So I have been doing FSRS for around 1 week and I have an exam in 25 days. I have around 500 cards and I have around 200 cards that I need to learn. The rest of of the 300(ish) cards I have learned or I have recalled with ease, or with some problem.
I have been only pressing 3 of the 4 buttons. I have pressed “Again” if I completely have no idea what the answer for my card is, “Good” if I recall the card half-way or I recall it with a lot of trouble, and “Easy” if I recalled the card with ease and no problem at all. I have not pressed “Hard” once as I read that it would be bad for FSRS’s algorithm to work.
I know FSRS is a card rescheduler and sets when the card will come up to best work with my desired retention. I see many people with lots of cards on “due” rather than lots of cards on “learn”. I am worried I am messing up FSRS by having cards that are on “learn”. I’m not asking about the intervals with FSRS as I will just let the cards reschedule via FSRS and see how that takes does for me. I am just asking if I don’t have a lot of cards on due, I will have less cards affected by FSRS thus lowering the effectiveness of it or am I just over thinking it? If you guys have any suggestions for what I should do about all my cards on “learn” can you please tell me?
Thanks!
Visual on what I am talking about: (I don’t have any on “due” as I did my revisions already for today)
TLDR; I have lots of cards on “learn” and not a lot on “due”. Is FSRS going to be less effective and are there ways I can get more cards on “due”?
Initially, the cards will not be scheduled by FSRS, because it is not scheduling learning and relearning cards. But after that initial go-through it will start working.
The only thing that is an issue between Hard and FSRS is if you are grading answers Hard when you really should have graded them as Again. So, don’t do that. Otherwise, clicking Hard won’t break anything.
Cards in learning are also affected by FSRS, just not as immediately, because they are still controlled by the learning steps. That history of grading your cards (the 1st grade each day, anyway), will build up to FSRS determining how Difficult the card is, and how Stable your memory is for it. Grading your cards accurately and honestly is the best plan – regardless of which algorithm you are using for scheduling.
This might be somewhat of an issue, because if you want to optimize your settings Deck Options - Anki Manual, FSRS wants you to have at least 1000 reviews. But if you’ve been keeping your cards in learning, you might have stocked up enough reviews already. Regardless, you can start with the default parameters and optimize them to your own history when you’re ready.
Lastly – what’s up with that? Do you have a lot of added learning steps? Or are you just “Again”-ing your cards so that they never graduate from Learning? You’ll probably be better served by letting those cards graduate to review, so that you and FSRS can make the best use out of them.
I only have around 30-50 reviews per day based on my deck size. Like I said, I have around 500 cards in total, around 200 in learn, and the rest are in the due. Since I have less than 1000 reviews should I go back to the default scheduling or should I just stick with FSRS.
You are exactly right. What should I do instead of “Again”-ing my cards a bunch? Should I just press hard? But my “Hard” value is also less than a day.
EDIT: Also, are my intervals fine for a brand new card? I have my “Learning Steps” set to 15m.
Since you’re already running FSRS (with the default parameters, I assume?), you should stick with it.
Those intervals look fine. If you trust the algorithm (which you can!), try not to think too much about the intervals displayed. [In fact, you can turn that of in Preferences for a while if it is a distraction. Preferences - Anki Manual] The important thing is that you are clicking the right grade for your answer.
As sprvlcn said – Again is always and only – I got the answer wrong. But everyone draws the lines between Hard-Good-Easy in their own ways. When I started, I found this helpful to think about the grades in relation to each other Studying - Anki Manual .
Some of the specifics don’t apply when you’re using FSRS, but the general principle is the same – was the amount of delay between the last time and this time right or not?
It was too long of a delay, because it was a struggle to answer = Hard
It was about the right delay, because I thought about it and got the answer = Good
It was too short of a delay, because I didn’t even have to think about it = Easy
Most cards you get right should be “Good.”
I have a related situation where my ‘Learn’ cards are starting to outnumber my ‘Due’ cards, but not as extreme as the OPs.
In my case I believe it’s due to me making heavy use of the Hard button and being strict with graduating via Easy. I’ve made my conditions for using Again, Hard, Good, Easy pretty clear and concrete, but after this recent change I noticed the similar phenomenon.
I’m guessing that this isn’t an issue with FSRS, as it was mentioned above that after a while FSRS schedules cards in learning too. Also, I do have cards moving to review, just that they trickle in slower due to the stricter conditions they have to pass.
So it seems my situation is similar yest slightly different than OPs. Will FSRS handle it fine, or should I consider my recent change in grading criteriea?
Personally I have never used hard a lot, rather using Easy when I just don’t know the information, the Good button when I know it but it takes me some time to remember and its not quite as fast as I like, and Easy when I know the information AS SOON as I see the card. I only use hard when I see a card and it takes me forever to recall the information but I still manage to recall it.
I used to only have around 30-50 reviews pre day but now I am getting 80-100+ reviews per day (which may not seem like alot but my deck is only 654 cards). I have been mainly using only again and good as my main two buttons and I think good is the sweet spot for the buttons. I have more cards that take me a bit of thinking before answering so thats why I have more cards due sooner.
My suggestion is just to not use the hard button as again and to use again more than hard. You should also try to see if you can make the cards smaller or simpler so you can have the same amount of information but also being able to memorize it too. I had a problem where I added a bunch of information to my cards and I couldn’t recall it all so I just kept pressing Again. But since then, I made my cards easier and I have been pressing good a lot more frequently. You can check this website out when you make your cards. Its called " Effective learning: Twenty rules of formulating knowledge" by Supermemo . com. I just think its how you are grading yourself on the cards. But with that being said, I think you should take Danika_Dakika’s advice and do what they said for grading your cards.
But I do have a question for Danika_Dakika too lol. I just added a bunch of new cards to my deck and my deck is a lot bigger now. I have 654 cards now and I use anki around 1 hour to 1 hour 30 minutes per day spaced throughout my day. I used the “Compute Optimal Retention” and I put 654 as the deck size, days to simulate to 18 (as my exam is 18 days away now), and minutes study/day to 90 and it said my Estimated Retention was 0.95. How is this value decided? What I mean is that how do I get that Estimated Retention? Do I have to study my whole deck everyday and then I get my Estimated Retention or is that Estimated Retention an estimate on how well I can recall my due cards?
Also is it normal for the amount of due cards to be less and less as my exam day gets closer? I used to have 150+ reviews but now I’m getting 80-100 reviews per day. Is it because of FSRS’s algorithm scheduling the cards?
I use hard because my cards are quick physical tasks where I can objectively grade out of 10 attempts how many I get correct.
0/10 successful attempts → again because it was a clear failure
10/10 successful attempts → good
10/10 with a clear feeling that it wasn’t challenging → easy
Anything in between means I can somehow manage the task, but I should definitely practice more often to get it down before graduating, so I grade those as Hard to repeat them in the learning phase.
Once cards are in review, I actually relax my criteria, and grade them good if I achieve the task with 80%(8/10 attempts). My FSRS desired retention is set to 0.87 which is somewhat between the success rate of learning(10/10) and review(8/10) criteria.
I don’t really agree with the idea of completely avoiding the hard button since I need to repeat cards while learning them. This to me is not the same as hitting again which classifies the card as ‘re-learning’ if I understood correctly.
To summarize, my cards are unique in that they can objectively be graded. However, being strict with that grading is leading to a similar pattern of many cards building up in learning and only slowly trickling in to review, so I don’t know if I should relax my grading criteria if it interferes with FSRS(let the cards into review after 90% successful trials during answering, especially since that is my target desired retention rate)
So, the first thing I see is that these graphs only cover about … 175 reviews. That suggests a few things to me. You’re new to Anki (and maybe spaced-repetition systems in general), and you don’t trust that it’s going to work. You also might be afraid of graduating cards because you’re afraid they will lapse and the world will end.
Just in case that accurately describes you (no judgment, and you don’t have to admit anything), I’ll offer this –
Don’t worry about whether cards will lapse – because you will definitely have cards that lapse. Lapses are a part of learning, and they have never caused any world to end. They are not proof that SRS doesn’t work, they are evidence that it is working. SRS is teaching you to use your memory to stretch for things. Every time I stretch for an answer, it helps, with that answer and with training my memory – whether I get to the answer or not [anecdote, not from a brain scientist].
Right now, since you don’t trust the algorithm, you’re actively preventing it from doing what it knows how to do best. So, it will never be able to show you what it can do, and you’ll never trust it, and [… repeat]. It’s okay to trust the scheduling algorithm and let it demonstrate what it can do. [This is assuming you have time for that and you aren’t expecting some sort of my-test-is-in-3-weeks miracle.]
You have to start letting cards graduate from learning. Otherwise, you might as well just put this deck on index cards and flip through all of them every day.
Don’t think of FSRS as something that takes over “after a while.” FSRS can’t control the schedule of a card while it is in learning/relearning steps, but it is definitely paying attention to the card. It clocks how you rate that card (once) each day, and all of that factors into how it schedules that card later on. FSRS expects all of your learning steps happen in one day, which ends in graduation and the algorithm taking over.
The most important thing you can do to help FSRS be successful for you is to grade your cards honestly and accurately. You can set whatever strict rules you want for yourself, as long as they result in cards graduating from learning!
I do trust the algorithm, I just want to work together with in a way that lets it shine. That being said, I now have a better idea of the disadvantages of leaving cards in learning instead of graduating them(they are accounted for but not actively managed from what I understood).
Based on your feedback I will relax my criteria to let more graduate. I accept there will always be lapses, but I still want to keep those to a minimum. Nonetheless, if I hoard my cards in learning it prevents the algorithm from doing what it does best which I don’t want to stand in the way of.
Due cards. Estimated Retention will be based on you studying your cards as FSRS schedules them.
At 18d left, are you still adding new cards to your studying? When you stop adding New, that will always be the point when your review spacing starts to stretch out and your workload goes down. It’s not strictly due to FSRS, it’s just how SRS works in general. Anki Simulator is still a fun add-on to use, even though it makes its projections using teh SM-2 algorithm instead of FSRS. You can run numbers and see that every time you stop adding New, the load will start to recede.
Yeah I’m a bit upset about that because my teacher is behind on the material and I don’t understand some of it so I didn’t add it to anki until very recently. I think I might still add a couple of cards in the next 2 or 3 days but thats about it. I know anki wont work at its best in only 18 days but I think I can do okay for that section on my exam (if I try hard enough to memorize it). Do you think that If I still continue to add cards 2-3 days from now I will be able to retain my information? Or do you have any methods to memorize the subdeck in those 18 days?(around 150 cards give or take). Thank you so much once again.
If you want to know some context on why I’m still adding cards here it is: I have a main deck named “Science” and I have 4 subdecks in it. I have already finished making 3 of the 4 subdecks and actively reviewing them when I first originally made this post around 8 ago. But now my teacher is almost done explaining my 4th subdeck and I am just adding cards based on their explanations in class. I can do good on the 3 subdecks I but this 4th one that I’m still adding too might be a problem.
You can definitely do extra emphasis study on those new-er cards. For instance, setup a Filtered Deck Filtered Decks - Anki Manual to study after you do your main study each day – and base it on characteristics of those cards Searching - Anki Manual.
[This is just an example, you’ll use your own numbers …] Start with your review cards is:review -is:learn, narrow that to cards seen for the first time in the past 7d introduced:7, exclude cards already reviewed in the past 3d (today, yesterday, the day before) -rated:3, and ones already scheduled for the next 2d (since you’ll study those soon enough) prop:due>2, etc. Or instead you could focus on the cards you’ve seen the least prop:reps<5, or you’ve failed the most prop:lapses>3. And of course, you can adjust those figures as necessary as your timeline gets shorter.
