FSRS vs. SM-2: Why the Newer Spaced Repetition Algorithm is the Future of Fluency
The math of memory has changed. Here is how FSRS outperforms traditional flashcards and hacks the forgetting curve.
For over 20 years, the SM-2 algorithm defined how we study with flashcards. But a new standard has emerged. The Free Spaced Repetition Scheduler (FSRS) uses machine learning to predict memory decay with unprecedented accuracy. This post explains the difference between SM-2 and FSRS, and how to use the latter to optimize vocabulary retention.
We have all been there: You study a vocabulary list on Monday. You review it Tuesday. By Friday, you feel confident. But when you look at that same list two weeks later, the words look foreign again.
This is the "Leaky Bucket" problem of human memory. We pour information in, but it constantly drips out.
To plug the holes, serious learners rely on a Spaced Repetition System (SRS). However, a massive shift has recently occurred in the cognitive science community. The algorithm that has powered apps like Anki and SuperMemo for decades is being dethroned by a smarter, AI-driven successor: FSRS.
Here is how the math of memory is changing, and why upgrading your algorithm is the fastest way to fluency.
The Baseline: Hacking the Forgetting Curve
To understand why FSRS is superior, we must first define the problem it solves. In 1885, psychologist Hermann Ebbinghaus identified the Forgetting Curve, proving that memory decay is exponential. Without review, you lose the bulk of new information within 24 hours.
Spaced Repetition is the technique of interrupting this curve. The goal is to review a word just as you are about to forget it.
- Too soon? You waste time reviewing something you already know.
- Too late? You have to relearn the word from scratch.
- Just right? You strengthen the neural pathway and flatten the curve.
The difficult part is calculating that "Just Right" moment.
The Old Standard: The SM-2 Algorithm
For decades, the industry standard was the SM-2 algorithm. Created by Piotr Woźniak in the late 1980s, SM-2 is the default engine behind most flashcard apps.
SM-2 relies on static rules. If you mark a card as "Good," it simply multiplies the previous interval by a fixed factor (usually 2.5).
- Review 1: 1 day
- Review 2: 3 days
- Review 3: 8 days
The Limitation: SM-2 is a "one-size-fits-all" approach. It assumes that "Apple" (a simple cognate) and "Anachronistic" (a complex abstract concept) decay at the same rate. This rigidity often leads to "Ease Hell"—a state where the algorithm punishes you too harshly for a minor lapse, forcing you to review mature cards excessively.
The New Standard: What is FSRS?
The Free Spaced Repetition Scheduler (FSRS), developed by researcher Jarrett Ye, is a modern algorithm that replaces static rules with machine learning.
Unlike SM-2, FSRS does not treat all memories equally. It calculates two distinct variables for every single card in your deck:
- Retrievability (R): The probability (percentage) that you can recall a specific card at this exact moment.
- Stability (S): The time required for that retrievability to drop from 100% to 90%.
Why FSRS is Better for Language Learners
Language acquisition is nuanced. FSRS adapts to your personal learning history in real-time.
- It Learns Your Patterns: If you consistently remember verbs but struggle with nouns, FSRS detects this pattern and adjusts the intervals for each grammatical category separately.
- Smart Resiliency: In SM-2, forgetting a card usually resets its interval to zero. FSRS understands that a "lapse" doesn't mean total amnesia. It adjusts the interval intelligently, preserving your long-term progress.
According to benchmark tests on over 20,000 users, FSRS can reduce study loads by 30-40% while achieving higher retention rates than SM-2.
The Missing Piece: Algorithm vs. Content
Switching to an FSRS-based schedule is a high-ROI change. It ensures you are reviewing at the exact moment of maximum efficiency.
However, the algorithm only solves "When" to study. It does not solve "What" to study.
You can have the world's best scheduling algorithm, but if you are feeding it single words like "Run" or "Blue," you are still suffering from Shallow Processing. As we discussed in our post on Context Learning vs. Rote Memorization, the brain needs sentences, not isolated data points.
How to Automate FSRS and Context Mining
The ultimate study workflow combines Deep Processing (Context) with Optimized Scheduling (FSRS).
Setting up FSRS manually in Anki can be technical—it involves optimization code and parameter tweaking. This is why we built FSRS directly into ContextCards.
When you use ContextCards:
- The Content: We instantly generate sentences to ensure semantic encoding.
- The Schedule: We run a custom implementation of FSRS in the background.
You don't need to tweak retention rates or calculate "Stability" values. You simply study the word in context and let our engine decide exactly when that card needs to reappear to lock it into long-term memory. Optimize your retention with FSRS + Context Learning for free.