// the find
open-spaced-repetition/fsrs4anki
A modern Anki custom scheduling based on Free Spaced Repetition Scheduler algorithm
FSRS4Anki replaces Anki's SM-2-based scheduler with an algorithm grounded in actual memory research — specifically a stochastic model of memory stability and retrievability. It's for Anki power users who want their review intervals driven by data from their own review history rather than Anki's decades-old default. The optimizer is the interesting part: it fits the algorithm's parameters to your personal forgetting curves using ML.
The algorithm has peer-reviewed papers behind it (KDD 2022, IEEE TKDE 2023) — this isn't someone's weekend project dressed up as research. The optimizer personalizes 17 parameters to your review history, so the scheduler actually adapts to how your memory works rather than using population averages. FSRS is now built into Anki 23.10+, which means the scheduler half of this repo is largely obsolete in a good way — the algorithm won adoption at the platform level. The archive of candidate notebooks shows genuine iterative research methodology, not just a black-box model drop.
The primary artifact is Jupyter notebooks, which means the 'source of truth' for the optimizer is not a proper library — it's a notebook you run manually, and reproducibility depends on your Python environment. The scheduler is JavaScript injected into Anki's custom scheduling slot, which is a fragile integration point that breaks when Anki changes internals. For anyone not on Anki 23.10+, setup is nontrivial and the docs split across two tutorials is confusing. The repo itself is becoming a research archive rather than an actively maintained tool now that FSRS is in Anki core, so its long-term maintenance story is unclear.