Optimize your Anki – You’re Overtesting Yourself on too few cards.

Guide by jadmin 312 created at 2018-06-29

Author: Vladz0r Original Post: vladsperspective.wordpress.com

Anki hasn’t been treating you right. You want to make real gains, but you can’t imagine doing more than 20 or 30 new cards/day, can you? After all, you keep hitting “Hard” on yesterday’s vocab, and it’s just not sticking. Anki is supposed to be a tool you use to improve your learning, not something you hate and struggle to you. You should be learning to use the tool a bit better whenever you can.

Here’s the gist of this radical shift to Anki that will make learning much less frustrating:

You can alter your cards, so that they all appear less frequently. You can utilize a fixed 300% interval with a 2 day graduation rate, and then see a sentence card today, then 2 days from now, then 6 days after that, then 18, then 54, and then in 162 days. That’s 5 reviews of a card in 8 months. And you can actually know what’s on the card.

Japanese is Cumulative

One reason why we can use such high intervals in Japanese is because we’re often not learning completely new information. You couldn’t do this with unrelated things like “The top 10000 Jeopardy Facts” but for testing your comprehension, and seeing a vocab word in context, you absolutely can use a 300% interval. You can also use a high-ish interval % for vocab cards, especially if a card contains kanji you’re familiar with. If you learn a sentence that’s like 「私は負けない」 and then later something like  「いや、そんなわけないだろう」and then later「僕は負けるわけにはいかないんだ。 」You’ll be over-testing your older cards if you keep them at Anki’s rate. We went higher rate, more sentences, more exposure.

The second reason is because Anki overtests you. When you hit Hard, you make a card’s interval decrease, so you wind up seeing a card more often. It seems ok in theory, and the memory formula, the whole “Anki’s so awesome because it tests me right before I forget something!” is great in theory, but generally we can increase the rates. Ja-dark (source below) mentioned that you’ll often know things in Anki for several months past a card’s interval, once we test a card 5 or 6 times, even with a 300% interval. In the default settings of Anki, by the time you get a card to a 2-month long interval, you’ll probably actually know the information a year from now, if it’s cumulative information. (having a kanji you know, being an old sentence that’s become easier, etc.)

Additionally, when you screw up a card, you don’t have to completely reset it. I’ll remind you of the 2->6->18->54 day intervals. With this altered system, if you hit “18 days” and then forget the card 18 days later, you hit Try Again. You test it 10-20 minutes later, and then it gives you the option of “18 days” instead of “54 days.”This revision process is good enough, because re-learning is different from learning something over again, especially with something like a language that is very cumulative.

Important Note:  Card Creation Principles
Your Anki cards should be fun from the start. You ought to have seen these words from a TV show or anime, or a website. You should add funny pictures of them. If a card is frustrating you, and is wasting your time, delete it. If you hesitate over whether you should delete a card, that means you should delete it. You can do Core10k vocab cards, but you should know what Kanji is on them. You should do sentence cards, but you should be familiar with most of the words in the sentence. Ideally, they should come from things you actually read or experienced, but we can always use some fundamental Core sentences and Grammar examples to start out with, learn new kanji meanings, and refine what we know.

Never cram. Never study a card you hate. Seek to improve it first.
I’ll have info in other posts on how to up your card creation game.

So how do I fix this shit already and make da gains?!

Part 1: Anki Addons (install all of these)


Low Key Anki: Pass/Fail - LINK
-this changes Anki’s behavior so that we simply answer Yes or No for our cards.

The reality of a card feeling “hard” or “easy” and us judging the difficulty ourselves is actually really arbitrary.

The number of revisions, or times we see a card, is what improves this, but we shouldn’t be drastically reducing the interval increase by hitting Hard all the time. Also, we can increase the standard interval increasing.


Low Stakes Anki: No Penalties or BoostingLink
-There’s a tldr that Ja-dark wrote on the addon page, so check that out.

Change order of review cards in regular decks: Link
-You can re-sort your deck so that it goes in order of Increasing intervals. The default is by Decreasing, but I prefer Increasing. This adds a bit of predictability to our study sessions, as we get through our harder, more recently added cards first, and can build momentum and blow through the easier cards.
-Go to Tools>Add-ons>change_order_of_review…>edit> and change 「order = “ivl dsc”」 to 「order = “ivl asc”」to modify the addon. Save and restart Anki.
-Note: This won’t work if you have a deck hierarchy but Filtered Decks are easy and I wrote a quick guide below 

put ALL due -learning- cards first ×Link
-this is useful so that you can keep learning New cards without seeing like 80 new cards at once that haven’t passed the 10 minute step interval. It improves the behavior of Anki a bunch, if you’ve had issues with trying to learn 30+ new cards at once.

Install all the addons, restarting Anki, then modifying the 3rd addon as I mentioned. (And then restarting Anki again).

Anki settings:


In Anki, go to Tools>Preferences>Basic, then set the Learn ahead limit to 2mins
-With the default Anki behavior, we can see a card, hit [10 minutes] on it, and then immediately hit [2 days] if there are no other cards we’re studying at the time.

Creating your new Options groups:
Click on a deck and hit the cog at the right of it. and then hit Options.
We’ll make a new Options Group, so hit the cog in the new Window, then hit Add. We’ll call it 20 250%

Now, fill out the settings as follows:


New Cards: (pic)
-Steps(in minutes): 1 10
-Graduating interval: 2 days
-Easy interval: 2 days
-Starting ease: 250% (or otherwise)


Reviews: (pic)
-Maximum reviews/day 999
-Easy bonus: 100%
-Interval modifier: 100%


Lapses: (pic)
-Steps (in minutes): 1 20
-New interval: 100% (this keeps it from resetting our card)
-Minimum interval: 2 days
-Leech threshold: 5-7 lapses. (I use 5, and then hopefully figure out why I keep failing that card.)
-Leech action: Tag only

I make a 200%, 250%, and 300% for each amount of cards. I have a 225% for certain cards, though. You can hit [Add] from your current Option, rename it from, for example, “20 250%” to “20 300%” and then modify the Starting ease, and you’ve made your new Options group in a few seconds. Here is what part of my list looks like.

Recommendations for Intervals:
200-225% for Kanji, or Vocab if you’re building your foundation. Anything that you feel is foundational, and doesn’t seem to be built up from something, ought to use a lower %. I use this for classes sometimes for memorizing facts.
250%+ for Vocab if you know kanji, especially if you’re using the Kanji+Furigana method
300%+ for Sentences in i+1 format

My Intervals:
Kanji Writing (RTK): 200-225% – these are fundamental, and I want to test them a good amount, so I keep them at a low-ish %. I use 225%, but I was using 200% for the first 1000 or so.
Vocab Writing: 300% – I can already write the kanji for these, and I know the word, so the difficulty is generally low enough to warrant a high %.
Vocab (New): 250% – these have Kanji+Furigana on the front, picture and meaning on the back, with an example sentence if I have one. I keep the same interval for my Eng->JP cards and Listening-Only cards (which uses the audio and hiragana with common homophones listed on the front, Kanji+Furigana, picture, meaning on the back.)
-I primarily learn vocab with kanji I can read/write, or from words I’ve learned in context, so I can get away with a 250% interval. For a beginner, I’d recommend a 200-225%. I think that your kanji knowledge becomes vast and words become really easy to learn, you can crank it up to 300%+ like ja-dark suggests.
Sentences (From Sub2SRS, Core10k): 300%. These sentences are sorted in (i+1) by MorphMan, meaning I generally don’t see sentences with more than 1 new word, and they’re for comprehension and context.
Grammar Sentences: 250% – I may spend more time on these, and they tend to have less familiar words for me.


Reviewing on AnkiDroid, Anki for iOS with these hax pro settings:
So, the settings we set in Anki and the add-ons we use kind of accomplish the same thing. The settings work on Anki for mobile devices, AnkiWeb, or an Anki client without the addons, and the addons will override a lot of Anki’s settings. When studying on AnkiDroid, hit “Good” every time you study a card. The interval rate won’t change, because we set the Easy bonus and Interval modifier settings we use, but the increase for the review what we want when we hit “Good.” on the mobile versions.

(Recommended) Using a Filtered Deck and a Deck Hierarchy if you have multiple decks:

A filtered deck will let you do your reviews separately from your new cards. I often set a timer and do 15-20 minutes of reviews -> 15-20 minutes of new cards. I feel like breaking it down this way makes it more manageable. I also kind of have to do it because I use a deck hierarchy (multiple decks) and want to keep the Increasing Intervals.

I suggest keeping your decks bunched together like this

Then, click on your Japanese deck and hit the F Key, or go to Tools>Create Filtered Deck.
Your search should look something like this:
Search: deck:”Japanese (New)” is:due -is:learn
Limit to: 999 cards selected by Increasing intervals.
Here’s what mine looks like, though I cut out some decks. I do my writing reviews + new cards all at once, outside of my Filtered deck:

Mindset, General Overview:
-Every revision is an improvement. Think  “plus 1” every time you answer a card. As an enthusiastic man once said,”We’re going to keep winning, and winning, and winning — whether they like it or not!”
-You’ll mess up the cards you learned recently. A bunch of them.
-This is completely natural. This strategy is something that’s for long-term learning. Through revisions and passing/failing the card, and having time go by, we learn the card better. When you mess up, simply hit Try Again, then 20 minutes, then 2 days again. We’re getting our revisions in. The magic will happen, and through exposure and re-testing the card, you will nail the cards down. You’ll find that 1-2 weeks from now, the cards that you messed up today will stick well, and you’ll be happy you decided to do upwards of 30 new vocab/day along with your sentences and other card types you’re experimenting with.
-Keep your cards fun, engaging, interesting. You’ll learn how to spice up your cards in some other posts, so that even the Core10k can become more fun. Don’t study frustrating cards.
-Doing like 50+ cards/day is possible, depending on what they are. In fact, if you know your kanji, I’d encourage it. On a good day, I do my 50 new kanji in RTK, 30+ new vocab, 50+ sub2srs listening sentences, some Eng->Jp and Listening-Vocab cards. Basically the 50 kanji and whatever else I have the energy for. If I didn’t have kanji, I would probably be aiming for 200 cards a day, mainly from sub2srs reading and listening cards, followed by new vocab, some core10k sentences, grammar, and some experimental things like cloze-deletion cards I want to try someday.

Credits:
My main source for this information is Ja-dark and the research he has done.

AJATT’s – http://www.alljapaneseallthetime.com/blog/all-japanese-all-the-time-ajatt-how-to-learn-japanese-on-your-own-having-fun-and-to-fluency/
AJATT gets a bad rep because he wanted to monetize his website, but the core of his website is a gold mine of inspiration, self-improvement ideas, and strategies you can implement to try and master  Japanese.