I've been using ANKI for about half a year now, for both Kanji reps and vocab.

I use these decks:
- RTK deck keyword --> kanji
- Kanji --> keyword deck w/ yomi (maybe I will use this as a replacement of the one above soon)
- core 2k deck
- Nayr's core 5000 (been jumping between core2k and this)
- JLPT 5 and JLPT 4 decks (single words only)

However, since I don't know a thing about the technical / setup aspects of ANKI, I feel like I might not use it to its full potential. I did change the steps in many decks from "1 10" to "1 20 40 80" recently, because I read somewhere it's more efficient than reps after 10 mins only. And I complete all reps before adding new cards.
But other than that I use standard settings.

I'm not sure what else I am missing, or what I could do differently to maximize the output of my studies.
Any advice is welcome.

"1 20 40 80" sounds good. For vocabulary decks, I have been using "1 10 15 30 70" which has been working well. Yesterday I changed it to "1 10 15 45 120 600 600" and a graduating interval of 4 days because something like that worked so well for my cardboard flashcards.

For grammar decks "1 60" seems to be enough.

Try different things. If something works well, change it anyway after some time to see what happens.

maxwell777 Wrote:- RTK deck keyword --> kanji
- Kanji --> keyword deck w/ yomi (maybe I will use this as a replacement of the one above soon)

maxwell if you have a few reviews per day for the first deck, I suggest you not to ditch it! This sounds exactly what I do, "kanji -> meaning + readings". It works pretty well so I would suggest it if one has still to start and don't want to lose time on writing, but if you have already a production deck and it doesn't take you much time, then I think the benefits will surpass the disadvantages! If you have difficulty try to add a japanese word beside the english keyword.

PS: sorry for the off-topic!

tikka Wrote:"1 10 15 45 120 600 600" and a graduating interval of 4 days because something like that worked so well for my cardboard flashcards.

a graduation interval of 4 days seems weird because you do 2 10 hour intervals and then jump to 4 days. It' makes sense to me that each interval is the same multiple of the previous step so it's a nice smooth curve. Also, I've always thought the jump from a few hours to 1 day way my crux so my exception was that I doubled up the one day interval. But hey, it's not about theory, it's about how well it works for you, so nothing wrong.

Here's the intervals I use for new vocabulary: 0.5 3 15 60 180 540 1440 with a graduating interval of 1 day so I get that double 1 day bump.

Other things you can try is tweeking your starting ease depending on your accuracy at answering the cards. According to supermemo, the most efficient learning is when your accuracy is around 70%, but that got a little demoralizing for me. Lately, I've been happy to trade efficiency for the enjoyment of getting around 90% accuracy. I'm doing a lot of review, so when I start adding new vocab, I'l probably settle between 75-85%. Whatever you decide, your main tool to adjust your accuracy will be starting ease. Anki's default is 250, but that's way too high for me.

I just use 1 5 10 (30)* graduating to one day for vocab. Recently, I've been adding thirty per day, which keeps my reviews between 150 and 180.
* I only use this if I decide to add new cards during the day, and it's only meant to force me out of the deck when I hit that interval (work ahead is only ten minutes); when using that interval, I don't review again until I get in bed, which is when I normally do new cards.

I changed the lapsing interval modifier so that it doesn't reset progress entirely.

yogert909 Wrote:a graduation interval of 4 days seems weird because you do 2 10 hour intervals and then jump to 4 days. It' makes sense to me that each interval is the same multiple of the previous step so it's a nice smooth curve. Also, I've always thought the jump from a few hours to 1 day way my crux so my exception was that I doubled up the one day interval. But hey, it's not about theory, it's about how well it works for you,

Yogert's comments are worth further reflection. Not sure about the "smooth curve" comment, but I strongly agree with his conclusion.

With respect to the "smooth curve," science's understanding of human memory is rudimentary at best. The concepts of short-, medium-, long-term memory are not well understood; perhaps these buckets don't actually exist. Are memories handed-off from these buckets, and if so, how?

Assuming various memory buckets existed, a "smooth logarithmic function" may properly represent the process within one bucket. Those curves are all over nature. But each bucket may have its own function. And the hand-offs from short- to medium- to long-term buckets may "jump," an independent function in its own right. Maybe the memory for kanji, vocab, and grammar functions differently. And the memories of people I know varies widely. I would like to believe that people who study hard improve their memory function, but as people age, their memory may decline...

A simple algorithm just is not going to work well for everyone (or anyone, for that matter). Perhaps if we could visualize forgetting curves for a deck, that would be a good start to customizing our own ANKI intervals. Is there a plug-in that allows that?

For example, in my case, I believe that I can memorize new items very quickly, but my retention over the following week is weak. My sense is that ANKI gives me too many reviews on the first few days and too few during the following week or so. Or maybe not. That is why I am reluctant to mess around much with the ANKI intervals.

The dynamic approach of the more recent of Super Memo versions seems to have promise, so I might give that a go with vocabulary. Always tough to change platforms!

scooter1 Wrote:Assuming various memory buckets existed, a "smooth logarithmic function" may properly represent the process within one bucket. Those curves are all over nature. But each bucket may have its own function. And the hand-offs from short- to medium- to long-term buckets may "jump," an independent function in its own right. Maybe the memory for kanji, vocab, and grammar functions differently. And the memories of people I know varies widely.

I totally agree with what you are saying. I know for sure that my kanji forgetting curve is a lot different than my vocabulary forgetting curve because I am more accurate on young kanji cards than mature kanji cards, but it's reversed for vocabulary.

However, without knowing the exact shape of your forgetting curve, it would be difficult to guess the optimal interval function. I'm guessing that even if you did know your exact forgetting function, it would be a lot closer to a smooth log curve than tikka's example with two 10 hour intervals followed by a jump to 4 days. But that might even make sense depending on tikkas study schedule because everyone has to sleep. So depending on their study schedule, tikkas intervals have him/her studying cards in a much smoother log curve than it appears. But this all is a little to complicated for me to figure out, so I just go with a smoothish curve (rounded to prettier numbers) and get on with studying.

I do wish we had the tools in anki to figure this out a little better and it's even a project that I've been working on in my limited spare time. I've considered supermemo too, but I need a mobile app and as fas as I know there isn't one. I suspect that memrise and duolingo are working on sophisticated algorithms, but those apps are difficult to add material and customize.

"1 10 15 45 120 600 600" practically means that easy words graduate after three to four days depending on my study schedule, which is different from day to day. Difficult words will stick around a very long time since if you can't remember it you have to repeat it for several days. A mistake made for a word before it graduates will always lead to it being showed to me for those couple of days which will probably lead to a nice and EASY learning experience.

And the "easy" part is important for motivation etc as everyone knows.

I think in the end this weird algorithm is not that different from the more normal ones. With this, I guess I'll see a new word on 3-5 days on the first week. Isn't it more or less the same as with the more smooth-curved algorithms? The first days are a bit too easy and the last days are a bit too hard so it sounds balanced. And since the new words are always mixed with last week's cards and mature cards, the result is even more balanced in reality.

I'm not saying this is some miracle method nor that you should try it, I just want to see what happens.

thanks for your replies guys, that already helped a lot.

I'm doing something like 1, 2, 2, 5, 10, 20, 40, 80, 160 right now and it works quite well. Too bad I wasted time with the standard 1 10, before giving that more thought.

However in the meantime I'm through with RTK1, so it's great to be able to just focus on vocab and grammar.

One question, I still don't quite get how the "starting ease" percentage works. What exactly does that affect, and in what way?

thanks

When you use settings like "1, 2, 2, 5, 10, 20, 40, 80, 160" or "1 10 15 45 120 600 600", how do you deal with the huge increase in the number of reviews that you have to do? You essentially see each card 7-8 times before it graduates, so an addition of 10 new cards leads to 80 viewings of that card. The red "failed" stack will be huge, which is something that I personally find quite demotivating.

Well, in my case, I don't really think of red cards as "failed", but as cards I have yet to learn.

As for the many repetitions, I actually find it quite motivating to be able to read the same cards faster and faster.

Rivvie Wrote:When you use settings like "1, 2, 2, 5, 10, 20, 40, 80, 160" or "1 10 15 45 120 600 600", how do you deal with the huge increase in the number of reviews that you have to do? You essentially see each card 7-8 times before it graduates, so an addition of 10 new cards leads to 80 viewings of that card. The red "failed" stack will be huge, which is something that I personally find quite demotivating.

Same here. I don't really use steps but for my last deck my graduating interval was 21 days (then next review "hard" is 10 days and "good" is 40, my kind of spreading). And I always graduate new cards because the short term loop wasn't doing it for me. Which is why I also prefer to push "hard" instead of "fail". I barely use the "fail" button.

[RawToast]

Rivvie Wrote:When you use settings like "1, 2, 2, 5, 10, 20, 40, 80, 160" or "1 10 15 45 120 600 600", how do you deal with the huge increase in the number of reviews that you have to do? You essentially see each card 7-8 times before it graduates, so an addition of 10 new cards leads to 80 viewings of that card. The red "failed" stack will be huge, which is something that I personally find quite demotivating.

Personally I don't go to that extreme, but I found "1, 10" didn't provide enough reviews. I'd fail the new cards the next day (for a few days). I found simply adding an extra step for both new cards and lapses provided enough coverage for me. e.g. updating lapses from: 5 to 5, 15

When you add more steps you'll find your failure rate will drop. So you get a mixture of more work up front and less work in the long run. Although, I agree that 8 steps sounds rather extreme!

For myself, my mature card succes rate went from ~75% to ~90% by adding the extra steps (or I just got better...)

I agree that the default 1,10 is not enough spacing, I currently use 10, 10, 60. The 60 ensures that I at least get to see it again in the next review session, which can be the same day or the next. Often I am confident that I will remember the card after the 10,10 part, but still fail it after an hour has passed.

EratiK Wrote:I don't really use steps but for my last deck my graduating interval was 21 days (then next review "hard" is 10 days and "good" is 40, my kind of spreading)

You must have an amazing memory if you can remember a new card 21 days since your initial review, or do you add material that you are already somewhat comfortable with?

Rivvie Wrote:You must have an amazing memory if you can remember a new card 21 days since your initial review, or do you add material that you are already somewhat comfortable with?

Far from it, and I don't add anything, my last deck was core 10k. But yes, I was already comfortable with about a third of the deck from dictionnary look-ups of stuff in the wild. Then another third of the deck was either transparent or really easy in the context sentence (and with predictable on'yomi), stuff like 脱落, 引力 etc. And then there was a tricky third, but cycle through it for a couple of months with the 10-20 days interval (so that's 3-4 reviews) and if you don't already know the word by then you know the sentence so you can guess the word. I know it's kind of cheating but I don't review to master every word perfectly (I used to do production for that), just to have a sort of live mental entry so when I meet the word again in the wild I'd react, and if I never see it it doesn't matter if I don't remember it well.

The point of all that is anki is really a pain for me so I do as few of it as I can, and of course that means long intervals.

Rivvie Wrote:When you use settings like "1, 2, 2, 5, 10, 20, 40, 80, 160" or "1 10 15 45 120 600 600", how do you deal with the huge increase in the number of reviews that you have to do?

The reviews became so easy it's almost fun to do them. I added two more steps so it's now 1 2 5 10 15 45 120 600 600. I can't believe how easy it is now! It's totally worth it, no matter how many extra reviews I have to go trough.

Usually I've been adding 20-30 cards a day. Yesterday I added 130 and today 100. "Today you studied 1622 cards in 148 minutes" is what Anki says now - but I still don't feel like I'm going to die.

I want to see if it is possible to learn 700 words in a week. Maybe it kills my motivation and slows down long-term progress, but still.

tikka Wrote:The reviews became so easy it's almost fun to do them. I added two more steps so it's now 1 2 5 10 15 45 120 600 600. I can't believe how easy it is now! It's totally worth it, no matter how many extra reviews I have to go trough.

Usually I've been adding 20-30 cards a day. Yesterday I added 130 and today 100. "Today you studied 1622 cards in 148 minutes" is what Anki says now - but I still don't feel like I'm going to die.

I want to see if it is possible to learn 700 words in a week. Maybe it kills my motivation and slows down long-term progress, but still.

Interesting, I might give it a try, easy and fun reviews sound good to me. Might I ask what your level is and from where you add your cards? Do you use a premade deck?
I am still at a beginner level (~2000 vocab) and struggle with all the new info when I add more than 10 cards a day, I can't even imagine what is must be like at 100 cards... I do use a premade deck, but I think that the randomly presented vocab is killing my retention. Also, my reading speed is not that fast yet and the reviews take up a lot of the time I have for Japanese in a day, leaving me little time for native material.

tikka Wrote:I want to see if it is possible to learn 700 words in a week.

My personal best is 1000 in a single day

Didn't add any for a while after that though, which kind of defeats the purpose

Rivvie Wrote:Interesting, I might give it a try, easy and fun reviews sound good to me. Might I ask what your level is and from where you add your cards? Do you use a premade deck?
I am still at a beginner level (~2000 vocab) and struggle with all the new info when I add more than 10 cards a day, I can't even imagine what is must be like at 100 cards... I do use a premade deck, but I think that the randomly presented vocab is killing my retention. Also, my reading speed is not that fast yet and the reviews take up a lot of the time I have for Japanese in a day, leaving me little time for native material.

I've done half of RTK2 and the N5 & N4 vocab decks from tanos.co.uk. Im now half way trough the N3 deck. By the way, those decks become easier if you randomize the order. The original ordering is very confusing for some reason.

A couple of things I found helpful for learning new words:

Remembering the Kanji 2. I don't know what I'd do without it. The easy groups were a great way to start learning on-yomi and I can't see why the rest of the book wouldn't be useful sooner or later. Most (all?) of the signal primitives have more than one reading so people often think it's useless information but I think it helps.
There's probably some lists of the phonetic primitives online as well, in case you don't want to buy the book. They are not Heisig's invention and different sources will probably list them differently since there is a lot of overlapping primitives and some are listed as "pure" in RTK2 even if they strictly aren't.

In addition to phonetic groups, you should play with words to make them easier to remember. Someone else could probably explain this better than me, but anyway. 大抵, 活気, 生き物, 真っ赤, 反抗, 癖. Look at those words for example. They were easy to learn because they are very close to words I already know in my native language. Art, poop, forever-lasting skiing shoe, sausage, pitchfork, urine. The meanings are not related in any way but it still helps. Somebody mentioned a similar one in English,　好き -> ski. They don't have to make any sense, anything helps. Don't think about them too much and let yourself forget them as soon as you remember the word easily.

tikka Wrote:Remembering the Kanji 2. I don't know what I'd do without it.

Do you review single kanji to readings, or compound on front and reading on back (as Heisig suggests)?

cophnia61 Wrote:Do you review single kanji to readings, or compound on front and reading on back (as Heisig suggests)?

I do it from compound to reading and meaning. There's no point in memorizing the readings separately. Also, I think there's no need to make any effort to review the primitives, just the words.

If I forget a word, I usually check the related words from the book to reinforce the understanding of the signal primitive in that word, but that's it, there's no need to use SRS for the primitives or anything like that.

tikka Wrote:Remembering the Kanji 2. I don't know what I'd do without it.

I appreciate the advice. I still have a copy of RTK2 lying around which I bought years ago, but never used. I will give it a shot. Which anki deck do you use for reviewing?

On the word play for remembering readings, yes I do that sometimes too. It makes the reading very memorable, however in most cases I can't seem to think of a word which even sounds a little similar in my native language.

Rivvie Wrote:Which anki deck do you use for reviewing?

I made cardboard flashcards for RTK1 and added the readings to those.

Quote:On the word play for remembering readings, yes I do that sometimes too. It makes the reading very memorable, however in most cases I can't seem to think of a word which even sounds a little similar in my native language.

Have you seen Chapter 11 of RTK2?

anotherjohn Wrote:My personal best is 1000 in a single day
Didn't add any for a while after that though, which kind of defeats the purpose

Something like that might not be such a bad thing after all. If you add a month's worth of new cards at once and then don't feel like adding more for a month..

tikka Wrote:I want to see if it is possible to learn 700 words in a week. Maybe it kills my motivation and slows down long-term progress, but still.

I've now learned 600~ new words in six days. This is horrible. I quit. No more.