What is FSRS (and why it matters for learning Chinese)

FSRS is an open spaced-repetition algorithm that models your memory and decides, every day, what you review in Wanyu. This guide explains what it is, where it comes from and what it changes versus the classic method — no jargon, no hype.

The problem: forgetting runs on a schedule

Since Ebbinghaus's experiments (1885) we've known that forgetting follows a curve: after you learn something, recall drops fast at first and more slowly later. Every well-timed review flattens that curve — the next lapse takes longer to arrive.

That's why re-reading word lists works so badly: you waste time reviewing what you still remember, and you arrive late to what's already gone. The idea of spaced repetition is to review each thing just before you forget it. The hard question is practical: when is "just before" — for each word, and for you?

The classic method: SM-2 (1987)

Almost all the spaced repetition you've seen descends from SM-2, an algorithm from 1987 (classic Anki): answer a card correctly and its interval multiplies — 1 day, 3, 7, 16… — fail it, and it goes nearly back to the start.

It works, and it brought spaced repetition to the world. But it's rigid: the formula is fixed and identical for everyone, it doesn't estimate the actual probability that you'll remember a card, and it punishes lapses so hard that difficult cards get stuck in endless repeats (the "ease hell" Anki forums have complained about for years).

What FSRS changes

FSRS (Free Spaced Repetition Scheduler) is an open algorithm published by Jarrett Ye and the open-spaced-repetition community. Instead of multiplying intervals blindly, it keeps a small memory model per card with three variables: the card's difficulty, the stability of the memory (how long it takes to decay) and its retrievability (the probability you'd recall it today).

• Its default parameters aren't a hunch: they come from analyzing hundreds of millions of real reviews from real learners.
• Every time you grade a card, that card's model updates, and the next review is scheduled for when your recall probability falls to the target (around 90%).
• In the project's own open benchmarks, FSRS predicts recall better than SM-2. In practice: the same retention with noticeably fewer reviews, or more retention with the same effort.
• It's not "AI in the cloud": it's a formula that runs on your device, offline. Anki adopted it as an optional scheduler in 2023; today it's the de facto standard of serious spaced repetition.

What it means inside Wanyu

Wanyu ships FSRS out of the box (the open ts-fsrs library), with nothing to configure. The "For today" screen in Study is exactly that: the reviews FSRS scheduled for today, plus whatever new words fit. You grade with a tap; the algorithm does the rest.

And the metrics are the algorithm's, not a game's: real retention and words mastered — no streaks, no lives, no fireworks. The review engine is free forever, including with your own imported decks.

What it does NOT do (the honest part)

• It doesn't remove the effort: if you don't review nearly every day, no algorithm will save you. FSRS optimizes the work; it doesn't make it vanish.
• It doesn't teach grammar or pronunciation: it schedules your vocabulary memory. The rest is still studying.
• If you already use Anki with FSRS and you're happy, the algorithm here would be the same one. What changes is everything around it — we cover that in Wanyu vs Anki.

FSRS is free software: the algorithm and its benchmarks are published at github.com/open-spaced-repetition. Wanyu uses the open ts-fsrs library. Guide written in July 2026.