FSRS-5 vs SM-2: Why Modern Schedulers Review Less and Remember More
A synthesis of published research and the open-source FSRS project: how difficulty/stability/retrievability modeling aims to cut review workload compared to the 1987 SM-2 heuristic — and why we made FSRS-5 the default in ClozeForge.
TL;DR What the Evidence Says
The 37-Year-Old Heuristic Behind SM-2
Piotr Woźniak designed the SM-2 algorithm in 1987. It was a monumental breakthrough for early computer-aided learning, but it relies on a simplification: a single static 'Ease Factor' applied multiplicatively to every card. In reality, memory is better described as a multi-dimensional system. The FSRS project models three separate variables: Difficulty (how hard an item is for you), Stability (how long the memory trace is likely to last), and Retrievability (the current probability you can recall it right now).
Under SM-2, easy items tend to be reviewed more frequently than strictly necessary out of algorithmic caution, while a single mistake can send a hard item back toward a short interval — regardless of months of history.
How FSRS-5 Re-Engineers the Spaced Repetition Queue
FSRS (Free Spaced Repetition Scheduler) is an open-source algorithm developed by the open-spaced-repetition community, with parameters trained on large collections of anonymized review logs. Instead of a static multiplier, it computes when a card's retrievability R(t, S) is predicted to decay to your target retention threshold (90% by default). Items you know well get intervals that expand into weeks; items you struggle with get focused reinforcement — without erasing your historical progress. FSRS has been available as the native scheduler in Anki since version 23.10.
The FSRS project's stated goal is to achieve a comparable or better level of retention with fewer reviews than legacy SM-2 scheduling. Published comparisons come from the FSRS community itself; we cite them as the algorithm's design aim, not as our own measurements.
Why Fewer Reviews Matters for Real Learners
The case for smarter scheduling is not about leaderboard numbers — it is about adherence. Vocabulary research consistently shows that durable word learning needs many encounters per word (Pigada & Schmitt, 2006, observed roughly 8–10+ encounters; Webb, 2007 found around 10 repetitions for durable learning). Every wasted review of a word you already know is an encounter you could have spent on a word you don't. An efficient scheduler concentrates effort where retrievability is genuinely at risk.
Language learners rarely quit because grammar is hard; they quit because bloated review queues become unsustainable. Scheduling each item only when it is actually due keeps daily practice short.
Full Portability to Anki
We believe in total learner sovereignty. ClozeForge includes a 1-click Anki TSV exporter that exports all your practiced sentences, translations, and audio tags directly into Anki desktop — where FSRS is available natively since Anki 23.10. You own your memory vault forever.
FSRS is fully open-source (github.com/open-spaced-repetition), auditable by anyone, and interoperable with modern Anki FSRS profiles. Nothing about your schedule is a black box.
Related Research & Deep Dives
Ready to forge fluency with FSRS-5?
Start with 1,600 authentic cloze sentences today in our free live app.