Make desired retention more fine graned (with decimal places)

What do you think about decimal places in the “desired retention” option of FSRS?

In my particular case 82% is too mall and 83% is too big. I would like to be able to set 82.6% or like.

Is there any reason for that option to be integer?

1 Like

I think integers are precise enough for this case use.

3 Likes

This might make sense for larger values. There’s not much difference between 80 and 81. However, there’s a significant difference between 98 and 99.

I think I read somewhere that the DR vs workload graph depends on the user, so it can possibly make a difference even for smaller DR values.

In any case, I don’t think it hurts to add this feature, right?

It is a little bit awkward, but you can use filtered decks to do this e.g.:

deck:Kanji is:review prop:r<0.876 
2 Likes

Not related. You are talking about estimated retention of the card, Desired Retention is the option which controls then intervals in FSRS.

I think this would be the simple patch with no drawbacks and some advantages for a fraction of users.

It is related.

When you set Desired Retention (DR) you are asking Anki to schedule cards so they are due when their Retrievability (R / prop:r) is approximately the same as the DR.

Using prop:r in filtered decks can emulate this behaviour.

Note
It is normal for cards to not be scheduled on the exact day R drops below DR.
The real scheduler will apply Fuzzing and Load Balancing which will often cause cards to be scheduled a little early or late.
Filtering by prop:r will therefore give you slightly different results than the real scheduler because it will not apply any fuzz.

2 Likes

desired retention = show me a card when estimated retention equals desired retention