I often start studying cards on the same day as I have made some of them. (I know this is maybe not ideal behavior, but i have a mock test coming up. i’ll use spaced repetition between now and the real exam)
The problem is, when I want to answer “good” on a new card, anki might suggest an interval of, e.g. 8 days. It doesn’t realise that I got it right because I literally just revised this topic (while making the cards). How can I have it so that anki assumes i have reviewed the cards recently, and that before this I had forgotten them? So far my work around has been to simply answer all cards as “again” and then start reviewing for real, or to use custom study. I don’t love my “answer all cards as again” method because then my stats think I’ve done 200 reviews in one day.
FSRS does take into account the prior grades you’ve given a card, and how long ago you gave them. I wonder if the issue here isn’t FSRS – it’s that you’re not confident you’re going to remember for as long as FSRS thinks you will. That’s a question of whether FSRS is doing a good job predicting your memory curve.
Have you run Evaluate on your parameters? What’s your RMSE? How many reviews did it count?
When was the last time you optimized your parameters?
What is your desired retention and how does that compare to your actual retention?
The only way I see is having a seperate deck/preset for the cards where creation date is close to first review date. Then FSRS can more accurately calculate the first interval for them. BUT if almost all your cards are like this then of course this will be of no help.
My issue is with new cards, with no prior review history in anki. However, in reality, when I create the cards I revise material I had mostly forgotten beforehand. Anki obviously doesn’t know this so doesn’t take this recent revision into account when deciding what interval it should give new cards when i answer “good”.
I think what I want is for anki to essentially think that the previous interval for a card is the interval between the time created and the current time
Just throwing in that the “time created” might be different from when you actually revised the information, e.g. if you import via csv or import a new anki package without actually revising the contents / facts beforehand.
Also: Does FSRS really struggle with that? Technically if you create a card now, revising info you have about that topic so that that card could be created in the first place, then you did study that card. It’s the same with studying info that might come up in your actual life. Sure, anki won’t know that you had additional revisions of your material (since they were outside of anki) but if you grade yourself based on the actual difficulty you had during this revision in anki when the card comes up, then everything should work fine. There’s no reason to believe that the memory traces won’t get strenghtened if you revise / encounter material in the wild (instead of in anki). Every revision counts, no matter if in anki or in the wild.
I also have cards that I created the same day I started to study them in anki and FSRS has no issue whatsoever in adjusting to my memory curve. All you have to do is be honest and stay consistent with your grading (and optimize your parameters every once in a while).
Of course anki can’t always know when you’ve reviewed info out in the wild, but it seems useful to consider card creation. The issue isn’t whether or not reviewing outside of anki strengthens your memory - of course it does. The problem would be that anki might think that you previously remembered this card for ages, when actually you only remember it thanks to short term memory from creating it earlier in the day or just a day before.
Someone on the github thread mentioned also that shared decks would typically be created a while ago.
The system could be set up such that if the interval suggested by how long ago the cards were created was longer than the interval suggested by the current system, anki would use the shortest interval of the two. This would ensure it doesn’t break things for imported decks users.
I’ll also add that in the case that the imported cards were created recently, this will simply make the user study them a little too soon. Surely this is better than studying them too late - in the same way that it’s better to have retention set a bit high as opposed to so low that forgetting material makes things less efficient. Furthermore, in this scenario, the user could simply answer “easy” instead of “good”, if they think interval is too short. This is much easier than users like me having to go through all of the cards they created and answering “again” on the day created, then answering "good.
I’m not an expert regarding FSRS but cannot really see why that would be the case.
Short term memory stores information from a few seconds to minutes. If you remember it after a longer period of time, then that means that you have some (weak) traces of that info in your long term memory. If you then grade yourself accurately anki / FSRS has no problem to adjust to your individual memory curve.
I would strongly advise against this. Users typically cannot accurately assess when they’ll forget things. Playing the algorithm, especially if the algorithm tries to learn about your individual memory curve, could lead to worse retention overall. Which is why some people here suggest that the intervall indicators above the buttons should be turned off by default. It’s best to accurately assess how difficult the card was for you and to consistently grade it accordingly.
Not neccessarily but potentially true. At least according to my experience (and probably the simulations too, looking at the results @Expertium shared in several places), FSRS should have no trouble to adjust – if you grade yourself correctly and consistently.