How Literate Chinese Schedules Your Flashcard Reviews (FSRS)
Literate Chinese now schedules every flashcard review with FSRS, the open-source Free Spaced Repetition Scheduler, using the FSRS-7 model through the ts-fsrs library. Instead of moving every word up the same fixed ladder of intervals, the app keeps a separate memory model for each word and brings it back when your chance of recalling it is about to drop below 90%. It is the same family of algorithm that Anki offers as its modern scheduler, and it shipped on November 4, 2026.
If you just want to study, you can stop here: tap the honest answer and the scheduler handles the rest. If you are the kind of learner who likes to know exactly why a card showed up today, the rest of this post is for you.
- Every word has its own stability (roughly, how many days until your recall chance falls to 90%) and difficulty, updated after each answer.
- The app targets 90% recall, fuzzes intervals slightly, caps any interval at 2 years, and uses a 10-minute first learning and relearning step.
- Nothing retires forever. Words with 180+ days of stability show as mastered, but they still come back when due.
- Reading counts. Finishing a story gives the bank words you read without help half the credit of a "Know it" answer, without using your daily review allowance.
- The old 10-step ladder is gone; your existing progress was converted into FSRS memory states, not reset.
What is FSRS and how does it work?
FSRS is a spaced repetition algorithm that predicts the probability you can recall a card right now and schedules the next review for the moment that probability reaches a target you choose. It was developed in the open by the open-spaced-repetition project on GitHub, and it is built on what researchers call the DSR model of memory. Three numbers describe each card:
- Retrievability (R): the probability you can recall the word at this moment. It falls as time passes since your last review.
- Stability (S): the number of days it takes for retrievability to fall to 90%. A word with a stability of 30 days has about a 90% recall chance a month after you last saw it.
- Difficulty (D): a score from 1 to 10 for how hard the word is for you. High difficulty means each successful review raises stability less.
Between reviews, retrievability follows a forgetting curve. In current FSRS versions that curve is a power law: memory fades quickly at first and then more slowly, which matches real review data better than a simple exponential. When you answer, FSRS updates stability and difficulty and computes the next interval as the point where the curve crosses your target. The project's algorithm wiki documents every formula and parameter, version by version.
Its parameters were not guessed. The project's public SRS benchmark fits and tests schedulers against a dataset of about 727 million flashcard reviews from 10,000 Anki users, and FSRS-7 is the newest model in that benchmark.
How is FSRS different from SM-2 and the old Anki algorithm?
SM-2, the algorithm behind Anki's classic scheduler, multiplies each interval by an "ease" factor that starts at 2.5 and goes up or down with your answers. It has no idea how likely you are to remember a card; it only knows the last interval and the ease. FSRS models recall probability directly, which is why Anki's own FAQ says FSRS needs fewer reviews to reach the same retention and is much better at scheduling cards you review late.
Lateness is a good example. Say you miss a week of study and a word is ten days overdue. If you still remember it, that is strong evidence your memory was better than predicted, so FSRS gives it a bigger stability gain than an on-time success. A fixed ladder or a plain multiplier cannot tell the difference between "right on schedule" and "right after a long gap".
What changed from the old Literate Chinese schedule?
Before November 2026, Literate Chinese used a fixed 10-step interval ladder: every word climbed the same steps, and a word's only state was which step it was on. FSRS replaces the ladder with a memory state per word, so an easy word like 你好 can space out fast while a slippery one like 遗憾 (遺憾, yíhàn) keeps coming back until it sticks.
| Old 10-step ladder | FSRS-7 (since Nov 2026) | |
|---|---|---|
| Where intervals come from | One fixed sequence of steps, the same for every word | Computed per word from its stability, difficulty, and the 90% recall target |
| Hard vs easy words | Same pace up the ladder | Difficulty (1–10) changes how fast stability grows |
| Overdue reviews | A step is a step, however late | A late success earns a bigger stability gain |
| Longest interval | Top of the ladder | 2 years, with slight fuzz so reviews don't bunch up |
| Mastered words | Reached the top step | Shown as mastered at 180+ days of stability, still reviewed when due |
If you used the app before the switch, your progress came with you. Old step levels became stability, your past accuracy on each word became its difficulty, and words you had already mastered got a year of stability, with their reviews spread across the coming year so they don't all land on one day.
What does each flashcard answer do?
You answer with two buttons, because two honest choices are faster and more reliable than four you have to agonize over. "Know it" is FSRS's Good rating and "Don't know" is Again. A few extra controls let you override the schedule when you know better than the model.
| Action | What happens to the word |
|---|---|
| Know it (Good) | Stability grows. It grows more when the word is still young, when its difficulty is low, and when you got it right despite being overdue. |
| Don't know (Again) | A lapse: stability shrinks, difficulty goes up, and the word returns in 10 minutes for relearning. |
| Three misses in one session | The word is snoozed instead of coming back over and over in the same sitting. |
| Master | Sets the word to a year of stability. It still comes back when due. |
| Reading it in a story | Half of a "Know it" gain, once per word per day (details below). |
There is also a Boost option that jumps a word straight to about three months of stability, for words you already know well but never reviewed in the app. The study guide inside the app has a "Full breakdown of the scheduler" sheet if you want every control explained on your phone.
Why does Literate Chinese aim for 90% recall?
Ninety percent is the point where you remember almost everything you review without paying for it in a flood of reviews. It is also FSRS's standard default, and Anki's deck options guide describes the same 90% as a good balance of retention and workload. Push the target higher and intervals shrink quickly, so each extra point of retention costs more daily reviews than the one before. Push it lower and you forget more words between reviews, which hurts most when you are trying to read.
The 2-year maximum interval is a reading-first choice. A word you have not seen in two years is worth one quick look, and checking costs you a second.
How does a new word get its first difficulty?
Before you have ever answered a word, the app estimates how hard it is likely to be from its HSK band, its length, how many senses it has, and whether it uses rare characters. That estimate is blended 50/50 with the difficulty FSRS derives from your first answer. The result: 喜欢 (xǐhuan) and 纠结 (糾結, jiūjié) start in sensibly different places even if you get both right the first time, and your own answers take over from there.
Does reading a story count as a flashcard review?
Yes. When you finish a story, every word from your flashcard bank that you read with its reading hidden, and did not tap for help, gets credit worth half of a "Know it" answer. Each word can earn this once per day, and it does not use up your daily review allowance. The story footer tells you how many words it counted.
This is the part I care about most. Recognizing 已经 (已經, yǐjīng) in the middle of a real sentence is a genuine act of recall, arguably a better one than seeing it alone on a card, and the scheduler should know it happened. Half credit is deliberately conservative: context helps you guess, so a story read counts for less than a clean card answer. If you want to see this in action, try reading your first story with pinyin hidden. For why cards alone stall so many learners, see why flashcards aren't enough to read Chinese.
How many reviews will I get each day?
By default, up to 50 reviews a day, and you can change the limit. The study setup card shows the split as "N today, M waiting", so you can see when a backlog builds up. If you have time and want to clear it, "Review more" keeps going past the limit. Because overdue words that you still remember earn bigger gains, a backlog is less costly under FSRS than it looks: you are not starting over, just catching up.
What's next: parameters fitted to you
Right now every learner uses FSRS's default parameters, which the project fits on large collections of real reviews. The app already keeps a log of your reviews and syncs it, so that it can later fit the parameters to your own memory with the FSRS optimizer, the same way Anki's "Optimize" button does. That per-learner fitting is coming; it is not live yet.
Frequently asked questions
Is Literate Chinese's scheduler the same as Anki's FSRS?
It is the same algorithm family from the same open-source project. Literate Chinese runs the FSRS-7 model through the ts-fsrs library with a 90% recall target and a 2-year interval cap. Anki offers FSRS as an option you switch on in its deck options and tune yourself; in Literate Chinese it is simply on.
Do mastered words ever stop appearing?
No. A word shows as mastered once its stability reaches 180 days, but it still comes back when its recall chance drops to the target. With a 2-year cap, even your strongest words get an occasional check.
Why are there only two answer buttons?
"Know it" and "Don't know" map to FSRS's Good and Again ratings. Two clear choices keep reviews fast and honest, and FSRS still adjusts each word's difficulty from your pattern of answers over time.
Did the update reset my flashcard progress?
No. Your old step levels were converted into stability and your past accuracy into difficulty. Words you had already mastered received a year of stability, with their reviews spread over the coming year.
