Findings For newly introduced cards, the average intraday retention corresponding to different initial ratings are: Again: 74.2% Hard: 91.7% Good: 96.2% Easy: N/A (the card is graduated immediately) For lapsed cards, the average intraday retention is 85.3%. Data and code are available at: Anki-button-usage/analysis.ipynb at main · open-spaced-repetition/Anki-button-usage · GitHub The current default learning steps are set at 1m 10m, with the default relearning step at 10m. This translates to the following intervals for new cards based on ratings: Again: Review after 1 minute Hard: Review after 6 minutes Good: Review after 10 minutes Lapsed cards are reviewed after 10 minutes. Suggestion For typical users: New cards: The ‘Again’ interval is excessively long The ‘Good’ review interval is insufficiently long Lapsed cards: The interval is excessively long To elevate the intraday retention from 74.2% to 90%, the interval should be reduced by a factor of 3.5. Thus, 60s / 3.5 ≈ 17s. To decrease the intraday retention from 96.2% to 90%, the interval should be increased by a factor of 3. Hence, 10m * 3 = 30m. To raise the intraday retention for lapsed cards from 85.3% to 90%, the interval should be reduced by a factor of 1.6. Therefore, 10m / 1.6 ≈ 6.25m. [image] [image] [image] Given these calculations, the proposed adjustments are: Default learning steps: 17s 30m Default relearning step: 6m For a more moderate approach, consider 30s 15m as the default learning steps.

Here is an extreme case: [image]

My suggestion only applies for the default setting for average users. For your own case, I need your data to draw the conclusion.

[image] L.M.Sherlock: To elevate the intraday retention from 74.2% to 90%, the interval should be reduced by a factor of 3.5. Thus, 60s / 3.5 ≈ 17s. To get this conclusion, you need to make an assumption that a very large fraction of these reviews were done after 60s. But, what if a significant number of users in the dataset are not using the default learning steps. Also, what if they are using the default learning steps but they allow a longer time to pass before they get back to the reviews (if they are busy with other tasks). As an extreme example, what if all of the reviews that were actually done after 1 min were Pass and only those which were delayed (because of different learning steps or busy schedules) were Fail. In this case, retention after 1 min is 100%.

You are right. According to my initial analysis, the median interval after pressing again is 114s. Seems most people tend to delay it. However, it could be another reason to shorten the learning steps because if people tend to delay, a short step may make up for it.

This is very interesting! Are you implying that shorter again intervals might be better to prevent incorrect memorization of information? [image] L.M.Sherlock: 17s 30m Any thoughts regarding 17 seconds leading to brute force memorization?

[image] Arc: Are you implying that shorter again intervals might be better to prevent incorrect memorization of information? He is aiming for 90% retention I think.

[image] L.M.Sherlock: You are right. According to my initial analysis, the median interval after pressing again is 114s. Consider that even if you do your reviews in one go, you’ll have a queue of learning cards getting due around the same time. Going through them means what was a 30 second interval becomes, say, 60 seconds. So, some cards will have a shorter delay and some will have a longer delay. Although that’s not always happening. If you spend the same amount of time for every single review, cards will appear at the right time and will get reviewed at the right time.

I guess from a mathematical point of view the extremely short intervals make sense, but 17 seconds is just absurd. That would mean 1 or 2 flashcard reviews in between… there is no chance a user can experience a genuine memory recall, and the number of total reviews will increase significantly. Aiming at a 90% retention rate or any specific target shouldn’t be enforced in the very short term. Seeing a card again in 17 seconds would not have a more meaningful impact than reviewing the card in 10 minutes or the next day (after memory consolidation during sleep).

Suggestion about the learning steps after analyzing the Anki Dataset

Anki Scheduling

L.M.Sherlock November 7, 2024, 1:07pm 13

Please wait for Anki 24.11. It will provide a API to export your data in specific format.

Topic		Replies	Views
FSRS 5: <1d Scheduling and Learning Steps FSRS	167	4645	November 25, 2024
Several FSRS-related suggestions FSRS	167	3819	October 29, 2024
Clarify what optimal retention means FSRS	77	4310	January 9, 2025
[Feature Request] FSRS Should Ignore Learning Cards Reviews FSRS	36	1163	June 29, 2025
[FSRS] Separate Desired Retention for Young and Mature cards FSRS	32	1345	July 18, 2026

Suggestion about the learning steps after analyzing the Anki Dataset

Related topics