How Alma flashcards actually work: FSRS, deck options, Due Today and Stats

Disclosure: Alma publishes this article and makes one of the study tools it discusses.

If you are a medical, nursing, NP or PA student who lives in flashcards, Alma’s system is built around the FSRS spaced-repetition algorithm but with its own settings, deck options and stats. It is not Anki, does not promise any score or memory outcome, and you control how aggressive it is with intervals, new cards and exam crunch time.

This guide walks through what actually happens when you tap Again, Hard, Good or Easy, how FSRS thinks about intervals, how Alma’s Deck Options, Due Today, Catch up, Quiz me on my misses and Stats fit together, and when AI helps versus when you should lean on primary resources.

What does a study session look like in Alma?

Open a deck and press Study Now. The card shows its front, and clicking it or pressing Space flips it. Before flipping, the learner can optionally tap Guess, Unsure or Sure (keys 1 to 3); that confidence tap is stored with the rating and feeds the confidence chart in Insights, and it can be turned off from the card’s More menu. After the flip, the learner rates the card Again, Hard, Good or Easy (keys 1 to 4), and each button shows the interval it will give, such as “10m”, “3h” or “4d”. A card rated Again or Hard that is due again within 20 minutes comes back later in the same session. Ctrl or Cmd plus Z undoes the last rating within 30 minutes. When the queue is empty, “Session Complete!” shows the cards studied, the time spent and the counts of each rating.

How do the Again, Hard, Good and Easy buttons in Alma actually work?

Alma’s study screen uses four buttons — Again, Hard, Good, Easy — under each flashcard. Internally, FSRS treats Again as a failed recall and Hard/Good/Easy as successful recalls, and schedules the next review using your desired retention setting.^[1][2][3]

FSRS — the Free Spaced Repetition Scheduler — models each card with difficulty, stability and retrievability.^[4][5] Stability is the interval in days at which your chance of recall falls to 90%.^[4][3][5] Retrievability is your current recall probability, which decays over time along a forgetting curve.^[6][4] FSRS predicts when retrievability will fall to your desired retention and schedules the card for that day.^[6][4][7]

When you rate a card, FSRS records the rating, updates stability and difficulty, and then computes the next interval and due date.^[3][5] Again is treated as forgetting; it cuts stability and returns the card to a short interval, often with same-day learning steps.^[1][2][3][5] Hard, Good and Easy all signal a successful recall, but with different effort, so intervals grow by different amounts.^[2][5] Easy stretches the interval most; Hard the least.^[2][5]

FSRS internally treats Again as “fail” and Hard/Good/Easy as “pass”.^[1] Pressing Hard when you actually forgot misleads the algorithm and can make future intervals unreasonably long, because it assumes a successful but hesitant recall.^[1] The most honest default is usually Good; FSRS then grows stability by the amount it was aiming for.^[2]

What recall probability is FSRS aiming for when cards come due?

FSRS schedules each card so that, when it appears, your recall probability is at your desired retention target rather than some fixed maximum: target retention is defined as the probability that you can recall a card at the moment it comes up for review, and FSRS uses the forgetting curve to find the day when retrievability will drop to that value.^[6][4]

What is FSRS and how is Alma’s scheduler different from Anki’s?

FSRS is a spaced-repetition scheduler based on the DSR model that predicts your recall probability for each card and lets you choose a target retention; it is integrated into several apps and can reduce reviews versus older SM‑2 style schedulers for the same retention.^[4][8][5] Alma uses FSRS but its scheduler is separate from Anki’s and not Anki-compatible.

Anki later integrated FSRS as an optional algorithm alongside its legacy scheduler, with FSRS parameters based on a large public dataset of review logs and benchmarks showing lower prediction error (fewer reviews for the same retention) compared with SM‑2.^[8][4] FSRS can also be used with optimisers that fit its parameters to your own review history, further improving accuracy.^[1][4]

Alma’s implementation uses FSRS and offers a desired retention setting, but is its own scheduler and is not guaranteed to match Anki’s behaviour. Alma does not claim any particular retention percentage or exam outcome; deck options affect workload, not a predictable score. As of October 21, 2026, Alma uses the default FSRS parameters and does not train them on each learner’s own review history, so an optimiser-based setup in Anki can behave differently.

How do Alma’s Deck Options control new cards, learning steps and desired retention?

Each Alma deck has options that let you choose study order, how many new cards and reviews to see per day, learning and relearning steps in minutes, a desired retention setting, and a leech rule. These per‑deck settings shape your workload and how quickly FSRS pushes intervals out, without promising specific memory results.

Both daily limits are true daily allowances, so closing and reopening a deck does not hand out a fresh batch. Cards still being learned are not counted against the review cap.

What does “desired retention 70–97%” actually mean?

The desired retention slider in each deck’s options sets FSRS’s recall‑probability target for that deck. FSRS documentation and related guides describe desired retention as the fraction of cards recalled successfully when due and note that higher targets increase workload while improving accuracy only modestly, whereas lower targets save reviews at the cost of more forgetting.^[1][6][4][3]

FSRS‑focused guides emphasise that the “target retention” dial is more meaningful than hand‑editing individual intervals: changing it shifts the whole deck’s schedule so that intervals collectively match the recall rate you ask for.^[6][7] In Alma, this means deck options give you one clear way to trade off daily workload against how often you are willing to relearn lapses.

How do leeches, Exam Mode and sub-decks behave?

Deck Options also include a leech threshold and action, an Exam Mode that compresses intervals toward a chosen exam date, and support for sub‑decks that roll up under a parent deck. These controls help you tame stubborn cards, front‑load critical topics before an exam, and study big imported decks from the top level.

In Alma, a card forgotten as many times as the leech threshold (the default is 8; 0 turns it off) is marked as a leech is marked as a leech; you can choose whether Alma just marks it or also suspends it. FSRS‑based tools commonly recommend rewriting leech cards rather than hammering them indefinitely, because repeated failures are usually a cue to improve the card, not review it more.^[9]

Sub‑decks appear under parent decks; a parent’s Study Now button studies all cards in its sub‑deck tree, with the parent’s daily limits capping the overall session and each sub‑deck’s limits capping its contribution. This is particularly useful when you import large hierarchies (for example, a curriculum organised by system) and want one queue across them instead of jumping between many small decks.

How do Due Today, Catch up and “Quiz me on my misses” work together?

Due Today appears at the top of the deck list with a badge showing the number of cards ready across all decks. Its play button starts a single session that draws from each top‑level deck’s due reviews and new cards, obeying that deck tree’s daily limits and ordering reviews mainly by due date. This lets you clear everything due without opening decks one by one.

Catch up is designed for when you have missed days and have a backlog. It selects a fixed‑size batch of the most overdue cards across all decks and orders them by how likely they are to be forgotten right now. Cards you do not see stay due for later; the dialog tells you how many remain.

Quiz me on my misses appears in the sidebar once a card has been rated Again at least twice in the last 7 days. Alma writes one practice question per such card, five by default and at most eight, in the learner’s exam style. Each question says which card it came from, and the dialog ends with a score such as “X of Y right”. Nothing from this quiz is saved, and it counts as one quiz against the plan’s quiz allowance. Catch up studies a capped batch (50 cards by default, 200 at most) and does not clear the whole backlog.

How does Alma’s 4 am study day and Stats page help you pace yourself?

The study day starts at 4 am in the learner’s own time zone, and daily limits reset then.

The Stats page begins with headline tiles: cards ready to review (across all decks), total cards, mature cards (those with intervals of at least 21 days), and retention, defined as the share of cards answered successfully. Guidance on FSRS and target retention focuses on “the fraction of cards recalled successfully when they are due,” which is the measure most aligned with what spaced repetition is trying to control.^[1][6]

Below the tiles you see Cards by State (New, Learning, Review), a Study Streak (current, longest, total study days) and a year‑long Study Activity heat map that visualises how consistently you have shown up. This kind of calendar‑style activity view echoes how many FSRS and Anki users think about the habit side of spaced repetition — consistent small sessions rather than irregular marathons.^[7]

None of these stats predict exam scores. They are mirrors for your spaced‑repetition behaviour: whether you are consistent, whether your target retention is realistic, and where your workload may be creeping beyond what you can sustain.

What is AI good and bad for in Alma’s flashcards and quizzes?

Alma relies on FSRS for scheduling and uses AI around that core: to help generate flashcards from your materials, to turn missed cards into small quizzes, and to explain answers. AI can speed up card creation and add variety in practice, but it is not a substitute for primary resources, official exam content or clinical guidelines.

Spaced‑repetition research emphasises that writing better cards and reviewing consistently matter more for outcomes than which specific scheduler you use, even though FSRS generally predicts recall more accurately and can reduce reviews relative to older SM‑2 style algorithms.^[4] Alma’s AI features sit on top of FSRS; they do not change the core scheduling logic and do not guarantee a particular memory or exam result.

Any time you use AI to generate or edit content — cards, questions, explanations — you should verify the facts against trusted primary resources (textbooks, guidelines, official exam materials). AI can hallucinate details or simplify in ways that are unsafe for clinical reasoning. Alma is a study tool, not a question bank of real exam questions and not a clinical decision tool. It is there to help you practise, not to replace judgment or authoritative sources.

How should you actually use Alma’s flashcards day to day?

Used well, Alma’s flashcards let you keep large decks moving with realistic daily reviews. In practice, that means setting sane new‑card and review caps, leaving desired retention near the default unless workload is extreme, using Due Today and Catch up to stay current, and watching the Stats forecast to avoid unsustainable spikes.

When you rotate into intense exam blocks, Exam Mode and a slightly higher retention target can bias FSRS toward more frequent reviews on that exam‑critical deck, while other decks stay on a slower track. After the exam, you can turn Exam Mode off and ease retention back down to something sustainable.

Throughout, remember that Alma’s scheduler, like Anki’s FSRS mode, is an implementation of an algorithm, not a guarantee. FSRS can make your reviews more efficient than older SM‑2 scheduling by predicting recall probabilities and scheduling against a target,^[4][8] but it cannot learn the material for you. That still depends on how you read, think, discuss cases, and practise questions off the cards.

Frequently asked questions

Does Alma guarantee I will remember 90% of my cards or pass my exam?

No. FSRS defines “desired retention” as the fraction of cards you want to recall successfully when they are due and schedules reviews to aim for that,^[6][4][3] but real recall depends on how you study, the quality of your cards and your consistency. Alma does not guarantee any retention percentage or exam outcome.

What happens if I often press Hard instead of Again when I actually forgot?

FSRS treats Again as a failed recall and Hard/Good/Easy as successful recalls.^[1] If you press Hard when you actually forgot, FSRS assumes you remembered with difficulty, which can produce unreasonably long intervals because the algorithm thinks the card is more stable than it is.^[1] That means more future lapses and inefficient scheduling.

Why does Alma start the study day at 4 am instead of midnight?

FSRS implementations commonly start the “new day” at 4 am so late‑night sessions count toward the previous day rather than pulling in tomorrow’s reviews.^[3] Alma follows this pattern to make nocturnal or post‑call studying less punishing: a 1 am session still belongs to the previous evening’s workload.

References

1. fsrs4anki/docs/tutorial.md at main · open-spaced-repetition/fsrs4anki · GitHub
2. Again, Hard, Good, Easy: what your rating actually changes — Kikusho
3. How Scheduling Works | True Recall
4. FSRS vs SM-2: Which Spaced Repetition Algorithm Should You Use? — Memori
5. The Algorithm · open-spaced-repetition/awesome-fsrs Wiki · GitHub
6. How to choose a retention target (the 90% question) | Yazu
7. The Best Spaced Repetition Intervals (And Why 1-3-7 Is Wrong) — Memori
8. Anki - Wikipedia
9. flowss — Study · Docs