FSRS: The Science Behind Optimal Memory Retention

What is FSRS?

FSRS (Free Spaced Repetition Scheduler) is a modern spaced repetition algorithm developed by Jarrett Ye in 2022. It uses a mathematical model of human memory to predict exactly when you're about to forget something, scheduling reviews at the optimal moment.

Unlike older algorithms like SM-2 (used by Anki since 2006), FSRS is based on the latest memory research and has been validated on data from over 20,000 real users.

FSRS by the Numbers

30%

Fewer reviews needed vs SM-2

90%

Target retention rate

20K+

Users in validation study

How FSRS Works

FSRS models memory using two key concepts: Stability and Difficulty.

S
Stability

Stability represents how well a memory is encoded in your brain, measured in days. Higher stability means the memory will last longer before you forget it.

For example, if a card has stability of 30 days, you have a 90% chance of remembering it after 30 days.

D
Difficulty

Difficulty represents how hard a specific card is for you personally, on a scale of 1-10. This affects how quickly stability grows with each review.

Easier cards (low difficulty) gain stability faster. Harder cards need more reviews to reach the same stability.

The Forgetting Curve

FSRS uses a power-law forgetting curve to predict your probability of recall:

R = (1 + t/S × 0.2346)^-0.5

Where R is retrievability (probability of recall), t is time since last review, and S is stability.

FSRS vs SM-2 (Anki's Default)

SM-2 was developed in 1987 and has remained largely unchanged since. While revolutionary for its time, it has significant limitations compared to FSRS.

Feature	SM-2	FSRS
Memory model	Exponential (simplified)	Power-law (accurate)
Personalization	Fixed parameters	Optimizes to your data
Retention control	Indirect (interval modifier)	Direct target (e.g., 90%)
Review efficiency	Baseline	30% fewer reviews
Research basis	1987 theory	2022 machine learning

FSRS Optimizer: Personalized Parameters

One of FSRS's most powerful features is parameter optimization. The algorithm can analyze your review history and tune its 19 parameters to match your personal memory patterns.

What the Optimizer Does

Analyzes your historical review data and success/failure patterns
Calculates your personal forgetting curve shape
Adjusts stability growth rates for different difficulty levels
Optimizes initial stability estimates for new cards

For best results, run the optimizer after you have 1,000+ reviews. More data = better personalization.

Why 90% Retention?

FSRS defaults to 90% desired retention, meaning cards are scheduled so you have a 90% chance of remembering them when they come due. This is the research-backed sweet spot:

Higher retention (95%+): Too many reviews, diminishing returns
90% retention: Optimal balance of retention and efficiency
Lower retention (80%): Faster learning but more forgotten material

You can adjust this in Memlex based on your needs. Preparing for an exam next week? Bump it to 95%. Learning casually? 85% might be fine.

Getting Started with FSRS

Memlex uses FSRS by default - no configuration needed. Just create cards and start reviewing. The algorithm handles everything automatically.

1. Rate honestly: Use Again/Hard/Good/Easy based on how easily you recalled the answer. Don't rate Good if you struggled.

2. Review daily: FSRS works best with consistent daily reviews, even if just 10-15 minutes.

3. Trust the intervals: If a card is scheduled for 30 days, trust the algorithm. Don't manually review it early.

4. Run the optimizer: After 1,000+ reviews, run the FSRS optimizer in Settings to personalize your parameters.