SM-2 vs FSRS: Which Spaced Repetition Algorithm Is Better (and Does It Matter)?

Every spaced repetition app has to answer the same question for every card: when should you see it again? The algorithm that answers it is called a scheduler, and two names come up more than any others. SM-2 is the classic, developed by Piotr Woźniak for SuperMemo in the late 1980s and used, in one form or another, by a huge number of apps since. FSRS is the modern challenger, built on a model of memory fitted to real review data, and available in Anki since version 23.10.

So which one is better? The short answer is that FSRS is more accurate, SM-2 is simpler, and both work. The longer answer is below, including the part most comparisons skip: how much the choice actually matters for your results.

Full disclosure up front: Space Repeat uses SM-2. We'll explain why, but we'll also be straight about where FSRS is ahead.

How SM-2 Works

SM-2 gives every card an "ease factor", a number that starts at 2.5 and represents how easy the card is for you. When you review a card, you rate how well you remembered it. Space Repeat uses four ratings: Again, Difficult, Good and Easy.

The first two successful reviews use fixed gaps: about 1 day, then about 6 days. After that, each new interval is the previous one multiplied by the ease factor. A card with an ease of 2.5 that you last saw 10 days ago comes back in about 25 days if you get it right again. Rate a card as hard and its ease drops, so its intervals grow more slowly. Forget it and it starts over from a short interval.

That's the whole idea, and its simplicity is its strength: it needs no setup, no history and no tuning, and you can predict what it will do. For a fuller walkthrough, see How Spaced Repetition Algorithms Work.

How FSRS Works

FSRS (Free Spaced Repetition Scheduler) models memory with three variables, as Anki's documentation explains:

Retrievability: the probability that you can recall the card right now. It falls as time passes.

Stability: how many days it takes for retrievability to fall from 100% to 90%. Successful reviews increase it.

Difficulty: how hard it is to increase that card's stability with each review.

Instead of fixed rules, FSRS uses formulas whose parameters can be fitted to your own review history with machine learning. And instead of tweaking intervals directly, you choose a desired retention, for example 90%, and FSRS schedules each card for roughly the moment your chance of recalling it drops to that level.

What the Evidence Says

The FSRS project runs an open-source benchmark built from hundreds of millions of reviews by around 10,000 Anki users. It measures how accurately each algorithm predicts whether you'll remember a card when it comes up, and FSRS comes out well ahead of SM-2-style scheduling on that measure. Newer FSRS versions keep improving on older ones.

Better predictions should mean better-timed reviews: fewer reviews of cards you'd have remembered anyway, and fewer cards that slip past the point of forgetting. That's why FSRS's developers and many Anki users report needing fewer reviews for the same retention.

Two caveats are worth keeping in mind. First, the benchmark measures prediction accuracy, not how much people learn, so it shows FSRS is the better model of memory rather than directly measuring exam results. Second, FSRS works best once it has enough of your review history to fit its parameters to you. Until then it relies on default parameters.

Where SM-2 Still Holds Up

It needs nothing from you. SM-2 works the same on day one as on day five hundred, with no optimisation step, no retention target to pick and no parameters to understand.

It's predictable. You can work out by hand when a card will come back, which makes it easy to trust and easy to plan around. Our free study schedule planner uses it for exactly this reason: the review dates it gives you are easy to follow on a calendar.

It's proven. Decades of learners have used SM-2-based scheduling successfully. Its main weakness is efficiency, not effectiveness: it may show you some cards a little earlier or later than ideal, but it still spaces your reviews out in the way that builds long-term memory.

Why the Algorithm Matters Less Than You Think

The gap between SM-2 and FSRS is a gap in efficiency: how many reviews you need for a given level of retention. Both are vastly better than no spacing at all, and both are far more sensitive to how you use them than to which one you use.

Reviewing consistently matters more. Any scheduler assumes you show up roughly when cards are due. A week of missed reviews costs more than any algorithm can save.

Rating honestly matters more. If you press Good on a card you barely remembered, no algorithm can schedule it well, because it's working from wrong information.

Writing good cards matters more. A card that asks three things at once is hard to rate and hard to schedule. One fact per card helps every algorithm. See How Many Flashcards Should You Study Each Day for keeping the daily load sustainable.

In other words, switching algorithms is an optimisation for people who already have a solid habit. If you're still building the habit, the habit is what will make the difference.

Which Apps Use Which?

Anki offers both, with FSRS available since version 23.10 as an option you switch on in the deck settings. Some other apps, such as RemNote, also offer FSRS. Space Repeat uses SM-2, with four ratings (Again, Difficult, Good, Easy), because it suits an app built to work with zero setup. For a side-by-side look at the apps themselves, see The Best Spaced Repetition Apps or Space Repeat vs Anki.

The Bottom Line

FSRS is the more accurate algorithm, and if you're an experienced Anki user who wants the fewest possible reviews, turning it on is worth trying. SM-2 is simpler, predictable and needs no configuration, and it remains a perfectly good way to schedule reviews. For most learners, the deciding factors are reviewing every day, rating honestly and keeping cards simple, whichever algorithm is doing the scheduling.