Interday intervals for lapsed cards results in disproportionately short intervals

I’m hoping the community will have success recreating what I believe to be a bug in FSRS scheduling in Anki version ⁨24.11

The steps to recreate this are as follows:

  • Lapse on a card’s review with re-learning steps set to an interval of <1 day
  • When the re-learning card is presented again, take note of the intervals (we will call this interval set A), but do not answer the card
  • Review the lapsed card on the following day, such that the re-learning interval is now >=1 day
  • See the new intervals for various answers (we will call this interval set B)

In every instance for my deck, interval set A is significantly longer than interval set B.

A similar phenomenon can also be observed when removing re-learning steps altogether and FSRS sets the re-learning interval to >1 day. Thus the issue could alternatively be framed as “intervals are far more aggressive when re-learning a lapsed card following a re-learning interval of <1 day”.

From a learning perspective, this seems to make little sense - If I have remembered the card after a day or more, presumably stability should be greater than the same outcome occurring within <1 day (i.e. interval set B would be expected to be the same or longer than interval set A).

Version ⁨24.11 (87ccd24e)⁩
Python 3.9.18 Qt 6.6.2 PyQt 6.6.1
0.1933, 1.1883, 4.9589, 25.8480, 6.8185, 0.8389, 2.2436, 0.0010, 1.6363, 0.3619, 1.1357, 1.8182, 0.2406, 0.2460, 2.4502, 0.0000, 2.3685, 0.8411, 1.6638

@L.M.Sherlock Have you noticed this behavior?

It should be fixed in 25.02.

1 Like