In this lesson, you’re going to learn my recommended Anki settings and how settings work in a more practical manner.
That’s because I don’t only want you to have “ready-made” templates, but also a complete understanding of how the advice I’m giving you works.
Why? Because I want you to be able to think for yourself rather than use something you don’t even know the purpose of.
Just to be clear, Anki’s default settings do work.
However, in my experience, I found the default learning steps (more on this later) gives a really poor review experience for both newer and older cards.
Newer cards don’t get reviewed as frequently, and older cards start back to zero when you’ve lapsed even partially. (i.e. during tip-of-the-tongue moments)
Once you use these settings I’m going to give you, you’ll be able to avoid these two problems right away so you’ll have a smooth review experience every single time.
They’re made-for-you settings that I found the most optimal in my learning process. Again, I will also share with you the underlying concepts (and provided supplementary materials) so you can tweak them on your own.
That being said, I do have to note two things:
- My recommended settings are more geared for conceptual learning than for memorizing vocabulary or isolated facts
- Changing your settings this way won’t really make or break your overall retention for conceptual learning. It just enhances the review experience.
Which means if you’re constantly forgetting the majority of what you’ve studied even though you’ve already made Anki cards for them, then mere settings are NOT going to treat that problem.
Luckily, I have a solution for that — you’ll learn it at the end of this post.
The Made-For-You Settings
This is an updated version that uses the recommended settings in the “LeanAnki Method” part of the study system course.
For Daily Limits:
New cards/day = 9999. New information lacks depth, and should therefore be studied immediately. So you need to retrieve them in order to make them stronger — at least, before they fade into oblivion when more complex concepts are still far ahead.
Maximum Reviews = 9999. Retention of past knowledge makes future learning easier. But take note that this can only be true if you have encoded what you’ve learned before making flashcards. Otherwise, you only get fragmented, useless retention that cannot be used as an activated semantic context.
For the New Cards, I recommend these settings:
The Learning steps setting means that pressing Again shows the card in 10 minutes and pressing Good would show the card after 1 day, then after you press “Good” on that, you’ll see it again after 3 days.
So the first number is actually the number of minutes the card will show up again if you pressed “Again”.
The next two numbers determine the intervals when you press “Good”.
And like the tab says, all of these only apply to New Cards.
Now, I don’t recommend you pressing “Easy” right away because that tends to skip the “learning steps.”
You don’t want that. Just because a card is easy to remember now doesn’t mean it’s going to be remembered in the future — we have a bias to believe otherwise (cf. Stability Bias) and that’s the trap we will fall into if we don’t “filter the bad cards out.”
So, just to be on the safe side, go through all of your learning steps for your new cards by pressing “Good.” (Obviously, this doesn’t apply to graduated cards — more on that in a sec.)
Now how do you know if the card has surpassed the “learning steps”?
You’ll know it when the next interval in the “Graduating Interval” appears. (in this case, it’s 7 days)
When you see that option, it means that that specific card will graduate after you press “good” or “easy”. (Think of the steps as a “grade” in grade school — you have to go through all of them, right?)
When a card has graduated, the ease of retrieval (i.e. what buttons you press after answering that card) will then determine how often that card will show up in the future, rather than the learning steps.
Lastly, you might be wondering why the “Insertion order” is set to Random.
What this does is make the questions show up in an interleaved way. Interleaved practice helps you dissolve boundaries between subjects, and this is also why I recommend letting relevance determine deck creation.1
So going back to graduated cards…
When the card has graduated, it will use the “Lapses” and “Advanced” settings:
The Lapses settings, I don’t care so much about that. 10 minutes is already fine for avoiding “mindless recall” when reviewing lapsed cards, and I don’t want my poorly-encoded cards disappearing, so I recommend not changing the “Leech Action” to “Suspend Card.” (This usually happens when your cards aren’t encoded or future-proof)
Then here’s the “Advanced” settings.
I primarily used Anki when I reviewed for my board exams, and I noticed that sometimes there were cards that are supposed to be extremely easy but have very low intervals.
If you find that you have those cards, this means that you have forgotten them somehow after they have graduated, and that they have already used the default “New Interval”, which is 0%.
So I’ve set the “New Interval” setting to 0.60. This is crucial, because if you forgot something you’ve been successfully recalling for 3 years, do you really have to test yourself as if it’s a new card?
CLEARLY NOT. And it’s an inefficient use of time and energy for a well-learned card.
That’s why we concede by turning the new interval to 60%. It’s arbitrary, of course, but I wouldn’t recommend going below 50%.
This means that if you have a graduated card, and you’ve pressed “Again,” then the next interval will be set at 60% of the original. Meaning, if you’re supposed to see that card 100 days for a “Good” rating, then pressing “Again” will make it pop up after 60 days (0.6×100=60) instead of 100.
My rationale here, if it’s not clear yet, is that well-learned cards do not need to be reviewed the same way as a newly learned card.
If the depth of encoding2 is the same, then there is already a base of storage strength for the lapsed card. Since the large storage strength slows down the rate of forgetting, it means that it will be forgotten way slower than a new card — which generally has low storage strength3.
Regarding your reviews with graduated cards, you should press whatever option in the “Easy” “Good” “Hard” suits your judgement.
That’s because from that point on, you want the Anki algorithm to do its magic for you.
- The New Cards settings apply to, well, new cards that haven’t yet graduated
- Press “Good” only for the newer cards to go through all your learning steps and avoid Stability Bias
- A specific card has “graduated” when you finish the learning steps
- Press whatever review difficulty option (i.e. Easy, Good, or Hard) for graduated cards
Settings Aren’t an Exact Science
If you’ve noticed already, these intervals are incredibly simple to tweak.
That’s because they aren’t exact science.
If your goal is to just use Anki more effectively and more efficiently in a practical way…
Then do you really have time to overthink about the exact data — i.e. “the pErFeCt aLgOriTHm” — instead of actually spending that time creating better cards that will improve your knowledge base? Learning how to extract important information from a textbook?
Look, even if some garbage news site “knew” what the “most effective & efficient Anki setting” is…
It HIGHLY likely that it fails to take into account:
- The context of what you’re learning
- How specific your card is
- How future-proof it is
- How well they relate to your previous knowledge
- …and so on.
…simply because these can’t be quantified.
In short, even if there’s available exact science out there, it likely fails to account the big picture.
So I argue it’s better to keep this simple.
One caveat, though, is when you’re not using Anki as a tool for learning concepts, but rather just to memorize facts.
I can’t speak for that — perhaps the most efficient intervals do exist for it.
(But hey, I doubt you’re even gonna make it this far if you’re not gonna use Anki for learning!)
Anki designers were actually criticized for not having “optimal intervals”, but in defense of Anki, Nielsen argued:
I’ve heard this used as a criticism of the designers of systems such as Anki, that they make too many ad hoc guesses, not backed by a systematic scientific understanding.
But what are they supposed to do? Wait 50 or 100 years, until those answers are in?
Give up design, and become memory scientists for the next 30 years, so they can give properly “scientific” answers to all the questions they need answered in the design of their systems?
We get that exact science is good for increasing the chance success, but every action doesn’t have to wait for exact data, especially if you want to make progress immediately.
Instead of waiting for memory scientists to somehow find the data for us, we follow the first principle of spaced repetition and learn the nuances for ourselves.
That is, to set increasing intervals for reviews.
Simple, and practical.
One might ask, “What if I’m overtesting/undertesting?”
You’d often hear this from other “advanced beginners”, but frankly, it’s impractical to stress yourself over them.
In case you’re asking the same question, here’s the solution:
Adjust the settings accordingly.
Of course, it’s easy to say that the overall review time can be sped up when you use setting X or Y…
But frankly, there are more important things to address in the grand scheme of things.
Next Step: How to Use Anki Efficiently
Like I said, settings alone won’t make or break your ability to retain most of the concepts you’re learning.
Sure, it might help a bit — but the gains would be marginal at best.
It’s more of an optimization thing rather than a viability factor.
And, as I’ve said in the beginning of this Anki Fundamentals series…
Anki — or more appropriately, spaced repetition — is just one-third of the things you need for learning effectively.
Listen, if you’re constantly tweaking your settings and you’re not seeing any substantial improvements, then the problem is probably not with Anki.
If you’re interested to know more, then the next course — How to Use Anki Efficiently — is for you.
It’s totally free, no signups required to start.
- See this article for more.
- Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(6), 671–684. https://doi.org/10.1016/S0022-5371(72)80001-X
- Bjork, R. A., & Bjork, E. L. (1992). A new theory of disuse and an old theory of stimulus fluctuation. In Essays in honor of William K. Estes, Vol. 1: From learning theory to connectionist theory; Vol. 2: From learning processes to cognitive processes. (pp. 35–67). Lawrence Erlbaum Associates, Inc.