Back

Anki - High fail rate.

#1
Hey, I'm kind of new to the forum and wasn't sure if I could ask this. Hope this is ok. Thanks.

I've been learning the kanjis with Anki for a while now, (since 5th of September last year) and have done about 1100 so far (I've been learning new ones at a fairly slow rate). I've been reviewing almost everyday (missed a few days here and there, a few more recently but in general not that many).

However constantly I've always had a fairly high fail rate, for both young and mature cards it's been very slowly dropping since the start.
They're roughly almost stabilized (although slowly falling) at:
75.2% correct mature.
Average time 23.6 seconds on each.
Correct young 80%
Correct first time 62%
Total correct 77.5%

I'm going to continue of course. However these seem a little low to me, I've never been great at remembering stuff and I'm happy to be getting anywhere at all. However it's a little discouraging and it's been bothering me for a while now.
My strategy is, to read about the kanji first and make up an image and write it once. Then after doing this for however many I'm going to learn that day (normally less than 5 now). I'll then go through them in Anki with my reviews (or at the end of my reviews). On Anki I see the word, remember the kanji, write it once then grade how well I knew it.
Here are my graphs:
[Image: snapshot20.png]
Should I change my Anki setting or they way I'm learning or is this OK?
I'd love to here what you think or any ideas on how I could do stuff differently.
Thanks you.

Edit: I should perhaps also mention that I have very little other Japanese in my life. I'm currently a six-form collage student in the UK and as such have little spare time so I rarely watch anime etc. I cannot speak Japanese at all but do know the kana. I am intending to learn the spoken language when I complete the basic kanjis so that they can provide a backing.
Edited: 2012-03-29, 6:29 pm
Reply
#2
I think if you're going that slowly but you're only retaining 75%, then this is not working out for you. Either you need to work out some more vivid and memorable system for tying the keyword to the kanji (it's not clear from the post whether you're doing RtK or not?), or you need to abandon this entirely.

Learning kanji will *not* help you when it comes to learning the spoken language. Learning to write kanji won't even help you that much with reading. You can do just fine in Japanese if you postpone learning to write the kanji until you have a better base in the language. (That worked a lot better for me, personally.)

You can probably fiddle with your settings in Anki to make your intervals shorter, but basically, I think this is not working out for you.
Reply
#3
My mature fail rate is about the same at 77% and slowly falling, though I'm hoping it'll eventually go around. My correct first time was horrible back when I was adding, no exact number but I didn't even get a third of the new cards right, and making mistakes with them became just as important to retaining as adding a story to koohii (done with RtK 1 and doing core atm, so I have barely added any kanji this month). That said, I did the entire first book 1. Jan - 1. March, for 2 months, and I feel like I was adding reasonably quickly.

If you're adding slowly and having trouble retaining, ask yourself, can you keep it up? Heisig is effective, but doesn't help with the readings. As long as you keep at it, you will get over the hurdle, and by the end of this year you'll be done with the first book and your mature cards will be doing better too, if you maintain your current pace. There's no point struggling if it's just draining your motivation though, or if you can't wait a year for the readings, and if so, I'd just go for a different method - it's much more important to be able to do it for years, preferably also with more time spent per day, than to burn out on an effective but difficult method and give up entirely. Basically, as long as you can keep doing it, you will succeed, no matter if you're using Heisig or mechanical copying on paper, so choose the method you can keep doing, because stopping is what makes people fail.
Edited: 2012-03-30, 6:57 am
Reply
May 16 - 30 : Pretty Big Deal: Save 31% on all Premium Subscriptions! - Sign up here
JapanesePod101
#4
If you are worried about your fail rate, then stop rating things so high. Anki has three "correct answer" buttons: Hard, Good, Easy. Start rating most cards as "Hard" if you are not completely sure about the answer (how to write the kanji). If you feel like you know it 100%, then give yourself a "Good". Never give yourself an "Easy". This should start to lower your fail rate.

I believe, in general, that your fail rate for new cards should be high, your fail rate for young cards should be medium, and your fail rate for mature cards should be low. In Anki, a card matures when it gets an interval of more than 21 days. A mature card should be a card you're really confident about. It sounds like your cards are maturing too fast, before you really know them.

I think I've read somewhere that you should be aiming for a fail rate of mature cards from somewhere between 85-95%. Anything less than that and you're rating cards too easy. Anything less than that and you're rating cards too hard.

But really, this is probably one of those things that you don't need to worry about. Nothing is wrong by having a high fail rate, it's just sometimes discouraging to do poorly on reviews.

I also agree with Fillanzea in that you might want to start working on something else (grammar?) to help with your motivation.
Reply
#5
Thank you. It's very nice to get feedback.

Yes I am following RtK and perhaps my images are a little weak at times. Although I don't think my fail rates have changed much since having to start making my own images. I'm still quite enjoying learning so my high fail rate isn't a huge bother, although improving it would be nice.

I know that the kanji won't help with the spoken language (sorry it was sort of unclear from my opening post, what I meant. I'd like to learn the readings and spoken language together next). But the kanji still seem like a good thing to learn first to help with reading, which in turn can be reinforced with spoken language as you write down things. Also it's just fun to be able to pick out kanji here and there in Japanese texts. :)

Fadeway, I am hoping I'll keep this up. I have managed ok so far and I don't feel the need to quit soon. Normally I can't seem to remember anything from copying etc. so when I tried Anki and found I was actually remembering stuff I more or less fell in love with it. So I'm a little scared to try other methods when this is working.

Partner, I'm generally trying to mark cards as you said, setting kanji to "fail" if I make a mistake (even if it's only small), however then when it turns up a minute later (and of course I remember it correctly) I feel the need to then mark it as easy the next couple of times, simply to "return it" faster to have a delay or a week or two. Maybe my cards mature too fast, as you said. They seem to accelerate quite quickly as they go? I might feel I know a card well after 1 month and therefore mark it as "good" but can easily have forgot in by the time it turns back up 3 months later. Yes you're correct fail rate on matures should be between 80% to 95% so I'm not doing that well. Should I try rating everything as "hard" for a while to see if that helps? Or maybe change my delays to something?

Thanks it was great to hear what you guys think. Any other suggests or ideas of what I should change my Anki settings to if I should change them at all. Maybe I should just try learn the remaining 900 or so quickly and then focus more on getting the rate higher?
Reply
#6
I was basically trying to say that you shouldn't rate things as high. Never rate something as "Easy", and only rate things as "Good" when you are 100% confident it in.

I believe you can also mess around with the Anki settings to achieve this, but I'm not sure what exactly you should change.

(I just took a look at the Anki settings, and you might be able to acheive similar results by reducing your "Initial button 3 internal" and "initial button 4 interval" by a few days.

You might want to check on the Anki mailing list to see if anyone else has played around with this.)
Reply
#7
Never mark cards easy just because you want them to move up. It's also not recommended to call cards easy if you've mistaken them the same day (logic is, you just got it wrong, how can you mark it easy a few minutes later?). I have some cards bouncing between 1 and 15 day intervals, constantly getting mistaken, but they don't go over 15 days for good reason, because I mistake them. A card should only be marked easy when you're sure you won't get it wrong, when it's so easy it's more a question of getting it out of the way without outright suspending it - usually cards like common primitives whose primitive meaning is the same as their standalone one, like 口,日,水. Marking a card as easy just to bump its interval is setting it up for an almost certain mistake when it pops up as a mature. It also screws up the algorithm, as it relies on your answers not only for short-term spacing, but long-term as well - i.e. card that get a lot of "easy" scores will be moved up faster, even if you fail them.

Taking a lot of new kanji at once will greatly and exponentially increase your review count, testing your persistence as you add more. I'm glad to hear you can easily keep up your current pace, which, IMO, will certainly lead you to success with the book. If you feel like you can pick up the pace, by all means, do so, it'll not only decrease your total time with the book, but also give newer kanji more time for reviewing. Just don't ever burn yourself out, and keep in mind that it takes at least a week for the review count to go down if you decide to slow down again.

It's not a good idea to mark everything as hard, as that setting hardly increases the intervals at all. If a card took 14 days to show up, marking it as hard, in my experience, would increase the interval to 15 or 16 days, or it may even keep it at 14.
Edited: 2012-03-30, 12:08 pm
Reply
#8
Fadeway Wrote:Never mark cards easy just because you want them to move up. It's also not recommended to call cards easy if you've mistaken them the same day (logic is, you just got it wrong, how can you mark it easy a few minutes later?). I have some cards bouncing between 1 and 15 day intervals, constantly getting mistaken, but they don't go over 15 days for good reason, because I mistake them. A card should only be marked easy when you're sure you won't get it wrong, when it's so easy it's more a question of getting it out of the way without outright suspending it - usually cards like common primitives whose primitive meaning is the same as their standalone one, like 口,日,水. Marking a card as easy just to bump its interval is setting it up for an almost certain mistake when it pops up as a mature. It also screws up the algorithm, as it relies on your answers not only for short-term spacing, but long-term as well.
This.

If you fail a card, then the next time you see it you should mark it as "hard" (or "good" or whatever).
Reply
#9
Hey ellipsis.

I'm no anki nor kanji expert, so I won't guarantee this is all correct. That said, I think I read somewhere, that how quickly the intervals on your cards increase again after you mark a card as failed depends on the 'ease' of the card in question, which is determined by how often you mark the card 'fail', 'hard', 'good' and 'easy'. You said that you often feel the need to press 'easy' the next few times after missing a failed card, right? I suspect this may be a bad idea as it could increase the 'ease' of the card causing the intervals to jump up again faster than what your memory is capable of handling. I'm not sure though. Since part of the problem seems to be that you want the interval of your cards to increase rapidly again after you mark them failed you could try not reducing them to zero in the first place. This is done by going to deck properties -> advanced and changing the 'Button 1 multiplier'. I don't know if this is smart (it's 0% by default), but if you want to avoid reducing the intervals to zero each time without messing with the 'ease' it's an option. I'm afraid that's all I've got to say.
Reply
#10
BTW, the Anki 2 beta is out now. It has settings that will fine-tune the spaced repetition algorithm so that you get closer to the kind of fail percentage you want to achieve. (i.e., if you're failing too many cards, your intervals will get shorter so you fail fewer). I haven't played around with it too much but it sounds like it could be helpful.
Reply
#11
Those are great ideas. I guess I just assumed that hard/good/easy only effected the next gap, not future ones down the line. I think I will change my button one from x0 to x0.05 or something similar. So in that way if I make a mistake on a mature 5-month-gap card it will then be reset to a 1 week gap. I didn't often mis-mark things but if I misread or misunderstood a mature card and therefore got it wrong I would mark it "fail" and then find it irritating as it kept popping up over the next few days. To the point where I'd "bump it up" quickly.
Oh and I'm going to go try Anki 2 (after backing up my deck) it sounds cool.

Thanks for the feedback. Your long detailed responses are nice to read. I kinda feared I'd just get messages along the lines of "you suck, learn French instead". But the responses I've got are way better. Thank you. I'll go try this stuff and maybe get back to you on how it worked.
Reply