Spaced repetition in Gomademy: how reviews work

Three numbers behind a review

An interval is the stored gap between reviews, measured in days. An ease factor helps determine how much a successful review grows that gap. A success count records consecutive successful ratings since the last wrong answer.

A new item starts with an interval of zero, an ease factor of 2.5, and a success count of zero. Its first due time is ten minutes after it is learned. The zero is a starting value; it does not mean the next review is due immediately.

The normal ease rules have a lower boundary of 1.3 and an upper boundary of 3.0. Ease is a scheduling setting for that card, not a score for your intelligence or your overall Japanese level.

Correctness comes first; speed helps choose a rating

In the automatic due-review flow, a wrong answer receives Again, whatever the response time. For a correct answer, the measured time selects a rating:

Scroll sideways to read all table columns. You can also focus the table and use the arrow keys.

In this version, the automatic timer excludes inactive and background time. Each retry starts a new timer; the answer-tap time is captured once for the rating and, on iOS, the saved attempt history.

These 3-second and 8-second thresholds are product heuristics. They are not validated optimal cutoffs for every Japanese task or learner. A long sentence, an unfamiliar keyboard, or a distraction can make a response slower without proving that the word was forgotten. A quick correct choice also does not establish that you can produce the word in conversation.

Some flashcard review screens let you choose Again, Hard, Good, or Easy yourself. The timing table explains the automatic flow; it does not mean every review screen measures and grades you in the same way.

Why the gaps get longer

A successful review often increases the next interval. In the version explained here, Good leaves ease unchanged. A new or reset card with an interval below one day moves to one day. On the second consecutive success, Good multiplies the stored interval by ease, with a minimum of three days. Other established Good reviews multiply the interval by ease. A card returning after a missed answer has a separate first-return rule, explained below.

If every rating is Good and ease stays at 2.5, the schedule looks like this:

First due in 10 minutes → 1 day → 3 days → 7.5 days → 18.75 days → 46.875 days.

Each number is the next gap after that review, not the total time since learning the word. The actual due time is calculated from when you answer. Waiting until a card is overdue does not, in this policy, add an extra bonus to the interval.

The longer gaps give familiar cards fewer scheduled returns. That is a practical scheduling pattern, not a guarantee that these exact dates are best for you.

The same seven-day card, four possible outcomes

Imagine an established card with a seven-day interval, ease of 2.5, and at least two consecutive successes already recorded. Each row below starts from that same state; the rows are alternatives, not a sequence.

One card. Three possible next reviews.

Starting interval: 7 days · ease 2.5

An established card with at least two consecutive successes. All three rows start here.

1. Slow correct Hard8.4 days
2. Normal correct Good17.5 days
3. Fast correct Easy24.1 days

Scroll sideways to read all table columns. You can also focus the table and use the arrow keys.

Hard still counts as a successful rating. It grows the interval by 1.2, with a minimum of one day, and lowers ease by 0.15. Later Good ratings can therefore grow more slowly, even if the Hard answer was correct.

Easy raises ease by 0.15 before calculating the interval. That is why the example uses 2.65 rather than 2.5. Its normal rule has a minimum of four days; Easy on a new or reset card schedules four days rather than one.

The seven-day Good calculation assumes an established, non-lapsed card. For a new card answered Easy and then Good, the gaps are four days and then 10.6 days: 4 × 2.65. The three-day minimum on the second success does not shorten that longer interval. The card’s history matters as well as its current gap.

What happens after a wrong answer

Again schedules a return in ten minutes, resets the consecutive-success count, and lowers ease by 0.2, down to the 1.3 floor. What happens to the stored interval depends on the card’s previous gap.

If the gap was at least two days, half is retained. For the seven-day example, the stored interval becomes 3.5 days and ease becomes 2.3. The card is still due in ten minutes; the retained interval is not that immediate wait.

At the first successful review after this lapse, Good uses the retained gap. Hard uses that same gap with a minimum of one day and lowers ease again. Easy uses the retained gap × 1.3, with a minimum of four days, and raises ease. After this first return, the usual rules apply again.

For our example, a wrong answer followed by Good answers therefore gives: 10 minutes → 3.5 days → 8.05 days. The last value is 3.5 × 2.3. Each gap begins when its corresponding review is answered.

If the previous gap was under two days, the stored interval resets to zero. Subsequent Good answers start at one day, then three days. Easy after that reset schedules four days instead. The one-day/three-day sequence is not a rule for every possible answer.

A ten-minute due time is not a ten-minute lockout. On iOS, Practise mistakes can offer an immediate retry after a batch. That retry is graded again and can replace the previously scheduled due time.

Forgetting one word does not erase all your Japanese learning. It changes the schedule for that item. Check the reading and meaning, notice what was difficult, and try recalling it again.

What the next-review caption tells you

In this version’s automatic due-review flow, the feedback caption calculates the next gap from the same policy and card state used to save the review. It is a preview of a scheduled gap, not a guarantee that you will actually open the app at that time.

The caption rounds for readability, so the 17.5-day example may be shown as 18 days. A lapsed card can show both the immediate ten-minute return and the retained interval. Other flashcard review screens may still use fixed rating labels; do not read every rating button as an exact calculated date.

What “mastered” means here

The app’s mastery heuristic requires at least three consecutive successful ratings and a stored interval of at least 21 days. A card in the all-Good example reaches that condition after the fifth successful review.

This is an app label, not a measured probability of lifelong recall. It does not prove listening comprehension, handwriting, sentence use, or readiness for a particular exam. Keep meeting words in reading, listening, and conversation too.

Related to SM-2, with its own rules

The original SM-2 specification uses one-day and six-day initial intervals and a 0–5 response-quality system. Gomademy’s one-day/three-day Good sequence and four ratings differ from that specification.

Anki’s deck-options documentation describes settings such as starting ease, Hard intervals, Easy bonuses, and learning or relearning steps. The comparison helps explain the family of ideas; Gomademy does not implement every Anki behavior, and Anki also supports FSRS.

Neither reference proves that Gomademy’s exact speed thresholds or multipliers maximize retention. The useful question is whether the reviews help you recall manageable learning targets and leave time to use Japanese outside the queue.

How is FSRS different?

Gomademy’s scheduler in this draft follows fixed rules: an answer rating changes the card’s interval and ease using the steps above. FSRS takes a different approach. It estimates a card’s difficulty, stability (how slowly its memory fades), and retrievability (the chance of recalling it now). Its scheduling uses those estimates and a desired-retention target to choose a return time.

That means there is no universal FSRS answer to “What happens after a seven-day card?” Review history, the model’s parameters, and the retention target matter. A target is a scheduling goal, not a promise that every learner will remember that percentage of words.

Gomademy does not use FSRS in the scheduler explained here. It is a future option to evaluate, not a promised migration or a proven improvement for Gomademy learners. See the FSRS algorithm documentation and Anki’s FSRS options.

Try this with one word

Choose a word you have learned, such as 山 — やま — mountain. Before checking, try to recall the answer the prompt actually asks for. Then notice the next due time and your result.

If the word keeps returning wrong, make the learning target clearer: separate reading from meaning if needed, check suitable audio, and put the word in a simple context. The scheduler chooses a return time. It cannot make an unclear prompt useful on its own.