Better Anki Usage Guide
As you may or may not know, I love Anki, the spaced repetition flash cards app. I’ve been using it in my Korean learning journey for almost one and a half years now and I’ve written about it before. It’s a good app, but you can have a much better experience by making some modifications to it.
If you don’t know what Anki is or why it’s compelling, I recommend you read Augmenting Long-term Memory, a blog post by Michael Nielsen. Everything I know: Tips and Tricks for Anki by Jeff Shek also has a beginner’s guide.
I recommend using Anki for at least 2-3 weeks before reading the rest of this article.
1. Changing default settings
Anki out of the box has two problematic defaults, and one default that I don’t recommend personally.
1.1. Resetting card progress
Let’s say you have learned a card, but when you review it you make a mistake in recalling it, for whatever reason. Call it a brain fart, or pressing too quickly, etc. Or you simply forgot!
By default, Anki resets your progress to that of a new card. So even if you had worked the card up to a ~4 month interval, now you have to see it in 1 day, 2 days, 4 days, all over again…
Even if you forget a card, you still had learned it at one point. Which
means re-learning it is now significantly easier. By completely
resetting the card’s intervals, Anki is wasting your time. It slowly
conditions you to be unwilling to press
Again, which is not desireable
Solution: In your deck’s options group under Lapses, change
New interval to
70%. This means if you forget a card, the new
interval will be 70% of the old one. As for why I chose 70%, it just
felt like an appropriate amount–feel free to play with different
percentages to see what works best for you.
1.2. Taking too long to mark Leeches
After you initially learn a card (the card reached a 1 day interval in default anki), if you forget it 8 times, Anki marks it as a Leech and suspends the card, not showing it to you anymore.
A card being a leech generally means its too complicated, or you’re getting it confused with a similar one. 8 times means it’s going to take way too long to identify cards that are problematic.
Solution: In your deck’s options group under Lapses, change
Leech threshold to
4. You may also want to change
Leech action to
tag only. This way you can easily sort for leech cards and modify or
As for handling leech cards, you generally want to simplify the card, or delete it outright. With the time and energy that you wasted on trying to recall a leech you could have learned a bunch of other cards. If it’s a word that you don’t want to delete, you can suspend the card, and practice using that word outside of anki. When you feel more comfortable with it you can unsuspend it.
1.3. Mixing new and review cards
By default, Anki mixes new and review cards. However, reviewing is vastly more important than learning new cards–it’s okay if you don’t memorize new cards every day, but missing reviews really hurts your memorization progress.
By putting reviews first, you can always do reviews daily and then choose to learn new words or not depending on how busy you are, how tired you are, etc. If the cards are mixed you can’t do this, making it likely you’ll just skip Anki for that day entirely, which can can quickly snowball into not using anki at all anymore. This happened to me.
Solution: In Anki’s preferences, change
Mix new cards and reviews to
Show new cards after reviews.
1.4. Making better card templates
The Basic card template in Anki has two fields:
This is the minimum for flashcards obviously, but as you make cards it’s
likely you’ll want to have separate information.
Over time my personal Zah Korean card template has grown to eight
sound. On the front I show
a hint. Some people put sample sentences but I find this leads to
short-circuiting your recall; instead of recalling the word you
recognize that sentence.
I also have a Zah Colorcard template which has a
Because I initially tried to learn colors in Korean like regular
vocabulary, which makes no sense. Anki lets you pull in values easily in
the template’s html and css sections.
You can modify a card’s fields by going to Add or Edit, then clicking on
Fields…. You can add a new card type by going to Add or Edit, then
click on the type, and then click
1.5. Better styling with html and css
Default anki’s styling is very plain. I recommend modifying it to show
different fields in different colors. For example, my
field is rendered in green while
hanja is rendered in purple. I render
my color card’s hex value dynamically. Here’s what my cards look like;
the horizontal line denotes the front and back:
You can go to Browse, Add, or Edit and click on
Cards… to get to the
style dialogue. Some premade decks include nice styling for you, or
default to a night mode. For more in-depth info, you can visit
styling in the anki manual.
2. Why behavior-changing plugins are needed
Essentially, the way Anki treats cards is not intuitive, and actually
hampers your memorization progress. Let’s say you’re reviewing a
card–you have four options.
Again, if you forget, and
You would assume this means "I (the user) had a hard/good/easy time recalling this card today", right? Well, Anki actually sees it as "this card is intrinsically more difficult; please modify it permanently".
You see, every card in Anki has an individual
starting at 250% by default. The algorithm that calculates a new
interval goes like:
NewInterval = OldInterval × CardEase × DeckIntervalModifier Where: OldInterval: 1 day is the first default value CardEase: 250% by default DeckIntervalModifier: 1.0 by default, set on a per-deck basis
When you hit
Easy, Anki modifies the ease by
+15%, respectively… You can see now why people recommend not hitting
Easy at all; it quickly snowballs.
Note that a card’s
ease can’t go lower than 130%.
So you might be saying, why is seeing cards too often, aka "overlearning", an issue? Either way, I’m seeing more difficult cards more often.. Well, research suggests that after an initial period, the extra time spent doesn’t actually lead to better recall.
In one study (Rohrer, Taylor, Pashler, Wixted, & Cepeda, 2005), college students learned novel vocabulary (e.g., cicatrix–scar), cycling through a list of word–definition pairs either 5 or 10 times. The extra 5 cycles yielded a substantial benefit after 1 week, but the gain was no longer apparent after 4 weeks … From a long-term perspective, overlearning appears to be inefficient almost to the point of wasting time. 
Now take a look at what happens in Anki if you lower the ease too much:
So even if you know a card enough to hit
Good, if you had hit
3-4 times prior, the interval growth rate is permanently stunted due to
how Anki works. You would have to hit
Easy to balance it which is
Research also suggests that larger spacing is better in the long term.
In a 9-year longitudinal investigation, 4 subjects learned and relearned 300 English-foreign language word pairs. Either 13 or 26 relearning sessions were administered at intervals of 14, 28, or 56 days. Retention was tested for 1, 2, 3, or 5 years after training terminated. The longer intersession intervals slowed down acquisition slightly, but this disadvantage during training was offset hy substantially higher retention. Thirteen retraining sessions spaced at 56 days yielded retention comparable to 26 sessions spaced at 14 days. 
Our results can be summarized as follows. We find that over substantial time periods, spacing has powerful (and typically nonmonotonic) effects on retention, with optimal memory occurring when spacing is some modest fraction of the final retention interval (perhaps about 10%–20%). 
With that out of the way, here are the plugins I use.
3. Anki Subdecks
If you review from multiple decks daily, you can use a parent deck with subdecks to mix reviews. This is my own personal opinion, but I think it’s better for recall if you change contexts. This is more like recall in real life.
Making subdecks in anki is really easy–just make a parent deck, and drag
decks onto it. Or you can manually rename the deck in this format:
Parent Deck Name::Subdeck Name. If done right, it should appear like
The only issue is Anki will still do reviews one deck at a time. I tried
using the experimental V2 scheduler, but it didn’t work for me. So now
we have to turn to a plugin:
HoochieMama: Randomize Rev Queue.
After you install this, open Anki’s preferences, and in the
Muffins tab enable
Hoochie Mama! RandRevQ w/ subdeck limit.
That’s it! Now you can review from multiple subdecks, and new cards are still deck by deck. If you want to mix new cards as well, there’s Hoochie Papa. I personally don’t think that is as useful.
|Make sure to back up your anki decks before adding behavior modification plugins like this.|
4. Filtering Cards
There are times when you want to study something more specific than just cards in a deck.
For example, lets say you have a general language deck with some cards tagged as
and you want to only study those.
4.1. Custom Study Sessions
If you click on a deck, you should see a "Custom Study" button. This gives you several options, which you can read about here.
Increase today’s new card limit
Increase today’s review card limit
Review forgotten cards
Preview new cards
Study by card state or tag
Choose the last one, then "All cards in random order (don’t reschedule)", unless you only want to study cards that you’ve already seen. Then click Choose Tags, and the tags that you want to focus on.
Once you review the card, it will go back to the main deck. You can also delete the Custom Study Session deck.
The downside here is that the queries are per deck. What if you want to search multiple decks, or have a more complicated query?
4.2. Manually Filtering
The other option Anki has is to create a Filtered Deck, using Tools > Create Filtered Deck from the main menu.
This will give you a search prompt. Anki searches can get pretty advanced, so read about them here. Some useful ones:
deck:decknamefilters by deck.
-negates the filter.
tag:tagnamefilters by tag.
-deck:filteredfilters by normal (unfiltered) decks only.
You can group terms with parentheses:
(tag:tag1 or tag:tag2 and tag:tag3)
5. Visual Plugins
These aren’t really necessary but I like seeing stats. Give me allll the stats.
This is basically a fancy stats plugin. You can see a detailed breakdown by day, week, and month, and see your true retention rate. You want roughly 80-90% retention. [TODO]
This addon shows extra information of the due cards and returns the expected time to finalize (Due+New).
This adds more stats to the main window via columns. As you can see in that link, the default config adds quite a lot, so I modified my config to pare it down, which looks like this:
Note that the stats at the bottom are from the
More decks stats and time left plugin.
Adds two new graphs to the stats window.
Simply colorizes Again to be red, Good to be green, etc. It’s a small but nice thing to have when reviewing. There’s also Large and Colorful Buttons if you wanted even more colorful buttons.
This is an awesome plugin that visually shows your kanji learning
progress. Simply select a deck that has a
kanji field, and run Tools >
Generate Kanji Grid. It looks like this:
This was generated from my deck with the default settings.
6. Behavior Plugins
This is where things get fun. These plugins change the core functionality of Anki, so please, read about them and make sure you understand what they do before adding them.
|Make sure to back up your Anki decks before installing these.|
This adds a pane to the
Add Card window, letting you quickly search
through your Anki decks for keywords. It’s very useful for finding
duplicates, or searching among sample sentences.
The only con is the search doesn’t work for non-ascii text. When I have time I plan on fixing this.
This adds a command in your Tools menu called
Reset Ease + Force Sync After. It resets all Anki cards to have the
default ease value, 250%. You have to install it manually.
Easy to not change the card’s ease
value. By default, Anki modifies it by -20%, -15%, or +15%,
respectively. This is a bad idea, and I go into more detail [in my other
anki post][TODO]. This goes hand in hand with ResetEZ. It also needs to
be manually installed.
See the Making use of subdecks section above. This randomizes review cards in subdecks.
7. Other Plugins
There may be plugins for the language(s) that you’re learning. Japanese has several, for example. It’s worth searching "anki <language>"" to see what’s available.
Considering that we end up using Anki for years, it’s worth taking some time to improve it and read the documentation closely. Is there a plugin or change that you feel like I missed? Let me know.
I have also written a post on general things to avoid while using Anki that I recommend you check out.
9. Further Reading
 MAINTENANCE OF FOREIGN LANGUAGE VOCABULARY AND THE SPACING EFFECT, Bahrick Et al (1993)[https://s3.amazonaws.com/andrewzah.com/studies/Bahrick-et-al-1993-spacing-effect.pdf]
 Low-Key Anki[https://massimmersionapproach.com/table-of-contents/anki/]
 Targeting an 80-90% Success Rate in Anki[https://eshapard.github.io/anki/target-an-80-90-percent-success-rate-in-anki.html]
 Low-Key Anki: No Penalties or Boosting[https://massimmersionapproach.com/table-of-contents/anki/low-key-anki/low-key-anki-no-penalties-or-boosting/]
 Low-Key Anki: ResetEZ[https://massimmersionapproach.com/table-of-contents/anki/low-key-anki/low-key-anki-no-penalties-or-boosting/]