If you know me in real life you probably also know that I love Anki. It has quite literally changed my life: without Anki I would not have been able to learn another foreign language and thus not live abroad.

Having used Anki on and off for over five years total here are some personal notes on what I currently think to be effective or ineffective.

Effective card writing

First some notes on how to create better cards.

An answer should only be one item

For example for the question “What is the capital of France” the answer should only be “Paris”. The answer should not be “Paris, population of 2 Million” or “Paris and the flag looks like this”.

It should be absolutely clear whether you knew or did not know the answer. Let’s assume you have both the name of the city and the flag of the city as an answer. In that case if you only know the city, did you know the answer? (Do you press “Good”?) If you press yes you will not learn the flag as much as you would need and will increase the probability of forgetting it. On the other hand, if you only allow yourself to press “Good” when you knew both the capital and the flag, the card has become more difficult. Remembering the capital is likely easier than remembering the flag but having it as one card will make it appear more often and thus take more time.

For such a case I simply recommend to have two cards:

• “What is the capital of France” → “Paris”
• “What does the flag of Paris look like?” → flag

What about lists?

Sometimes you need to have multiple items as the answer. For example if you want to remember the order of something. In such a case I still try to keep the number as small as possible and really try to not have more than roughly six items. It is also usually easier to remember a word or phrase from which the order can be derived - effectively compressing the list to “one item”.

The question should uniquely identify the answer

Given the question on the front, the answer should be (reasonably) uniquely identifiable. For example “What happened in 1990?” would be too vague. There are a lot of things that happened in 1990! If you have cards like this you might be able to remember the question-answer set for you specific deck. However I found, that with longer intervals I tend to forget those “special answers” and even though I knew the answer if asked specifically I could not answer the vague card (which is quite frustrating).

Avoid orphan cards

I love learning random stuff all over the place. However creating random cards with little connection to existing knowledge is usually not very effective. For example if I know nothing about astronomy but suddenly create cards for the atmospheric layers of saturn, it tends to be hard to memorize. This is most likely because there is not enough “sticky context” around the topic. Without the context one needs to rely on rote memorization which is usually hard.

In such a case it might be better to first understand/memorize a rough big picture and then fill in the blanks. By doing so you will have an interconnected web of cards instead of separate orphan cards. If that sounds like too much work and doesn’t seem worth it, maybe the orphan card was not that important to begin with!

A concrete example of this for me was remembering historic dates without the context. Even though it was only a handful of cards, it was one of the hardest decks that I tried to remember.

Effective card management

Next are some notes on how to organize cards in Anki.

Different decks for different “information types”

I have many different types of information in various decks: languages, geography, linux commands, etc. For these I found it very useful to split them (roughly) into separate decks: One for geography, one for language 1, one for language 2, etc.

The reason for this is, that with separate decks I can create separate presets. Separate presets then can be optimized separately which leads to better learning intervals and overall decreases the time spent for learning cards. (At least that is what I tell myself) For example, learning geography is way harder for me than learning linux commands. Having them in separate decks allows be to quickly memorize commands while take time memorizing the geography deck.

Effective card learning

Finally some notes on effective learning of cards.

Think twice before increasing “New cards per day”

It is very tempting to increase the number of new cards per day. “If I learn 50 cards a day, I will be finished in two months!” I did that more often than I would like to admit. Every time was hell. The more cards you add, the more cards you review. The more cards you review, the less concentrated you get. The less concentrated you are, the more cards you forget. The more cards you forget, the more card you need to review.

Spending more than two hours on Anki on every day of the week for several weeks is neither fun nor effective (for me at least). So if you think you could increase the number of new cards per day, add one. Then do that for some time and add another one if you feel you have the energy.

Do a card improvement on “Again”

Whenever I press “Again” on a card I try to look at the card and improve it:

• Was the question ambiguous?
• Is the answer more then one item?
• Can I add a metaphor / rhyme / etc. to the backside to make it easier to memorize?

By doing so you slowly learn which cards are hard to learn and how to make it easier. This is something I only recently started, but found to be quite effective.

Fix cards on the spot

Even if you knew the answer, but the card is a bit weird or has a spelling mistake, etc. I try to always fix the card on the spot. For some time I flagged the cards and went through them on the weekend - but that didn’t last long and usually I would forget what it was that I wanted to fix.

Now I simply fix the card on the spot on my phone. Also this “continuous refinement” is weirdly satisfying!

Delete useless cards

I struggle with this one. However some cards (or even whole decks!) are just not worth learning. In those cases I give myself the permission to delete them.