How Spaced Repetition and FSRS Work in LangBlitz
Every time you rate a word card in LangBlitz, the app is quietly deciding exactly when you'll see that word again. This guide walks through how that decision actually gets made, in plain language, no background in memory research or algorithms required.
The Problem Spaced Repetition Solves
Human memory does not fade at a constant rate. Right after you learn something new, you forget it quickly if you do not reinforce it. But every time you successfully recall something, especially if some time has passed since you last saw it, the memory gets stronger and the next forgetting curve gets flatter. This is sometimes called the spacing effect, and it has been studied in psychology for well over a century.
The practical implication is that reviewing a word you just saw five minutes ago barely helps you remember it next month. Reviewing a word right before you would have forgotten it, even if that is weeks after you first learned it, helps enormously. The hard part is knowing when "right before you would have forgotten it" actually is for any specific word and any specific person, since it depends on how difficult that word is for you personally and how many times you have successfully recalled it before.
That is the problem spaced repetition software exists to solve. Instead of you guessing when to review something, the app tracks your performance on every individual word and calculates the ideal moment to bring it back.
What FSRS Is
FSRS stands for Free Spaced Repetition Scheduler. It is a modern, openly published algorithm that has largely replaced older approaches used by early flashcard software, which tended to rely on simpler fixed multiplier systems, for example doubling the interval every time you got a card right regardless of how difficult that specific card had actually been for you in the past.
FSRS instead models each word with a few numbers that update every time you review it. The two most important are stability and difficulty. Stability is roughly how many days can pass before your recall probability for that word drops to a certain threshold. A high stability word can go a long time between reviews and you will likely still remember it. A low stability word needs to come back soon or you will probably forget it. Difficulty reflects how hard that specific word tends to be for you, independent of how long it has been since you last saw it. Some words are just harder to hold onto than others, and FSRS learns this per word rather than assuming every word behaves the same way.
LangBlitz uses FSRS version 4.5, with scheduling weights that match the published reference implementation, so the underlying math is not a custom or simplified variant, it is the same algorithm used by other serious spaced repetition tools.
The Four Card States
Every word in LangBlitz sits in one of four states at any given time.
New means you have never reviewed this word before. It has not been introduced yet and carries no scheduling history.
Learning is the state a word enters right after you first see it, while its initial stability is still being established. Words in this state tend to come back for review relatively quickly, sometimes within the same day, as the app gets an early read on how well you are retaining it.
Review is the main long term state. Once a word has graduated out of learning, it moves here, and this is where stability really starts to grow over successful reviews, pushing the interval between reviews further and further out as long as you keep recalling it successfully.
Relearning is where a word goes if you rate it Again while it is in the Review state, meaning you forgot something you had previously learned well enough to graduate. This is not a failure state to be worried about. Forgetting things and relearning them is a completely normal part of how spaced repetition works, and a word that has been relearned once often ends up with a more accurate, more resilient schedule afterward than one that was never challenged.
These four states are not the same thing as the mastery percentage shown on the home screen and in Vocabulary World, even though the two are related. A card's state describes where it sits in its review lifecycle, brand new, being confirmed for the first time, stable long term, or being relearned after a forget. Mastery, covered later in this guide, is a separate, continuously shifting estimate of how well you would recall a word right now. A word can sit in the Review state, the most advanced of the four, for months, while its mastery percentage still rises and falls in between individual reviews depending on how much time has passed since you last saw it.
What the Four Ratings Actually Do
When you flip a card and rate it Again, Hard, Good, or Easy, that rating feeds directly into the calculation of the word's new stability and difficulty, which together determine the next interval.
Again tells the algorithm you did not remember the word. This drops stability significantly, increases difficulty somewhat, and schedules a short interval so you see the word again soon while it is fresh in your mind from this correction.
Hard tells the algorithm you remembered it, but not comfortably. Stability still grows, but by less than a Good rating would produce, and difficulty ticks up slightly, reflecting that this word is proving to be genuinely harder for you than average.
Good is the default, expected outcome, remembering the word without excessive effort. Stability grows by a solid, unremarkable amount, and difficulty stays roughly where it was.
Easy tells the algorithm the word was no challenge at all. Stability grows by more than a Good rating would produce, difficulty drops slightly, and the next interval is pushed out further than it otherwise would have been.
It is worth being honest with these ratings rather than always tapping Good to move quickly through a session. Consistently rating a genuinely difficult word as Good, when it actually took real effort to recall, pushes its next review further out than is ideal, which increases the odds you forget it before it comes back around. The four options exist specifically so the schedule can adapt to genuine differences in difficulty between words rather than treating every successful recall the same way.
Why Mastery Can Drop Without You Doing Anything Wrong
LangBlitz shows a mastery percentage for each CEFR level, both on the home screen and in more visual detail inside Vocabulary World. This number is based on a concept called retrievability, which is the algorithm's current estimate of the probability you would successfully recall a given word right now, at this exact moment, if tested.
Retrievability decays continuously from the moment of your last review, following a curve shaped by that word's stability. A high stability word decays slowly, so its retrievability stays high for a long time even without review. A low stability word decays faster. This means it is completely normal to open the app after a week away and see your mastery percentage for a level you had previously gotten to a high percentage tick downward slightly, even though nothing you did was wrong. The words themselves have not gotten harder. The estimated probability that you would still remember them right now, without review, has simply gone down a little because time has passed. Reviewing those words again brings retrievability back up and, if you recall them successfully, increases their stability so the same amount of time passing next time will cause less decay.
Smart Pacing and Review Intensity
Two settings directly shape how the scheduling interacts with your daily study habits. Smart new word pacing, when enabled, automatically slows down how many brand new words get introduced per session as your due review backlog grows, so a busy week where you fall behind on reviews does not also pile on a large number of brand new words on top of an already large backlog. It only tapers, it never fully blocks new words unless the backlog becomes very large relative to your own new words per session setting.
Review intensity, with options for Light, Balanced, and Thorough, adjusts the overall aggressiveness of the review load the app aims for. This is a separate setting from your CEFR practice level or new words per session limit, and exists purely to let you tune the overall daily time commitment the app is optimizing toward, without changing how the underlying FSRS math itself works.
A Worked Example
It helps to walk through what a single word's journey through the system might actually look like. Say you are introduced to the word "meticulous" for the first time. It starts in the New state with no history. You see it, read the definition and example sentence, and rate it Good, since you had not encountered this exact word before but the meaning made sense once explained. It moves into Learning, and because it is brand new, the app schedules a short first interval, perhaps later the same day or the next day, to confirm the initial exposure actually stuck.
You see it again the next day, still recall it, rate it Good again, and it graduates into the Review state with a modest starting stability. A week later it comes due. You remember it easily this time and rate it Easy. Stability grows by a larger jump than a Good rating would have produced, and the next interval stretches out to perhaps two or three weeks rather than one.
Now suppose three weeks later, when it comes due again, you have genuinely forgotten it and rate it Again. It drops into Relearning, stability falls back down significantly, and you see it again soon, maybe the next day. You get it right this time, it graduates back into Review, but with a difficulty value that has ticked up slightly compared to before, since the algorithm has now learned that this particular word is a bit harder for you to hold onto long term than average. From here on, its intervals will tend to be somewhat more conservative than a word you have never forgotten, even if you go on to answer it correctly many times in a row.
This is the entire point of tracking stability and difficulty separately for every single word rather than applying one universal schedule to everything you study. Two words you currently both know equally well right now can still be scheduled very differently going forward, based on how consistently you have actually remembered each one over time.
What This Means in Practice
None of this requires active management on your part beyond rating cards honestly. The scheduling happens automatically, word by word, review by review. Understanding the mechanics mostly helps in two ways: it explains why an "easy" word can occasionally come back sooner than you expect if it has a rocky history, and it explains why letting your mastery dip during a busy week is not something to be alarmed by, since a handful of focused review sessions afterward brings it back up quickly. The two quiz formats built on top of this same scheduling data, Daily Quiz and Practice Quiz, are covered in the next guide.