Hi I have a quick question about FSRS and filtered decks. I have been doing my cards for almost a month now, but the card count (without any new cards added) remains in ~800 count. I have the filtered deck as is:due prop:due>-3, as well as rescheduling my cards based on my filter deck (image below). For my FSRS, I have the following image as my defaults, with a desired retention rate of 0.90. I do not know if I am doing something wrong with my filtered deck settings or FSRS? I thought I was expecting my due/review cards to be lower by now, but that is my main inquiry right now.
it was also 90 for desired retention, I had that setting with the “advanced” scheduling
about 1 year ago as well, before taking a break from Anki. I had used the initial experimental FSRS prior to anki version having built in FSRS, so I was not sure if the algorithm was working correctly or it was just me. I guess my question now is, is that to be expected with higher retention rates?
My expectation was that the card counts were going to gonna be going down much faster, like to 600s to 500s by now
Is there a reason you’re limiting this to only cards that are due today and overdue for the past few days? Do you have more cards that are more overdue that aren’t being studied? Or are you studying those cards in a different filtered deck? Maybe it would help if we could understand your workflow.
From your parameters, it looks like this material is very challenging. You’re starting with very short intervals, and they aren’t growing very fast. I suspect that’s why your workload isn’t dropping.
What is your retention rate – Stats > Answer Buttons graph > Mature %-age correct – for the past month? If that differs significantly from 90%, that can also be why your workload isn’t decreasing. It causes FSRS to keep your intervals short because you’re chasing that desired retention.
I have an Overdue filter deck with the parameters " is:due prop:due<=-3" but I have done all these cards already, meaning there are no longer an overdue cards. Below are my stats to help
Yes, your retention is a lot lower than your goal of 90% – especially for mature cards. [It doesn’t look like your workload is quite as high as you think though. Some of the charts show cards, some show “reps” on cards but both look lower than 800.]
This is a situation where you have to choose what you want –
Do you want to get your retention to 90%?
Do you want to reduce your workload and be satisfied with retention lower than 90%?
That’s not something anyone else can decide for you. Some folks would advise you to check “Compute minimum recommended retention” before lowering your retention. I’m not necessarily always in that camp, but it will at least let you know that you’re not going to be hurting your outcomes by lowering it.
Another thing to consider – how long ago did you last Optimize your parameters? Your retention has definitely improved over the past month. It might be worth it to run Evaluate on these parameters to check whether they are a good fit.
Retention or workload? Retention is the chart at the top labeled “True Retention.” That also shows how many cards you’re studying in a day, the past week, the past month. (FSRS chooses what to count/not-count in its own special way, but it wouldn’t be off by hundreds.)
Then there’s your answer. Leave your desired retention where it is. Your workload will start dropping when your retention increases. Until then, as far as FSRS is concerned, you need to be seeing these cards more often.
Once a month is plenty, so don’t worry about doing that again now.
FYI, retention here means retention when you review your cards. How much information you retain across all the cards will depend on a few other factors.
Your retention for mature cards can be lower than desired retention because of overdue cards. That’s one. But also because you do not rate your cards honestly or maybe because FSRS doesn’t have a lot of cards in high stability region.
Now, as for what desired retention you should have is hard to say until we get the simulator. But if you want lower workload, it might be better to simply use the desired retention value “Compute minimum reccomended retention” gives you. If CMRR gives you a desired retention that’s lower than 90 per cent, there will be a steady drop in “retention across all cards” (because cards are waiting for longer before coming up for review). But you might be able compensate for it by learning more new cards per day.
