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/guide/statistics/srs-dashboard.md.
Skip to main content

SRS workload and card health

Open Stats → SRS and select the scope you want to inspect. Check the scope before comparing values: one set and All SRS Sets answer different questions.

Summary

Summary itemWhat it tells you
Cards to StudyCards available to Start SRS Review under the current schedules and limits, including eligible New cards.
Estimated RecallA probability-weighted estimate for eligible studied cards. It is a model estimate, not a counted test result.
True Retention · 30dObserved successful recall among qualifying SRS review events in the latest 30 study days.

The overdue breakdown in Cards to Study counts eligible, admitted scheduled cards due before the current study day. It is not a separate addition to the total.

See Recall and knowledge for how the latter two values are calculated.

Card State Breakdown

This chart describes the current collection, not the set of answers given in the selected period.

CategoryMeaning
NewNot yet introduced to SRS learning.
Learning / RelearningCurrently in learning or relearning steps.
Young (≤30d)Active Review cards with a scheduled interval of 30 days or less; unknown intervals also fall here.
Mature (>30d)Active Review cards with a known scheduled interval longer than 30 days.
SuspendedCards marked suspended.
DisabledCards excluded from review.

Each card belongs to one breakdown category. Disabled takes precedence over Suspended; otherwise the learning state determines its category. “Mature” describes an interval, not a guarantee that the card is permanently memorized.

Cards Studied

Use the time-range control to inspect study activity. The chart distinguishes:

  • Unique Cards: distinct cards studied per study day.
  • Repeated Attempts: answers beyond each day's first answer to a card.
  • Total Reviewed: all recorded attempts, including repeats.

For example, answering card A three times and card B once in one day gives 2 unique cards, 2 repeated attempts, and 4 total reviews.

The 7- and 30-day views use daily bars. The 90-day view groups three study days per bar. Multi-day values aggregate daily unique counts; a card studied on two different days can contribute once on each day, rather than only once for the whole range.

Future Outlook

Future Outlook shows the current scheduled workload for active Learn, Relearn, and Review cards. It excludes New, Suspended, and Disabled cards.

  • Cards due shows scheduled cards in a displayed future period.
  • Running total accumulates the chart's scheduled counts.
  • Daily load estimates recurring workload using eligible cards' current intervals, including cards due beyond the chart window.

Daily load is not simply the visible chart total divided by the number of days. A card with a seven-day interval contributes roughly one-seventh of a review per day to that estimate; intervals shorter than one day and unknown intervals contribute one per day.

The outlook is a snapshot, not a simulation of every future answer. New cards you introduce, failures, changed schedules, and your next grades can alter it. It also differs from Cards to Study because the latter applies the actual session limits and includes eligible new cards.

Grade Mix

Grade Mix counts Again, Hard, Good, and Easy answers in the selected range. Every saved review event counts, including repeated attempts. Past events remain part of this chart when their card is now suspended or disabled.

Hard, Good, and Easy are successful answers; Again is a failed answer. An increase in Again answers is a reason to inspect the material and workload, but it is not by itself the same calculation as True Retention.

Card Health

Stability summarizes the model's estimate of how durable a card's memory is, expressed in days. Difficulty describes the model's difficulty estimate on a scale of 1–10. Higher difficulty means the model treats the card as harder; it is not a difficulty rating you manually assigned to its text.

Use the histogram bars to inspect the count in a range. The displayed median is a middle value, not an arithmetic average. When a scope combines sets with different parameter profiles, difficulty can be marked Varies by set rather than implying a single directly comparable score.

Cards without usable memory history may not contribute to these estimates. Expect new collections to have sparse or empty health data.

Review Speed

Review Speed summarizes recorded SRS answer durations. It uses positive saved durations for its timed sample; missing, zero, or invalid timings do not produce a speed measurement.

The median is the middle recorded duration. Your set's Maximum Logged Time can cap newly recorded answers, while imported timings may have different characteristics. A shorter median does not by itself prove better recall—compare it with the grades and material you studied.

For empty charts, verify the scope and date range first, then check whether the relevant cards actually have the required history.