I ask because some times i fail the cards on purpose because i think it doesnt affect the schedule of a card failing it again the same day. I ask because i see the answer get logged in the card info
I have a few use cases but the easiest to understand is with clozed notes without buring siblings
{c1} {c2} {c3}
I study ordered by creation date and so i get asked c2 then c3. I fail c2 and c3 so both go to learning
once in learning sometimes i get the cards mixed and get asked c3 (showing me c2 on the screen) and then c2. I purposely fail c2 because i literally just saw it on the screen
Yes. In recent versions of FSRS, every recorded grade affects the scheduling, even in Learn, and even multiple reviews per day – however closely-spaced same-day reviews have a smaller impact.
So if i understand this, i can fail a card 100 times in a row and as long as they are same day it will have small/no impact?
Sometimes I also fail them until i get them perfect because in my mind once i saw it, im already working with sort memory, which is different from recalling it from thin air in 2 days
No, I didn’t say it would have “small/no” impact – I said “smaller.”
As much as possible, you should avoid excessively grading cards, and you should expect those grades will impact the cards in the future. Put in the work when you miss a card so that you have a better chance of getting it right the next time.
I see, “you should avoid excessive grading” got me worried tbh
what does it mean to put the work in this context? for me doing the flash cards is the work. I guess i can take the work out of anki and write them in paper so i can get my next anki grade of the card in learning fine
What should i do when you know the answer because i literally just saw it on the screen? i usually just fail it even tho i likely know it, so i get asked again. This happens to me every day at least 1-3 times
Theres also times that get mixed in learning a few similar cards that makes me fail them a few times but i would get them out of learning first try if they were alone in there. I can manipulate this a little by separating new cards from the study of the deck or stopping studying for 10 minutes so theres only 1 card in learning at a time
This algorithm unintentionally mixes short term memory and long term memory and in my experience these are not related. Short term is recalling something you recently saw and long term is recalling something out of thin air on cue
How long does it take to fix something in your brain is not related with how well you remembering it next week
You should aim to have a perfect recall using short term memory (aka once you saw the answer) and do as many tries as you need to achieve this. if you cant do a perfect recall, the card is too hard and should be simplified. This is why i argued last week in another thread that hard and easy buttons should probably removed from learning. A card cant be hard using short term memory. If its hard the card should be redone or you should keep studiying it. A card cant be easy using short term memory. You just saw it, of course you can repeat it very fast
These are not the same hard and easy you use when you are using long term memory on the card
Im tying to understand how much impact does it have in the algorithm so i can minimize it because this is new to me
I don’t understand why you are deliberately failing cards that you know. Anytime you give Anki a false answer, it seems that that defeats the algorithm and gives bad results.
(Answering “Hard” for forgotten cards is known to cause problems. So answering “Again” for known cards will probably also cause problems.)
From your original post, it seems that you have disabled burying siblings. Is there a specific reason for that? The cards you are deliberately failing are siblings of other cards you reviewed on the same day?
Im sorry i cant read the math. could you explain in layman terms
Getting something in your short term memory is inherently messy and should be removed from the long term algorithm because i think the process are not related at all
Even if they are related (they are the same info at the end of the day), they are two different separated process, what you do to commit something to short memory should not affect the long term algorithm even if a few cards happens to do well on both process
It makes specially little sense to let go of a card that you are learning by pressing hard until you do a perfect recall in short memory
(unless you are using hard as this is important info, and easy as i should know this already regardless of the reality of how well you know it)
The fail, hard, good and easy buttons are not the same and have different meanings in short vs long term
as an example, easy in long term means i recalled something on cue out of thin air and i cant remember any other soft recall that might have affected it. A soft recall would be for example, i used this info for something yesterday and i can clearly remember this usage
In sort term it makes no sense to say something is easy because you literally had a soft recall 10 min ago (with defaults values)
You are rating long term only the first time, after you fail, you are no longer rating long term memory because memory evaluation is always subdued to soft recalls, which literally just happened
Im not gonna argue your statistics, i cant understand or know how to run whats on your repository, but I still have 2 concerns
what should i do with a card in learning that i literally just saw the answer to on screen in the previous card
Cards in learning can be manipulated to have lots of fails or none at all. If they affect the algorithm, can you please help me understand how so i can minimize damage?
You are rating long term only the first time, after you fail, you are no longer rating long term memory because memory evaluation is always subdued to soft recalls, which literally just happened
You do realize that the formulae for Short-term and Long-term are different The model already takes into account that a review occurs in short-term or in long-term. That is the essence of FSRS 5 and 6.
What even are you arguing here
Cards in learning can be manipulated to have lots of fails or none at all. If they affect the algorithm, can you please help me understand how so i can minimize damage?
Well if you are just manipulating the cards by inflating the fails or not failing at all then you are not really using FSRS as it should be. You are using Anki more like an organizatory tool, less so a learning tool while you are doing your own thing.
what should i do with a card in learning that i literally just saw the answer to on screen in the previous card
If they are sibling cards, you can try to switch on the option of seperating them apart.
Whatchu, I recommend that you manually bury the cards as you review them rather than incorrectly marking them as failed.
The best solution would be to use your tools the way they were intended to be used (especially by ensuring related cards are siblings and enabling the automatic sibling burying), but manually burying cards that contain duplicate information may be a nice middle ground to solve your immediate problem.
So you saying to bury cards in learning? to leave them in learning until next day? Next day a random card in learning is not going to adhere to shorting rules, so you are just going to delay the problem
And if you mean before, some times you can bury when it makes sense. I do bury cards that where exposed to me by mistake or bad luck, but not with things that need remembering in order, like a poems or lists
Yes, I do mean to bury cards while learning. That’s what I meant when I said “bury the cards as you review them.” When you say that doing so will delay the problem, you’re absolutely right. It will delay the problem you created for yourself by not using Anki the way it was intended to be used.
As for “things that need remembering in order” like poems or lists, I recommend the LPCG (Lyrics/Poetry Cloze Generator) plugin. It’s specifically designed to help with such things since Anki reviews will always be in the order calculated by the scheduling algorithm, not the order you made the cards in. Anki doesn’t have any feature where “The cue for the next card is the previous card.”
As a last note, I genuinely appreciate your engagement with the Anki community! I hope you continue to engage for as long as you use Anki. When you interact on the forums, please keep in mind that none of us are paid to help you, and very few of us are developers who work on Anki. If we answer one of your questions, it’s very discouraging for us when you don’t take the time to understand our replies or read the existing documentation. It’s especially frustrating when you argue with us unconstructively. I’m always happy to help (when I have free time), but this kind of response makes me want to use my free time for other things.
The model already takes into account that a review occurs in short-term or in long-term. That is the essence of FSRS 5 and 6.