Hello, I was wondering why my hard button is 4 days and good is only 2 days. I thought that hard means that i should see the card sooner? Am i missing something? I would appreciate your help.
It’s hard to say anything without seeing your settings and card status.
Your card may be in the last step of learning. In this case, the interval of the “again” and “hard” buttons is controlled by your learning step settings. And the “good” and “easy” buttons depend on other settings..
It would be nice if they made it simpler. But, alas, it’s been anything but simple for a long time. The first one seems average, but the later ones have a completely different algorithm. And it also depends on this:
Is FSRS enabled? Which card, and where can I find its details to figure out what happened?
And what’s going on here?
I tried to replicate this on version:
Version 25.09.2 (3890e12c)
Python 3.13.5 Qt 6.9.1 Chromium 122
But I didn’t notice this behavior. Perhaps something else was actually happening, like changing the deck’s parameters from “1m 4d” to “1m 2d 4d” or something else.
What is your version?
Could you share the study history for the card in question?
It’s because of what you’re doing with your learning steps and intervals.
In your screenshot –
- The card is on the 4d learning step, so Hard repeats that step → Hard = 4d.
- Good graduates the card to Review, and you set a 4d Graduating interval, which is subject to Fuzz that will vary between cards → Good = 2d (fuzzed from 4d).
- Easy skips the rest of the steps and graduates the card to Review, and you set a 4d Easy interval, which is also subject to Fuzz → Easy = 3d (fuzzed from 4d).
First – rather than messing around with this, you should enable FSRS, and follow those instructions about your steps.
But if you want to use the default SM-2 algorithm, and tightly control your steps/intervals like this, you at least have to use it as it was intended – Deck Options - Anki Manual . If you have a 4d learning step, your Graduating interval should be more than 4d – you’ve already remembered it for 4d, so you don’t need to do test that again. And your Easy interval should be even more than that – because that’s for cards that can skip the rest of the learning steps and go straight to Review, because you know them so well.
Exactly when it appears:
Perhaps I should write in more detail here, and the “easy” parameter can also be changed in a way that won’t be clear to someone.
But still, turn on FSRS and let it think. Here’s an example of what might happen next:
You keep learning, but you’re constantly struggling. So what does such a simple algorithm do? It constantly increases the interval. But if you started forgetting as soon as the interval became 5 days, why won’t you forget when it becomes 7? In the isolated case of random forgetfulness, this might work, but more likely, you need to decrease the interval when things get difficult, and to do this, the user needs to understand what’s happening and where to set what. But alas, this is difficult for a beginner to understand.
For example, I set the minimum. Although I should have done this right away, not now. But the algorithm, not so smart, doesn’t take statistics into account, so I’ve had trouble with this card many times, and in fact, I have to go back to the 4-day interval and multiply it by 0.5 to get 2 days.
That’s the problem with this button, so it’s easier to have two buttons: I know or I don’t know. And sometimes, when in doubt, it’s easier to tell yourself you haven’t learned it than to deceive Anki and yourself.
It’s not worth it for this user to arbitrarily mess around with the SM-2 settings. If they want to understand how that algorithm really works, they should go to the primary sources to learn about it before they start changing random settings. But none of what you’re suggesting is related to what they are asking about – because their card was still in Learn.
I showed a screenshot of where it actually was when it was shown. It was still a tutorial for new cards.
And the algorithm isn’t perfect. What does “Hard” mean? Does it know half, less than half, or more? Yes, you’d say more than half. But ask users how they interpret that. If they read it somewhere, they’ll certainly say it’s something they barely remembered, but they remembered it all. But other users think “Hard” means they know at least half. Here’s what the help page says: “Hard: Select this button when your answer is correct, but you had doubts about it or it took a long time to recall.” What does “but you had doubts about it” mean? It’s balancing on the edge when there are two options, or zero or one. That’s the same 50%, and how can I say our users misunderstand? Perhaps we should remove those words about doubt from the help page altogether, and just say what took a long time to recall, but was absolutely certain (without a doubt).
So, what happens is this: the user forgets 50% of the information in X interval. Let’s say, after studying some more, they’ll forget 25% of the information in X. But they’re given an interval of X*1.2—meaning they’ll forget not 25%, but more likely 30%. And if they do, this button won’t help them the first time, and it won’t help them the second time. If, in reality, the user was only a little slow in recalling the information, but actually remembered 100%, then this algorithm is well-designed. So, it’s not the algorithms that are bad, it’s the users who are bad, they don’t need to figure this out… just do as I said?
None of this has anything to do with the user’s question.
I agree. I gave you a heart back then. But the user showed the wrong screenshot. I provided mine, after running a real test with their dates. I showed what can happen when the “Hard” button is used incorrectly. I suggested they use FSRS. I explain why this is the case, what needs to be improved, and perhaps even in the help section. I could do this for our team right now, but I’m not sure you agree with the inaccuracy of the text. Since I see your suggestions on GitHub, what they finish and change there will need to be looked at and clarified later. Thank you for your continued support!