Anki 24.10 Beta

[vaibhav]

@Keks

For some reason, the w9 parameter affects the complexity of the cards.

@L.M.Sherlock, if this is actually true, it indicates a serious bug. So, proper analysis is required.

[L.M.Sherlock]

It’s not a bug. The w[9] really affects cards when their review histories are incomplete.

open-spaced-repetition/fsrs-rs/blob/cd80a968917ab2dc8a989f1b28e6d294a1a7e435/src/inference.rs#L109-L135

[Keks]

Now it’s clear what the problem is. This happens in those decks where “Ignore cards reviewed before” is set to a future date. This completely breaks FSRS.

And as I pointed out above, this happens not only when changing w9.

[L.M.Sherlock]

Is it a problem? It’s intended:

How does the scheduler work?

Please see the section “Ease Factor to Difficulty”.

[Keks]

It was an unexpected behavior.
The workaround doesn’t seem to cause any problems.

[L.M.Sherlock]

I don’t know what the unexpected behavior is. Do you mean the difficulty shouldn’t be changed when the w[9] is changed?

[Keks]

Optimize all presets

I have always lacked the ability to make exceptions in the “Optimize all presets” function. I had the idea to enter such a search query so that it always finds 0 maps and thereby prevents the optimization of the preset I need. The question is, won’t I break something in this way? [image] Or maybe there is a better way to do this? For example, set ignoring reviews to an unrealistic date. For example, the year 2200. The second question is not quite on the topic and not quite on FSRS. Is there a…

Optimize all presets

I mean, it looks like it works, but I’m not sure if it could lead to some undocumented behavior. @L.M.Sherlock sorry for pinging you twice, but I just want to verify that setting “Ignore reviews before” to a future date is the same as turning the optimizer off, with no side effects?

There is a need to exclude a preset from Optimize All Presets.
But it turned out that setting the date for a future date changes the FSRS operation in an unpredictable way.

This led to the fact that many cards received difficulty close to 0 or 11, even though they had only Good answers in the history of the review.

[Aleksej]

I noticed earlier that I had messed my preset queries for the movies2anki decks up. They had “deck:*m2a*”, which included cards from filtered decks with “nonm2a”. I changed them to “note:movies2anki*”. Also, it seems that optimizing on all m2a decks may be better than separating a set of clearer shows with the same VAs.

I am not sure that changed much when I did it (the presets for the m2a decks may have become easier on average, like 53→48), but now it’s like this: there are fewer “extremely difficult” cards and more “extremely easy” cards. The m2a decks have spikes of extreme difficulties, too, although they don’t comprise most of them.

Whole collection graph including suspended:

-is:suspended shows fewer cards on the right end.