Jarrett Ye

Definition:

Jarrett Ye is a researcher and software developer, best known as the primary author of the FSRS algorithm (Free Spaced Repetition Scheduler) — the most significant advance in SRS scheduling since SM-2, and now the default scheduler in Anki and used natively in Sakubo.

In-Depth Explanation

Ye developed FSRS through the open-spaced-repetition community project on GitHub, beginning around 2022. His core motivation was to address the well-documented limitations of SM-2 — a 1987 algorithm that, despite its reliability and widespread adoption, used fixed scheduling formulas that could not adapt to individual learners’ memory profiles or leverage the large-scale data that modern SRS platforms could provide.

FSRS is grounded in Piotr Wozniak‘s two-component model of long-term memory, which proposes that each memory trace has two independent properties: retrievability (the current probability of successful recall) and stability (how resistant the memory is to forgetting — roughly, how quickly retrievability decays). Wozniak proposed this model theoretically; Ye’s contribution was to operationalize it with modern machine learning, fitting the model to large datasets of actual user review data to derive parameters that outperform SM-2’s hand-tuned constants.

The result is a personalized, data-driven scheduler that predicts forgetting more accurately than SM-2 and adapts over time as more review data accumulates. Benchmarks comparing FSRS and SM-2 on real-world Anki data have consistently shown FSRS achieves higher retention at the same or lower review burden — or equivalent retention with fewer reviews.

Ye released FSRS as completely open-source (github.com/open-spaced-repetition), invited community testing and critique, and worked with Damien Elmes to integrate it into Anki — making what began as a research project the new default scheduler for millions of users. FSRS has since been adopted by other platforms.

History

• 2022: Ye begins developing FSRS through the open-spaced-repetition GitHub project, motivated by the limitations of SM-2 for modern large-scale SRS applications. Initial versions are published, benchmarked against SM-2, and iterated on the basis of community feedback and data.

• 2022–2023: FSRS goes through multiple versions (FSRS-4, FSRS-4.5, FSRS-5), each incorporating more user data, refined model parameters, and improved scheduling accuracy. Ye publishes benchmark results comparing FSRS and SM-2 on real Anki review data, demonstrating statistically significant accuracy improvements.

• 2022–present: Damien Elmes integrates FSRS into Anki as an optional, then preferred scheduler. This brings FSRS to Anki’s massive user base. The adoption of FSRS by Anki is the largest-scale deployment of a new SRS algorithm since SM-2 was encoded in Anki in 2006.

• 2023–present: FSRS is adopted by Mochi, and other modern SRS platforms. The algorithm is actively maintained, with Ye and the open-spaced-repetition community continuing to release updates as more data becomes available.

• Present: Ye continues to develop and maintain FSRS, publish benchmarks, and engage with the SRS research community. The open-spaced-repetition project has extended FSRS to multiple programming language implementations (Python, Rust, JavaScript) for easy integration into any platform.

Common Misconceptions

“FSRS is just another SRS algorithm like SM-2.” FSRS represents a fundamental redesign rather than an incremental improvement to the SM-2 algorithm. Where SM-2 uses a fixed exponential progression model, FSRS uses a parameterized neural-inspired model with per-card parameters optimized from the learner’s actual review history — producing substantially better stability predictions for individual cards, particularly those with unusual difficulty or response patterns. The algorithm’s open-source specification and documented accuracy superiority over SM-2 on large review datasets distinguish it from earlier refinements.

“FSRS requires a lot of technical setup.” FSRS is natively integrated into Anki (since version 23.10) as a fully supported scheduling option, accessible through the deck settings without any additional scripts or plugins. Users switching from the default SM-2 scheduler simply select FSRS in preferences — parameter optimization runs automatically on review history. The algorithm’s technical complexity is entirely abstracted from the user experience.

Criticisms

Some SRS researchers and practitioners have noted that FSRS’s accuracy improvement over SM-2, while statistically significant in aggregate, may produce modest practical differences for typical learners whose card difficulty distribution is roughly normal. The algorithm’s reliance on sufficient review history for parameter optimization means new users or those with small card collections see less individualization benefit in early months of use. The academic validation of FSRS has primarily relied on large-scale Anki review datasets; independent replication in controlled SLA studies is still limited.

Social Media Sentiment

Jarrett Ye and the FSRS algorithm are widely discussed in the Anki and SRS communities — the algorithm’s integration into Anki as the default scheduler option (replacing the decades-old SM-2) is one of the most significant community events in the Anki ecosystem in recent years. Community members share optimizer settings, compare retention rate improvements, and discuss FSRS parameter customization. The open-source development process and Ye’s active community engagement have earned the project significant community respect. The algorithm is now the recommended scheduler in virtually every community “how to set up Anki” guide.

Last updated: 2026-04

Practical Application

Enable FSRS in Anki (Tools ? Preferences ? Scheduling) and run the optimizer on your existing review history — this takes minutes and personalizes the scheduling parameters to your retention patterns. Set your desired retention rate (0.90 is standard; higher = more reviews; lower = fewer reviews but more forgetting). FSRS is the default scheduler in which applies the algorithm natively to vocabulary review — building on Ye’s research to optimize the spacing of word reviews for maximum retention efficiency with minimum review time.

Related Terms

• FSRS (Free Spaced Repetition Scheduler)
• SM-2 (SuperMemo 2)
• SRS (Spaced Repetition System)
• Anki

See Also

• Piotr Wozniak
• Damien Elmes
• Forgetting curve
• Long-term memory

Research

• Ye, J. et al. (2022–present). FSRS: A spaced repetition scheduling algorithm. GitHub: open-spaced-repetition/fsrs4anki. https://github.com/open-spaced-repetition/fsrs4anki
  Summary: The primary source — the FSRS algorithm repository, including documentation, benchmark results, and implementation. The most authoritative reference for FSRS’s design, theoretical basis, and performance versus SM-2.

• Ye, J. (2023). FSRS vs SM-2: A benchmark comparison. open-spaced-repetition project. https://github.com/open-spaced-repetition/fsrs-vs-sm2
  Summary: Systematic comparison of FSRS and SM-2 scheduling accuracy on real-world Anki review data. Demonstrates FSRS’s measurable improvements in retention prediction. The key empirical basis for FSRS’s adoption by Anki and other platforms.

• Wozniak, P.A. (1995). Two components of long-term memory. supermemo.com. https://www.supermemo.com/en/blog/two-components-of-long-term-memory
  Summary: Wozniak’s two-component model (retrievability and stability) that provides the theoretical foundation FSRS builds upon. Understanding this model is essential for understanding FSRS’s design philosophy and how it improves on SM-2.

• Anki Documentation — FSRS. https://docs.ankiweb.net/deck-options.html#fsrs
  Summary: Official Anki documentation for the FSRS scheduler, explaining how to enable it, its parameters, and practical guidance for users. Represents the mainstreaming of Ye’s algorithm into the world’s most widely used SRS platform.

Note:

• Jarrett Ye’s public profile is primarily through the open-spaced-repetition GitHub project (github.com/open-spaced-repetition). Personal biographical details are limited in public sources; this entry documents what is publicly verifiable.
• FSRS continues to be updated. The version number (FSRS-4, FSRS-5, etc.) reflects revisions to model parameters and architecture; the core algorithm and storage model are consistent across versions.