Vocabulary decays on a predictable curve: a word you learned today is half-forgotten in a week if you never see it again. Spaced repetition systems (SRS) schedule each review just before that decay wins, so a word needs a handful of well-timed re-exposures instead of dozens of untimed re-reads. For a language whose vocabulary is 100% memorization work - no cognates, no alphabet shortcuts - it is the single highest-leverage habit a Mandarin learner can build.
How the scheduling works
Every card carries a memory model: how stable the memory is and how hard the item is for you. Each rating (again/hard/good/easy) updates both, and the next review lands at the interval where recall probability hits the target - usually ~90%. Modern algorithms (FSRS, used by LinZi and recent Anki) fit the model to your actual review history; the classic SM-2 uses fixed multipliers. The practical difference: FSRS schedules shorter intervals on items you keep missing without bloating the easy ones.
The mistake almost every flashcard app makes
One direction. A card that shows 苹果 → "apple" trains recognition; the skill that fails in conversation is production - seeing "apple" and producing píngguǒ. They are different skills and they decay separately, so LinZi mints two cards per word (zh→en and en→zh), spaced so the pair never lands back-to-back, with a third handwriting card earned once the word graduates. A daily "10 new words" budget means ten words - both card faces ride along as today's allotment.
A pace that survives month three
- Start at 10 new words/day, not 50. Reviews compound - 10/day is ~40-80 reviews within weeks; 50/day becomes a second job and burns out.
- New-word budgets reset at midnight; reviews arrive on their own clock. A card rated "1 day" comes due ~24h later, not tomorrow morning - mornings are for new words, due cards trickle in all day.
- Never let a session end on an empty queue. "Learn more" extends the day when you are hot; the streak matters more than the count.
If you already live in Anki, keep it - LinZi imports .apkg decks. SRS is the engine; the content, audio and curriculum are what decide whether the engine has anything worth remembering.
Common questions
How many new cards per day is right for Chinese?
Ten new words (≈20 card faces with both directions) is the sustainable default - it generates 40-80 daily reviews within a month. Double it only after a full month without backlogs.
Is FSRS better than Anki's SM-2?
FSRS fits a memory model to your review history instead of fixed interval multipliers - measurably fewer reviews for the same retention, especially on mixed-difficulty decks like Chinese vocabulary.
Should I learn to write characters or just recognize them?
Recognition first - it is the bottleneck for reading. Handwriting earns its place once words are stable, which is why writing cards unlock after a word graduates rather than on day one.