For AI agents: the complete documentation index is available at https://docs.flashnerd.com/llms.txt, the full documentation bundle is available at https://docs.flashnerd.com/llms-full.txt, and this page is available as Markdown at https://docs.flashnerd.com/reference/srs/scheduling.md.
Skip to main content

Scheduling, retention, and learning steps

Open a set's gear button and choose Scheduling Algorithm to control how learning steps and longer review intervals are scheduled.

Settings reference

SettingDefaultAllowed values
Desired Retention90%70–99%
Maximum Interval36,500 days1–36,500 days
Learning StepsOne 10-minute step1–5 steps
Relearning StepsOne 10-minute step0–5 steps; an empty list is allowed
Step length10 minutes initially1–59 minutes, 1–23 hours, or 1–3 days per step

Steps must be the same length as or longer than the preceding step, comparing the actual durations. For example, 10 minutes followed by 1 hour is valid; 2 hours followed by 30 minutes is not.

Scheduling Algorithm settings with 90 percent Desired Retention, a 36500-day Maximum Interval, and 10-minute learning and relearning steps
Retention, interval limits, and learning steps control different parts of the review schedule.

Desired Retention

Desired Retention is the scheduler's target recall probability at the next review. It is not a guaranteed score or a requirement to answer that percentage correctly today.

Increasing the target generally produces shorter review intervals and more work. Decreasing it generally permits longer intervals and more forgetting. Start with the default and examine your workload before making a large change. Changing the daily new-card allowance is a separate way to control how quickly the workload grows.

True Retention in Statistics describes observed answers. Estimated Recall is a model estimate. Neither should be read as another name for the Desired Retention setting. See Recall and knowledge statistics.

Maximum Interval

Maximum Interval caps the interval chosen for a card in Review. The default of 36,500 days is a high cap; it does not mean all cards will be scheduled that far away.

Choose a smaller cap only when you want to prevent very long intervals. Existing due dates are not a live preview of every settings edit; subsequent scheduling uses the applicable settings when you review.

Learning and graduation

New cards enter the learning sequence. Once a card reaches Review, its longer-term intervals are calculated from its history and answer grades.

Grade in Learn or RelearnStep behavior
AgainReturn to the first step.
HardRepeat the current step. At the first step, the interval can be between the first two steps.
GoodMove to the next step; graduate after the last step.
EasyGraduate immediately to Review.

Relearning is used after a Review card receives Again. If you configure no relearning steps, the scheduler returns it to Review without an intermediate relearning sequence.

Change a step sequence

  1. Open SRS Set Settings → Scheduling Algorithm.
  2. Under Learning Steps or Relearning Steps, choose Add Step or remove an existing step.
  3. Enter a positive length and choose its Time Unit.
  4. Confirm the sequence is nondecreasing. Correct any Input required or step-order message.
  5. Tap Save Change.

Changing a time unit can clamp a value to that unit's maximum. Recheck the value after switching units.

Short steps and review-ahead

A card scheduled a few minutes ahead can appear before that exact time when your global Review Ahead Limit includes it. Check daily availability before assuming the learning step was ignored.

Advanced parameters

Desired Retention and learning steps are normal scheduling controls. The 21 values in Advanced Settings are model weights with a different purpose. You do not need to enter custom parameters to use SRS. See FSRS parameters.