+- kanji koohii FORUM (http://forum.koohii.com)
+-- Forum: Learning Japanese (http://forum.koohii.com/forum-4.html)
+--- Forum: Remembering the Kanji (http://forum.koohii.com/forum-7.html)
+--- Thread: Optimal failure rate (/thread-271.html)
Optimal failure rate - laxxy - 2006-11-18
Surely everyone reviews differently, but I wonder what people think is the best failure rate for a stack of cards, when does one learn faster?
When a system is designed with few reviews and large intervals, one forgets cards between reviews, fails most of them and learning is inefficient. On the other hand, if a system produces too many reviews, not increasing the review period fast enough, one will get close to 100% of the cards every time (because the well-learned cards clog the system) which would also mean that the learning is inefficient (imagine a system with just one box of cards as an extreme example).
My feeling is that something around 90% (perhaps ~95% for the well-learned cards, like those in higher stacks with a Leitner system) should be about right, but I wonder what other people think about this.
Optimal failure rate - Pangolin - 2006-11-18
Funnily enough, this evening I found that I had 230 kanji to review, mostly in the last stack which included the very earliest frames (eg 一 to 十) I just did 137 of those and had a 95% score. I generally get between 80 and 95% except on very recently learned kanji and find the review rate very manageable. I only got this vast amount today because these are the ones I banged in en masse when I first started reviewing here. It seems like a long time since I had seen many of these up for review, and I was starting to worry that I might have forgotten a lot of them, so I was pleased with the 95% result. This seems to suggest to me that the system as I'm using it is working fairly well. Is this the sort of thing you are thinking about, or have I misunderstood your point?
Optimal failure rate - synewave - 2006-11-19
When I first started using the site I reviewed far too heavily to get everything to the right as soon as possible and continued to review 100 from the final stack daily as well as adding 15-20 new cards Monday to Friday.
I started to get bored doing this coz I was remembering almost 100% from my final stack reviews. As laxxy suggests, I'm pretty sure this was due to the frequency of review.
To get back into the Leitner way of things, I asked Fabrice to remove all my cards so I could start again. Two benefits of this were I could start using the site as intended. This did mean that I had loads of reviews to get through before cards were back in the 4th pile again. However I'm just going through my first 1 month reviews now and I'm doing much better than I had feared. On average I reckon I'm at least over 90% successful from the final stack.
The second benefit to having all the cards deleted was that when I added them all back together, they were randomised.
So in my experience, using the site as designed utilizing the spaced repetition is more efficient time-wise and, I hope, better at forcing the kanji into long term memory.
Optimal failure rate - CharleyGarrett - 2006-11-20
My intuition also suggests that 90% is an optimal failure rate.
Optimal failure rate - mspertus - 2006-11-20
CharleyGarrett Wrote:My intuition also suggests that 90% is an optimal failure rate.Phew! I was worried that I was in trouble because I was failing 90% of my flashcards
Optimal failure rate - leosmith - 2006-11-20
mspertus Wrote:Phew! I was worried that I was in trouble because I was failing 90% of my flashcardsWise guy!:lol:
Somewhat related, supermemo suggests, and defaults to, a 90% success rate.
Optimal failure rate - Pangolin - 2006-11-20
leosmith Wrote:I'm using SuperMemo and this site in parallel, adding the same kanji at the same rate on each. (Will I go blind?) I don't think I'm observant enough to tell the precise difference, but it seems that with SM reviews are much more unpredictable (to me) and spaced out. I think this may become more apparent as time goes by. It's interesting, and somewhat scary, to look at the calendar that SM makes for you, which in my case stretches well into 2007.mspertus Wrote:Phew! I was worried that I was in trouble because I was failing 90% of my flashcardsWise guy!:lol:
Somewhat related, supermemo suggests, and defaults to, a 90% success rate.
Optimal failure rate - laxxy - 2006-11-20
Pangolin Wrote:i used to use Twinkle and this site, and switched to Twinkle alone after some time.leosmith Wrote:I'm using SuperMemo and this site in parallel, adding the same kanji at the same rate on each. (Will I go blind?) I don't think I'm observant enough to tell the precise difference, but it seems that with SM reviews are much more unpredictable (to me) and spaced out. I think this may become more apparent as time goes by. It's interesting, and somewhat scary, to look at the calendar that SM makes for you, which in my case stretches well into 2007.mspertus Wrote:Phew! I was worried that I was in trouble because I was failing 90% of my flashcardsWise guy!:lol:
Somewhat related, supermemo suggests, and defaults to, a 90% success rate.
But I do have a couple issues with the Twinkle spacing algorithm, nothing major, but maybe things that could be improved. This question is related to that.
looks like everybody understood what I had in mind, but I agree that the title was confusing, sorry -- it should be something like "What recall rates should an optimally designed spacing algorithm induce", it sounded kinda too long though.
Optimal failure rate - PepeSeco - 2006-11-20
By the way, the topic seems related to KANJI's question regarding optimal learning schedules...
Optimal failure rate - KANJI - 2006-11-22
Some of you expressed relief about the fact that some failure is experienced by others. I'm relieved too. When I had my first go around with RTK a decade ago, I was really disappointed with a 10% failure rate, ha, ha.
Since starting again, I got the idea from this site that of course some failure is expected. But I salute laxxy for starting this thread specifically on this.
PepeSeco Wrote:By the way, the topic seems related to KANJI's question regarding optimal learning schedules....You're right. This thread is yielding additional benefit in hearing about different study strategies and their results.
synewave Wrote:The second benefit to having all the cards deleted was that when I added them all back together, they were randomised.Randomized? How so? Isn't that the case already?
Optimal failure rate - PepeSeco - 2006-11-23
KANJI Wrote:You can probably set up a formula with the following variables to calculate a schedule:PepeSeco Wrote:By the way, the topic seems related to KANJI's question regarding optimal learning schedules...You're right. This thread is yielding additional benefit in hearing about different study strategies and their results.
- total number of kanji to be learnt
- number of kanji per day (or other time unit) that can be added
- failure rate (mine is about 13%)
- number of review piles
- spacing of the reviews in time units
I think this is more or less what you wanted to understand when you started the other thread, or?
Optimal failure rate - mspertus - 2006-11-24
Just for reference, I've been tracking my success rate since Oct. 9 (a few weeks after entering all the RTK1 kanji), and it's averaged 84%. I think I'll eventually get up to 90%, but it won't be soon.
Mike
Optimal failure rate - KANJI - 2006-11-24
mspertus Wrote:Just for reference, I've been tracking my success rate since Oct. 9 (a few weeks after entering all the RTK1 kanji), and it's averaged 84%. I think I'll eventually get up to 90%, but it won't be soon.Hey, buddy, if you're into the thousands, take the 84% rate and run with it!
Mike