How OOWN turns practice into a review schedule

Recognizing the right answer among several options and typing a word from memory ask different things of a learner. OOWN has to account for that difference before it updates a review date.

FSRS uses a small set of ratings. OOWN offers several exercise formats. Connecting the two requires an interpretation of what happened in the exercise.

OOWN's scheduling implementation, reviewed on October 20, 2026, uses FSRS-6. Its job begins after an exercise result has been turned into a rating the scheduler can understand. The names Again, Hard, Good, and Easy below refer to algorithm ratings, not four buttons that appear in every exercise.

Different exercises provide different signals

For a basic card, reporting that you know the answer maps to Good. Reporting that you don't maps to Again.

For recognition exercises, OOWN considers both whether you reached the solution and how many incorrect attempts came first. A correct solution without misses maps to Good. One miss before solving maps to Hard. An unsolved item or more misses maps to Again.

Written exercises also consider mistakes and revealed letters. A completed answer with no mistakes or letter hints maps to Easy. As the combined rate of mistakes and hints rises, the rating moves through Good and Hard to Again. An unfinished answer maps to Again, even if some letters were entered correctly.

These mappings are product rules. FSRS doesn't itself define what a typo or a matching error should mean. OOWN makes that interpretation consistently before passing the result to the scheduler.

Practice and measurement can overlap

Suppose you answer a word, see it again moments later, and get it right. That second encounter can help with practice. It may offer little independent evidence about whether you'll remember the word tomorrow.

OOWN can stop repeated answers within a session, and answers that arrive too soon, from changing the FSRS state. Some exercise types don't update the schedule at all. Words introduced through the introduction flow that day are also held back from scheduling updates until the next day in the study space's time zone.

The activity can still have educational value. Withholding the update means the schedule doesn't treat every successful interaction as a new reason to lengthen an interval.

This becomes especially useful when an app combines several ways to practice the same word. The amount of activity on screen and the amount of new retention evidence needn't rise together.

Where the next date comes from

When an answer qualifies for scheduling, OOWN gives FSRS the item's previous state, the rating, and the time of review. The scheduler updates the state and calculates a new interval using the configured desired retention.

The default target is 90%: the model aims for a 90% chance of recall at the scheduled review. It doesn't mean 90% of words are learned forever, or that a particular answer is guaranteed. With other inputs fixed, a higher target calls for shorter gaps; a lower target accepts more risk of forgetting in exchange for longer gaps. A setting in the scheduling code isn't necessarily an option you can change in the interface.

What personalization means here

Each word already has its own evolving difficulty and stability. The implementation can also read a parameter set and retention target associated with a study space.

Automatically fitting those parameters from a learner's history is a further step. In the implementation reviewed on October 20, 2026, that optimization worker remained planned. The scheduling foundation can accept fitted parameters, while ordinary scheduling can run with defaults.

OOWN also records information about the state before a review, the interval, and the parameters used. That history supports examining scheduling decisions and future parameter fitting.

The practical part is yours: try to answer before revealing help, use help when you need it, and report a missed answer honestly. The scheduler needs evidence of what happened, not a perfect-looking score.

We began with a familiar problem: a word looked familiar, then wouldn't come when you needed it. A schedule can't promise that this will never happen. It can use each qualifying answer to revise an estimate and choose the next gap. The date is a plan for another attempt, not a verdict on your memory.