Closed janispritzkau closed 3 months ago
I recommend using two decks with different presets: one for recognition, one for reproduction. The initial interval will be longer in your recognition deck, and shorter in your reproduction deck.
I'm gonna try that. Are there things to consider when pulling in new cards? Like
- Which card type should come first? Or randomize?
I think recognition first is better.
- Intentionally keeping a gap between learned notes of one card type?
It's hard to say. I have not recommendation here.
Before posting, please read the following:
Research
Enter an [x] character to confirm the points below:
Question
In one of my vocabulary decks I have two card types, one for recognition and one for reproduction (fill in the blank). I have been using FSRS pretty much from the beginning (4 months). Since I added the other card type, the initial interval has increased significantly, to the point that I had to manually adjust the first 4 parameters to prevent the interval from going straight to 3-4 months when pressing good or easy. I suspect the problem is that the second card type is usually shown shortly after the first one, and so there is a knowledge transfer that FSRS is not aware of.
I've been trying FSRS v5 for a few days, it seems to have exaggerated the problem even more. I wonder if I am using it wrong or if this is just something that has not been considered yet. These are the parameters (v4.5) which I get when optimizing on the entire history:
It gets slightly worse if I only optimize on the last 3 months:
Additionally, I use "again", "hard", and "good" pretty consistently. However, I tend to choose between "good" and "easy" a bit arbitrarily, to nudge the next interval in a certain direction. I'll be happy to answer any follow-up questions.