FSRS for Chinese: Smarter Spaced Repetition for HSK
Short answer: FSRS is a modern spaced-repetition algorithm that schedules each word for the moment you are about to forget it. For Chinese - where you face thousands of characters and words - that means measurably fewer reviews than the older SM-2 algorithm for the same retention. Wordbrush uses FSRS by default.
FSRS vs SM-2
| FSRS | SM-2 | |
|---|---|---|
| Released | 2022, actively developed | 1980s (SuperMemo) |
| Model | 17-parameter memory model, fit to your data | Fixed ease-factor heuristic |
| Reviews for same retention | Fewer (per the open FSRS benchmark) | More |
| Used by | Wordbrush (default), Anki (opt-in since 2023) | Anki (legacy default) |
The open FSRS benchmark evaluates the algorithm on 70+ real review collections totalling roughly 1.5 billion reviews, and reports more accurate recall prediction than SM-2 - the heuristic Anki shipped by default for years. Fewer wasted reviews is the whole game when you are memorising HSK 1-3.
A worked example: one word, one lapse
Abstract claims are easy to wave away, so here is the same word run through both schedulers. Take 舒服 (shūfu, "comfortable"). You review it on day 1 and day 3 and get it right both times, then on day 10 you blank and press Again. That is the lapse. Watch what each algorithm does with the next interval it hands you.
| Review | You pressed | SM-2 schedules next | FSRS schedules next |
|---|---|---|---|
| Day 1 | Good | ~6 days (fixed second-rep step) → due day 7 | ~4 days from a modelled stability → due day 5 |
| Day 3 | Good | 6 × ease (2.5) = ~15 days → due day 18 | stability rises to ~9 days, tempered by how little time had passed → due day 12 |
| Day 10 | Again (lapse) | interval collapses to the relearn step (~1 day) and ease is docked 2.5 → 2.3 for every future interval → back on day 11 | post-lapse stability is recomputed from the word's difficulty and prior strength, not reset to zero → back in ~2 to 3 days, then growth resumes from that modelled state |
The numbers are illustrative - SM-2's depend on the card's ease factor and Anki's step settings, FSRS's on the parameters it has fit to you and your target retention (90% here) - but the shape of the divergence is real and it is the whole point. Before the lapse the two roughly track each other. At the lapse they split. SM-2 throws away most of the memory history: the interval drops to about a day and the ease penalty quietly shrinks every future interval for that word, so a single bad day taxes you for weeks. FSRS treats the lapse as new evidence about one word's difficulty, recomputes its stability, and brings it back in a couple of days without permanently punishing the schedule. Over thousands of HSK cards, each of which will lapse now and then, that difference is the reviews you do not have to sit through.
Why it matters more for Chinese
Chinese vocabulary load is front-heavy: characters, tones, and homophones all compete for memory. A scheduler that adapts to your forgetting curve, per word, keeps the daily review count sane instead of ballooning. That is exactly what Wordbrush does for every word of HSK 1-3.
Study HSK with FSRS, free →Questions
What is FSRS?
FSRS (Free Spaced Repetition Scheduler) is a modern, open-source review algorithm that models your memory and schedules each card for the moment you are about to forget it, using far fewer reviews than older algorithms.
Is FSRS better than SM-2 for Chinese?
FSRS predicts recall more accurately than SM-2, the 1980s algorithm Anki used by default. In the open FSRS benchmark across 70+ collections it needed fewer reviews for the same retention, which matters a lot for the thousands of characters and words in HSK study.
How does Wordbrush use FSRS?
Wordbrush schedules every HSK word with FSRS by default and re-fits the parameters to your own review history, so the timing adapts to how fast you personally forget Chinese vocabulary.