I created a blank FSRS deck and I added 10 test cards. Each time I see a card, I mark it “Again”. I would expect that after a while doing this, it should infer that these cards are so difficult that I should be seeing them sooner and sooner. Eventually, I would expect it to show me the same card immediately after I “Again” it.
But this doesn’t happen. After pressing “Again” 154 times in a row at 1 second per card, each card has difficulty 74%, and it still cycles through many other cards before showing me the first one.
This makes it very difficult to learn each card because between any two reviews of the same card there are reviews of too many other cards and I can’t remember the first one by the time it returns.
Why doesn’t it converge to showing the same card immediately after enough failures? What is it doing instead?
How can I make it show the same card immediately after enough failures?
For new cards Anki uses learning steps. You can tweak these in your deck settings. The fsrs4anki helper addon can give you recommended learning steps based on your parameters, just install the addon, open the stats window while holding shift and scroll down a bit.
As I understand it, short term and long term memory is very different, and FSRS is only made for long term memorization, so It is not very good at predicting intra day learning steps. IIRC you can enable FSRS to control learning steps by removing your learning steps, but again it’s not very good.
Your learning/relearning steps are what control how soon you’ll see a card after grading it Again. What are your steps?
But in addition to that, the other 9 cards that you already graded Again are queued up in front of that 10th card. (Well, generally in front, but fuzz factor will add some variability to that queue.)
No other cards are being prioritized in front of due Learn/Relearn cards, but if you feed Anki a waterfall of lapses, all it can do is feed them back to you.