If you have looked under the hood of a flashcard app recently, you have likely seen two algorithm names: SM-2 and FSRS. Several competitors have started marketing FSRS as a scientifically superior replacement for SM-2, which sounds persuasive — until you understand what the difference actually means for day-to-day vocabulary study.
This post breaks down both algorithms, explains when the gap between them matters, and clarifies why the flashcard app you use consistently will always outperform the most mathematically elegant algorithm you open twice a week.
What Is SM-2?
SM-2 (SuperMemo 2) was developed by Piotr Wozniak in 1987 as part of the SuperMemo software project. It was the first practical spaced repetition algorithm built for everyday learners and became the foundation Anki adopted at launch.
The logic is straightforward. Every card carries an ease factor that starts at 2.5. Each time you review a card and rate your recall — easy, good, hard, or again — the algorithm adjusts the ease factor and calculates the next interval.
Rate a card "easy" several times and it stretches to weeks, then months. Rate it "hard" and it returns in a day or two until you build reliable recall. The interval formula works roughly like this:
- First review: 1 day
- Second review: 6 days
- Subsequent reviews: previous interval × ease factor
If your ease factor drops to 1.3 from consistently rating cards hard, intervals grow slowly and the card stays close. If it reaches 3.0, a well-mastered card might not resurface for three months. The algorithm acts as a personal schedule, moving difficult vocabulary into frequent rotation and well-known words toward long-term storage.
What Is FSRS?
FSRS (Free Spaced Repetition Scheduler) is a newer algorithm released in 2022, developed primarily by Jarrett Ye and the open-source community. It approaches interval scheduling from a different direction.
Rather than multiplying a fixed ease factor, FSRS models memory decay as a mathematical function of time. It tracks two variables for each card: stability (how long a memory is likely to persist before dropping below a recall threshold) and difficulty (how inherently hard this specific card is for you). After each review, both variables update and the algorithm predicts the exact interval that keeps your recall probability at a target rate — usually 90%.
The difference sounds significant. In certain conditions, it is. But the claim that FSRS is universally better needs context.
Where the FSRS Advantage Is Real
FSRS produces its most meaningful gains under two specific conditions:
Large, mature decks. When a deck has over 1,000 cards and you have been studying it for six months or more, FSRS has enough historical data to build an accurate personal memory model. At this scale, its interval predictions reduce total review load while maintaining the same recall rate — you see each card fewer times to achieve the same retention.
Established Anki users with existing review history. FSRS calibrates best when it can analyze months of existing review logs. If you are migrating from Anki and have years of review data, enabling FSRS (now available as a built-in option in Anki 23.10 and later) can meaningfully improve scheduling efficiency.
Where SM-2 Holds Its Own
For most vocabulary learners, the difference between SM-2 and FSRS is negligible in practice.
First, FSRS requires a training phase. Its predictions are most accurate after hundreds of reviews per deck. When you start a new vocabulary set — a GRE word list, a medical terminology deck, a JLPT N2 set — FSRS has no history to work from and its early intervals are no more accurate than SM-2's.
Second, peer-reviewed comparison studies consistently show that retention rates between SM-2 and FSRS are nearly identical for decks under 500 cards or for learners with fewer than six months of history on a given set. The gap that does exist — around 8 to 12 percent fewer reviews needed with FSRS over a full year — is real but minor compared to the effect of review consistency.
Third, SM-2 has been stress-tested by millions of language learners, medical students, and exam candidates over 35 years. It works, and it works immediately from the first review session.
What Vocabulary Learning Actually Requires
The marketing around FSRS often frames algorithm choice as a performance lever. It is not, for most learners. The real levers are simpler.
Using a spaced repetition app built on either algorithm, your first week with a new vocabulary deck looks like this:
- Day 1: Learn 20 words. Most cards return tomorrow.
- Day 2: Review yesterday's cards. Words you recalled easily push out to day 8. Difficult ones stay close.
- Day 8: Easy words return. Hard ones from day 2 have been consolidating through daily review.
The mechanism doing the work here is review consistency, not algorithm precision. A learner who opens their SM-2 deck every morning for 90 days will outperform a learner who uses FSRS but skips every third session. The algorithm optimizes the intervals; you still have to show up.
How Flashi's SM-2 Implementation Works
Flashi uses SM-2 as its spaced repetition engine. The algorithm runs automatically — you never configure intervals or ease factors. You create a deck, review cards, rate each one after seeing the answer, and Flashi schedules the next session.
What Flashi adds on top of SM-2 is AI-assisted deck creation. Instead of manually typing 200 vocabulary cards the night before your exam, you paste a word list or topic description into the AI flashcard generator and receive a complete, reviewable deck in seconds. The generated cards include definitions and context; SM-2 immediately begins scheduling them.
This pairing removes the two largest friction points in flashcard study: building the deck and tracking review timing. You get battle-tested scheduling from day one with no setup overhead.
The Practical Takeaway
FSRS is the more mathematically sophisticated algorithm. SM-2 is the proven standard that has helped more vocabulary learners achieve retention than any other method in the history of flashcard software.
For most purposes — exam prep, language learning, professional terminology, standardized test vocabulary — SM-2 is the right algorithm to build on. If you eventually have a 2,000-card deck you have maintained for years, it may be worth exploring FSRS. Until then, the energy spent comparing algorithms is better spent adding ten new vocabulary cards and reviewing yesterday's set.
The best spaced repetition algorithm is the one attached to a deck you open every day. Start there.
Download Flashi free on iPhone and build your first vocabulary deck in under two minutes. No subscription, no account required.