The Jazzcat Output Method Converstation

The Jazzcat Output Method Converstation
Photo by Manja Vitolic / Unsplash

I talked to Jazzcat about my retention and wanted to share it here 😄

Jazzcat is a member in The Moeway Discord famous for having a very weird and unique approach to Anki.

However, this approach has led them to obtain one of the highest ranks via vocab quizes on the server, and their Anki graphs look amazing:

Jazzcat

Basically they find Anki easy. This is what mine looks like:

Bee (me)

So they must be doing something right.

I spoke to them about my personal problems and retention issues, and have reposted our convo here for all to learn :)

Conversation

Me

i have really bad comprehension in anki and i was wondering if you had any input?

i was in difficulty hell but i fixed that by getting rid of fsrs params, but my retention is still trash. since i started 2 years ago its always been around 50% on young, and like 90% on mature

6 weeks ago i started reading 1 hour a day visual novels, for the past week i have done 4 hours a day (about 15k - 20k characters a day)

today i got rid of learning steps for new cards to let fsrs handle them (as i saw you post about that), and i have 10m relearning step for lapses leeches on, threshold 10 with suspend (i will just remine it. sometimes i make kanji cards too) i am doing rrtk450, i am at around 350/450 and its helped. i also make kanji cards myself if i know all the kanji apart from one, i just do kanji -> keyword in english

my retention for my mining deck is 47% on young, 84% on mature

last week for kaishi (pic 2) its actually quite bad, but i stopped doing kaishi because i dont believe in "learning in anki" and just want to mine cards

however i have a deck which is for output (pic 3) for the tmw quiz

and this retention is significantly better (although its a lot less cards! 1/2 new max a day)

i was wondering if you could teach me more about your output anki method or advise me on anything im doing wrong? seeing my retention for my quiz deck is crazy and i think output helps me a lot, but im not sure how to turn lapis into output

i have also attached my anki settings, i have "learn ahead" set to 0 minutes too. i think doing multiple smaller sessions per day would help me too

Jazzcat reply

  1. Empty learning steps: just to be clear, if you do this I recommend using that first Again (and only that first Again) like “I know this but I know it is/will be an especially hard word for me”
  2. As long as those numbers keep up, I don’t think there’s anything wrong with 50pc retention gradually becoming 85 at all. The fact that it becomes 85 means you are learning those words, the fact that you’re getting 50pc up front means you’re testing yourself over large enough spaces to make it challenging, and that challenge eventually brings benefit
  3. There probably is an easy enough way to convert lapis to output. But studies show the reason people benefit more from writing their own cards is because they gain a sense of what they actually need to put on the card to make that particular word stick for them personally. This makes it a little tough to show people how much better it is, because any time someone shows you a card, they’re showing you one optimized for someone else, and not for you, haha. Figuring out what you need on that card to make that card stick for you is something only you can do, but you can do a better job of making it work for you than anyone else can
  4. In my case, explaining how I got here is a long story, but the only type of card I’m doing right now has the meaning on the front (maybe in English, maybe in Japanese, maybe with a cloze sentence with the target word cut out of it, maybe not, sometimes even in Spanish if that forms the most natural link in my mind: any time captions are on a Spanish soap opera they always say “se miran fijamente” for that moment where two people dramatically stare at each other at the end of a scene and 凝視 is that in Japanese so I just have cards <-> for those two haha).
  5. So maybe I have something like, “loaning someone something (informal, without interest, just expecting the item back at some point)” -> I say taiyo or write たいよ or whatever. Flip the card, I see 貸与 on the back, but the point is I’m connecting the concept to the sound/spoken word “taiyo.” Now when I see 貸与 I have the meaning I associate with those kanji to hint me to the right meaning, but I also have something that I don’t get any other way. I have the fact that these can be read たい and よ hinting me to the word たいよ, that I just simply already know as a word. I’ve found that this is putting far, far less of a gap between my reading and listening comprehension than other methods did. (Though to be fair I was at the peak able to write over 3k kanji from memory, and maybe still can, so I was very imbalanced strong at kanji before switching to this)
  6. If you really feel it’s helping you then ignore me, but the time on single kanji -> English keyword is better spent elsewhere IMO. I still have a deck setup for handwriting kanji in my own cards, but now I only use it for kanji that aren’t in Ringotan because I’ve come back around full circle to thinking Ringotan just does a brilliant job. The algorithm is very inobtrusive, especially if you set it to “review somewhat less.” It repeats things a few times after you get them wrong, but it consistently assumes you’re going to get them down after a few reps and escalates them out to larger gaps again. And I think a huge part of what’s going on with SRS is just the fact that you’re trying to recall these things over, say, three month gaps, gives a signal to your brain that says “Okay, we need to store this in a way that can last three months then.” (This is also why I don’t think your low retention on new is a problem, as long as the cards are maturing and you’re getting to 85 overall
  7. oh, I looked at the output deck’s stats first, young in the other were staying at 50 over a year. well, it sounds like you’ve only just started immersing consistently, so still

My reply

Empty learning steps: just to be clear, if you do this I recommend using that first Again (and only that first Again) like “I know this but I know it is/will be an especially hard word for me”

Can you clarify this more? So if I see a new card:

  • I know it, but it'll be difficult

I press Again (0)

  • I know it and its easy

Easy (4)?

But if I don't know the card at all as I have forgotten it, do I do Again or do you suggest burying it / suspending it?

Ringotan

Okay thank you! I think what I'll do is:

  • Switch to tag leeches
  • Every Sunday, go through my tagged leeches and add them to review on Ringotan (if this is possible, I have yet to read up on Ringotan)
  • Suspend those cards, as I'll see them again later

For me I find knowing the kanji in the word makes it easier for me, so I think this will work for me. Especially if I potentially have to write it, I think that would help too Thank you so much for a lengthy reply!! I appreciate it

Jazzcat

Basically, if you don’t know the card at all, and you press again, you might just be setting yourself up to hit Again 10 times in a row and suspend it haha. If you don’t know the word at all, then burying it lets you try again over even a larger gap if you want without any penalty in the algorithm. Just putting it down and trying the same word in a few days, or a week, by trying over that larger gap, that spaced effort itself is helping to plant the seed for a stronger memory. And then when you rate, no matter what rating you give, you’re always saying something consistent: this word has enough of a hook in my brain I can already recall it over at least 1d successfully. That’s a somewhat objective measurement for a point at which you’re ready to begin more spacing. Instead of Anki getting the idea that you need to see words on average 1d 1d 1d 1d 1d every single time as a rule, when maybe it’s just that it takes on average five times for you before the coffee hits right during your review and you actually pay close attention and “actually learn the word” that time haha

My plan

I plan to continue doing what I'm doing:

  • FSRS params set to default
  • Multiple smaller anki reviews a day
  • No steps on new cards
  • 1 step on relearning

And also:

  • Bury cards more often if I do not get them right away when reviewing Anki
  • Ringotan, in Topokanji order