition Scheduler, is an open-source algorithm that decides when you should next review each flashcard or concept. It models your memory of every item with three numbers, difficulty, stability and retrievability, and schedules a review for the moment your predicted chance of remembering it falls to a target you choose, most commonly 90%. Anki added it as an alternative to its older SM-2 algorithm in version 23.10, and Lumi uses it for NEET revision scheduling.

This article explains what those three numbers mean, how FSRS differs from a fixed revision timetable and from SM-2, and where its limits are.

The three numbers FSRS keeps for every card

FSRS describes each card's memory with difficulty, stability and retrievability, often shortened to DSR. The FSRS project defines them precisely (ABC of FSRS, https://github.com/open-spaced-repetition/awesome-fsrs/wiki/ABC-of-FSRS):

• Retrievability (R): the probability that you can recall the item right now. It falls with time since your last review.
• Stability (S): the time, in days, for retrievability to fall from 100% to 90%. A card with a stability of 30 days will still be recalled with about 90% probability a month after its last review.
• Difficulty (D): how hard it is to increase this item's stability. Difficult cards gain less stability from each successful review.

How FSRS decides when you see a card again

FSRS schedules each card for the day its retrievability is predicted to hit your desired retention. With a desired retention of 0.9, a card is due when your chance of recalling it has fallen to about 90%. Each successful review raises the card's stability, so the next gap is longer. A failed review cuts stability and brings the card back soon.

The desired retention setting is a trade-off between memory and workload. Anki's manual puts it plainly: the default is 90%, higher retention means shorter intervals and more reviews per day, above 90% the workload increases very quickly, and above 97% it can be overwhelming (Anki Manual, https://docs.ankiweb.net/deck-options.html). Lumi uses 0.9.

FSRS vs a fixed revision timetable

A fixed timetable, such as 'revise on day 1, 3, 7, 21 and 60', gives every item the same gaps. That is better than no plan, and far better than cramming, but it ignores the most important fact about your memory: you know different things to very different degrees.

With a fixed timetable, a concept you mastered on the first go still comes back on day 3, wasting a review, while a concept you keep confusing waits until day 21 and is gone by then. FSRS separates them. The mastered concept gains stability fast and drifts out to months. The confusing one is caught every few days until it holds. Over hundreds of NEET concepts, that difference adds up to many hours.

FSRS vs SM-2, Anki's old default

SM-2 is the algorithm Piotr Woźniak published for SuperMemo in the late 1980s. Its first two intervals are fixed at 1 and 6 days, and every later interval is the previous one multiplied by an ease factor that starts at 2.5 and moves up or down with your answers (SM-2 description, https://super-memory.com/english/ol/sm2.htm). Anki's version keeps the same idea: the starting ease defaults to 2.50, so answering Good multiplies the interval by about 2.5.

SM-2's rules were hand-set. FSRS, by contrast, is fitted to data. Its approach was described in a peer-reviewed paper at KDD 2022, which modelled memory from a large dataset of real reviews and optimised scheduling against it (Ye, Su & Cao, 2022, https://doi.org/10.1145/3534678.3539081). Its parameters can be optimised on your own review history, so the model adapts to how you in particular forget.

How much better is it in practice? The FSRS project states that users need 20 to 30% fewer reviews than with SM-2 to reach the same retention. That is the project's own figure, not an independent trial, so treat it as a claim with evidence behind it rather than a settled number. The project also publishes an open benchmark that compares prediction accuracy across algorithms on about 10,000 Anki users' collections, roughly 727 million reviews (SRS Benchmark, https://github.com/open-spaced-repetition/srs-benchmark), and anyone can rerun it.

What FSRS cannot do

FSRS schedules memory. It does not know whether you understood the mechanism behind an answer or just memorised the answer. For NEET, that means it is excellent for the recall load of Biology and much of Chemistry, and a supporting tool rather than the whole answer for Physics, where you also need practice solving new problems.

It also depends on honest grading. Anki's manual notes that FSRS can adapt to almost any habit except one: pressing 'Hard' when you actually forgot. The algorithm then treats a failure as a success and pushes the card out too far. If you use an FSRS-based tool, mark forgotten items as forgotten.

How Lumi uses FSRS

Lumi runs the reference FSRS implementation rather than a home-grown interval rule, at a desired retention of 0.9, over roughly 2,700 NEET concept atoms. Each atom carries its own difficulty, stability and retrievability. That is one of three models Lumi names publicly: FSRS decides when you revise, a two-parameter Item Response Theory model estimates your ability, and a knowledge-tracing model predicts which concepts you would get right now. The how-it-works page on lumineet.com describes all three.

The honest summary is that FSRS is a well-documented, openly benchmarked way to decide when to review. It is not magic. It makes your revision time go further, provided the revision itself is active recall.

A worked example of FSRS scheduling

Here is how FSRS behaves in practice, in outline. You learn a concept today and answer it correctly the next day, so it gets a modest stability and comes back after a few days. You get it right again, stability jumps, and the next gap stretches to a couple of weeks, then to a couple of months. Each success raises stability by an amount that depends on the card's difficulty and on how close to forgetting you were when you recalled it.

Now suppose you get it wrong at the two-week review. FSRS lowers the stability sharply and increases the difficulty, so the concept returns within days and grows more slowly from then on. A concept you find easy and a concept you keep confusing can start on the same day and end up months apart in their schedules. That is the whole point: effort goes where forgetting is happening.

Related on Lumi

• How many times should you revise a NEET chapter? The numbers
• Spaced repetition for NEET: remembering 2,700 concepts without cramming
• How AI study tools actually work, and how to spot the fake ones
• Why you forget NEET Biology within a week, and how to stop