FSRS vs SM-2: Why Modern Spaced Repetition Algorithms Matter

If you use flashcards seriously, the algorithm running behind the scenes matters more than you think. The scheduling algorithm decides when you see each card — and a smarter algorithm means less time studying for the same (or better) retention. For decades, SM-2 was the only game in town. Now FSRS v5 is changing the rules.

In this article, we'll break down both algorithms, explain where SM-2 falls short, show what FSRS does differently, and explain why DeckStudy uses a modern approach to give you a genuine edge.

A Quick History of Spaced Repetition Algorithms

Spaced repetition — reviewing material at increasing intervals — has been studied since Hermann Ebbinghaus mapped the forgetting curve in 1885. But it wasn't until software got involved that the technique became practical at scale.

SM-2: The Algorithm That Started It All

SM-2 was created by Piotr Wozniak in 1987 as part of his SuperMemo project. It was revolutionary for its time: a simple formula that adjusts review intervals based on a single "ease factor" per card and your self-reported recall grade (0–5).

Here's the essence of SM-2:

• Each card has an ease factor (starting at 2.5)
• After each review, the next interval = previous interval × ease factor
• If you fail a card, it resets to a short interval
• Your grade (Again / Hard / Good / Easy) adjusts the ease factor up or down

Anki adopted SM-2 in 2006, and for nearly two decades it was the default algorithm for serious flashcard users. It works. Millions of students have passed exams thanks to SM-2.

But SM-2 has real problems.

Where SM-2 Falls Short

1. The Ease Factor Death Spiral

This is SM-2's most notorious flaw. When you press "Again" on a card, the ease factor drops. A lower ease factor means shorter intervals. Shorter intervals mean you see the card more often — which means more chances to press "Again" — which drops the ease factor further.

The result: cards you find difficult get stuck in an ever-tightening loop. You review them constantly but never escape to longer intervals, even after you've actually learned them. The algorithm doesn't recover gracefully from early failures.

2. One-Size-Fits-All Parameters

SM-2 uses the same starting parameters for every learner. But people have wildly different memory characteristics. A medical student reviewing anatomy and a language learner studying vocabulary might need completely different scheduling, even for cards of similar difficulty. SM-2 ignores this.

3. No Forgetting Curve Modeling

SM-2 doesn't actually model memory. It uses a fixed multiplier formula that approximates spacing, but it has no concept of "what is the probability this person remembers this card right now?" It schedules mechanically rather than predictively.

4. Poor Handling of Lapses

When you forget a card in SM-2, it essentially starts over — short interval, reduced ease factor. There's no nuance between "I blanked on a card I knew yesterday" and "I've never successfully recalled this card." Both get treated roughly the same.

FSRS: The Next Generation

FSRS (Free Spaced Repetition Scheduler) was developed by Jarrett Ye and has gone through several iterations, with FSRS v5 being the current state-of-the-art. It takes a fundamentally different approach.

How FSRS Works

Instead of a simple multiplier, FSRS models your memory using three core concepts:

• Stability (S): How long the memory will last before the recall probability drops to a threshold (typically 90%). Higher stability = longer intervals.
• Difficulty (D): An intrinsic property of the card that reflects how hard it is for you specifically.
• Retrievability (R): The estimated probability that you can recall the card right now, based on time elapsed and stability.

FSRS uses a machine learning model trained on hundreds of millions of real reviews to predict these values. When it schedules your next review, it's answering a real question: "When will this person's recall probability for this card drop to 90%?"

Why FSRS Is Better

Personalized parameters. FSRS can optimize its model parameters to your individual review history. After a few hundred reviews, it learns your personal forgetting patterns and adjusts scheduling accordingly. Some people have stronger long-term memory; some learn new cards faster. FSRS adapts.

No ease factor death spiral. Because FSRS tracks stability and difficulty separately, a few early failures don't permanently cripple a card's scheduling. If you start consistently recalling a card, its stability rises appropriately.

Actual forgetting curve modeling. FSRS computes the probability that you'll remember a card at any given moment. This means it can schedule reviews at exactly the point where your recall probability hits your desired threshold — not earlier (wasting time) or later (forgetting).

Better lapse handling. When you forget a card, FSRS doesn't blindly reset it. It uses the card's history to estimate a new stability value that reflects partial learning, recovering faster than SM-2's reset-and-rebuild approach.

FSRS vs SM-2: Head-to-Head

What the Research Shows

Benchmarks conducted on large-scale datasets show that FSRS v5 consistently outperforms SM-2:

• Up to 30% reduction in the number of reviews needed to maintain the same retention rate
• More accurate prediction of whether a user will remember a card (lower log loss in prediction models)
• Better handling of difficult cards — fewer cards get "stuck" in low-interval loops

In practical terms: if you spend 30 minutes a day on reviews with SM-2, switching to FSRS could save you roughly 10 minutes daily — while maintaining or improving your retention. Over a year, that's more than 60 hours saved.

How DeckStudy Uses Modern Scheduling

DeckStudy is built with modern spaced repetition at its core. Rather than sticking with a 40-year-old algorithm, DeckStudy leverages advances in scheduling science to give you:

• Smarter intervals: Review cards at the right time — not too early (wasting effort) and not too late (forgetting)
• Adaptive learning: The system adjusts to your personal memory patterns over time
• No death spirals: Difficult cards get appropriate scheduling without getting trapped in frustrating loops
• Configurable goals: Whether you're cramming for an exam next week or building knowledge for life, the scheduling adapts

Combined with AI-powered card generation, DeckStudy means you spend less time both creating and reviewing cards — leaving more time for actual learning.

Should You Care About Your Algorithm?

If you study with flashcards for more than 15 minutes a day, absolutely yes. The algorithm is the engine of your study system. A better algorithm means:

• Less time reviewing cards you already know well
• More attention on cards you're about to forget
• Fewer total reviews for the same retention
• Less frustration from cards stuck in short-interval loops

You wouldn't use a navigation app that ignores traffic data. Don't use a flashcard app that ignores 35 years of scheduling research.

Frequently Asked Questions

Can I switch from SM-2 to FSRS without losing progress?

Yes. FSRS can analyze your existing review history to estimate stability and difficulty for each card. You don't start from scratch — your past reviews inform the new model.

Is FSRS harder to use than SM-2?

Not at all. The complexity is under the hood. As a user, you still just rate each card (Again / Hard / Good / Easy) and the algorithm does the rest. With DeckStudy, you don't need to configure anything — it works out of the box.

Does FSRS work for all types of content?

Yes. FSRS has been tested on language learning, medical flashcards, programming concepts, and general knowledge. The machine learning model generalizes well across domains.

What if I only study casually?

Even casual learners benefit from better scheduling. If you only study a few minutes per day, it's even more important that those minutes are spent on the right cards.

Study Smarter, Not Harder

The difference between SM-2 and FSRS isn't academic — it's practical. Modern algorithms mean less wasted review time, better retention, and a more enjoyable study experience. Try DeckStudy free and experience what spaced repetition feels like when the algorithm actually understands how you learn.