Is there an add-on or a tool that can simulate how many future cards and review workload for a deck that someone will have using the FSRS algorithm instead of SM-2? The Anki Simulator uses SM-2 as far as I understand, does FSRS have something similar where you can input mature card correct rate, young card correct rate, etc.?
Does FSRS4Anki simulator fit your needs?
Thanks for the link. I was thinking more of something that had a correct mature rate involved, however I think that may not be possible with the FSRS Simulator, since it assumes that customized parameters take your mature correct percentage rate into account. I’m not entirely sure though.
Is the effectiveness of FSRS affected if you often encounter items of knowledge outside of Anki? I could ask the same question about SM-2, of course.
For example, if I’m studying a language, suppose that some word will be due tomorrow in Anki, but I’ve forgotten it, so when I review it, I will mark it Again. But in addition to using Anki, I’m also reading and watching videos in the target language, so let’s say I happen to encounter that particular word “in the real world” today, one day before it was due for an Anki review that it would have failed.
So I will remember the word temporarily, and when it comes up for review tomorrow, I will mark it Good instead of Again. Then FSRS will assume that I remembered it well and increase the interval, possibly to a very high amount, which is probably the wrong thing to do. Because I didn’t really retain it through the previous interval.
Language learning in particular is different, I think, than things like medical studies, where the chances are probably considerably lower that you will coincidentally encounter items of knowledge outside of Anki.
When you were researching FSRS, did this issue come up?
3 Likes
If anyone is using my edit of “deck and card info sidebar during review”, the difficulty it shows is apparently wrong: the code recently added to FSRS4Helper calculates it differently.
All spaced repetition algorithms are affected by that case. But immersion could also increase your memory stability of words that you encounter.
1 Like
Does anyone know if there is a way to calculate different card loads for a given retention? Ideally on a graph of some sort so that I can decide on what retention that I would like to use for my currently optimized FSRS parameters? I have attached an image of what I mean, is there a way to generate this graph for personal parameters?
2 Likes
I’d like to optimize my parameters since i use the default one for like 1 month. But I’ve been missing alot of reviews lately. Should i do it?
By the way, i’ve lots of suspended cards as well. Does it matter?
It doesn’t matter. If you have enough reviews to optimize the parameters, please just do it.
1 Like
I have Wednesday and Saturday as the Easy Days.
I reviewed all the cards in a big subdeck on the Anki Thursday, which included the OS Friday. I don’t remember if I ran Reschedule then, but I probably did (before the Anki day changed to Friday).
Next time I started Anki is now, OS Saturday, but Anki Friday. That subdeck had about 30 cards due (its burden is 145), which I reviewed, and rescheduling does not make any card in the subdeck due today, but only changes future days (depending on Easy Days).
This is after rescheduling with only Wednesday as an Easy Day.
Quick question: If I have burying siblings enabled, is “Auto disperse siblings reviewed on other devices after sync” working correctly? Or only non-buried cards are auto rescheduled?
It works correctly.
1 Like
It might be interesting to see Evaluator’s output for individual cards in the Browser.
Is it useful? For an individual cards, the randomness dominates the pattern of memory.
1 Like
To choose what to put into separate decks, I sorted cards by difficulty or ease, selected some near the top, and used “High Yield Tags” on them.
There is also Tag Statistics (but it uses the whole collection).
If true retention is lower than the desired retention, it is probably because of late reviews, but maybe it is also because of RMSE? Can optimal retention take RMSE into account?
It can’t.
Do you think that if I increase DR until TR reaches the calculated OR, that TR is likely to be more “optimal” than the low one?
I don’t think so. The optimal retention is based on the hypothesis that the model is accurate.
1 Like