Spaced repetition for kanji: ratings, intervals, and honest recall

What FSRS is estimating

Kanji Mastery uses FSRS-6. The model tracks difficulty, stability, and retrievability for each review item. Difficulty describes how resistant the item is to learning, stability represents how long the memory tends to last, and retrievability estimates the chance of successful recall now.

A successful review usually increases stability and moves the next review farther away. A lapse brings the item back sooner. The schedule becomes personal because repeated results for the same item affect its future intervals.

Rate the answer you actually produced

Choose Again when you failed to retrieve the target before revealing it. Hard is for a correct answer that required serious effort or felt close to failure. Good is a normal correct retrieval. Easy should be reserved for answers that were immediate and unambiguous.

Do not upgrade a rating because the answer looked familiar after it appeared. Recognition after reveal is not the same as retrieval. Consistent ratings help the scheduler learn; trying to force longer intervals by pressing Easy produces reviews that arrive after the memory has already failed.

• Again: incorrect, blank, or revealed before recall.
• Hard: correct with substantial hesitation or uncertainty.
• Good: correct with ordinary effort.
• Easy: immediate, confident recall that felt overlearned.

Review one defined skill at a time

A card that asks for meaning, every reading, stroke order, and several words at once is difficult to rate. Partial success gives the scheduler no clear signal. Define the target before answering: recognise the meaning, read a particular word, or write the character.

Kanji Mastery keeps writing and reading scores separate for this reason. You may recognise a character perfectly but still be unable to write it. Use drills for exposure, then use the corresponding test when you want to measure a specific skill.

Control the flow of new material

New cards create future reviews. Adding a large batch on one enthusiastic day can produce an uncomfortable queue several days later. Choose a daily amount that still leaves time for due reviews, and reduce new cards temporarily when the due queue grows.

Missing a day does not destroy the system. Complete a manageable portion of overdue reviews, prioritise cards you are actively using, and resume a steady rhythm. The purpose of scheduling is durable learning, not maintaining a perfect dashboard.

Add context when a card keeps failing

Repeated lapses often mean the memory lacks a useful cue. Attach the kanji to a familiar compound, compare it with the character you confuse it with, or create a concrete visual distinction. Simply shortening the interval may increase repetition without fixing the reason for the mistake.

After changing the cue, test the item again from the same prompt you ultimately care about. A better explanation plus spaced retrieval is stronger than either one alone.