So I generally get between 70% and 95% or so on my reviews, depending on what I’m reviewing and when I learned it. Burns tend to be closer to 60%.
It feels to me like as long as I’m burning things and slowly studying new vocab, I’m making progress. But there’s also a fair amount of churn as burns or other reviews get missed, and if we project out over the couple of years it will likely take me to get through 60 levels, some of those burns will have dropped out by then as well. The more years that is, the more burns disappear.
So: what average percentage should you ideally be hitting to make real progress? And what percentage means that you’re actually just treading water, not really getting any better?
Any and all answers are welcome, from wonky statistical analyses to wild-ass guesses based on limited experience. But we don’t just need another “how many do you get right?” thread because there are a zillion of those.
Also, no advice on starting reading to cement gains please. I know already.
I assume you’re talking about the percentage right shown after reviews on the summary page. The one shown during the review session doesn’t mean much.
The short answer is that roughly 67% right means you’re standing still. It’s an oversimplification, but getting a review right means the item goes up one SRS level, while getting a review wrong usually means it goes down two levels.
It depends i guess.
Especially for newer lessons i think the corect rate should be quite high, since you’ve recently been busy with remembering those and creating mnemonics. So if you have just learned a few batches and get only 67% right, you might have a net plus, but it feels to me your learning is inefficient. During these stages it’s not about getting it right, but planting the seeds to remember when the item is in guru or master or after burning.
It’s a bit vague perhaps, but you might say that becoming adept at internalizing kanji and vocabulary is the real progress. Post-60, you’ll still be learning new words. How will you remember those?
[quote=“cluffas, post:4, topic:50555, full:true”]
So we can give letter grades just to stress ourselves out:
S = 94%-100%
F = 0-94%
[/quote] Sorry, I’m too much of a perfectionist.
I would like to add to this that that’s 67% average over a longer period of time. A single review session with a score below that can happen, but doesn’t mean you’re not progressing as seen over a longer period.
I’m a firm believer that no one has yet managed to actually move consistently backwards or even stand still. At a given point the cycling items come back so many times that inevitably the accuracy goes up.
It often obscured though because most of us are continuously throwing new items into the mix by lessoning.
So… your theory post to me is the correct terminology. The only way to stand still is to not do the reviews at all.
for items which are in apprentice, they only drop down 1 level, so one needs more than 50% right on those. for items in guru and above (i think) they drop down 2 levels, so it’s 67%. so assuming that your items are distributed evenly over the 8 levels, and your errors are distributed evenly as well, you’d need about 58% right
however, i’d assume that the errors aren’t distributed evenly. i’d assume one would make most errors at the lowest levels, where one doesn’t know an item yet. and because items at higher levels drop down 2 levels instead of one, there’d tend to be more apprentice items than guru or higher. so your minimum required would be slightly below that 58%
and additionally, because burned items can’t drop down anymore, you would eventually get items burned even by pure chance. if you were getting 50% right, 1 in 2^8 items would get burned without ever being wrong…
i think i like playing with numbers too much
(edit: don’t take this seriously! don’t try to beat wanikani using infinite monkeys! )
Hmm? What? I’m just giving the math in terms of what review percentage results in a higher review load. It’s not theory, it’s math (even if simplified math). If, for example, you consistently averaged 50%, your workload would increase, even if you never did another lesson. That’s not up for debate. I wasn’t trying to make any claim about learning as a whole, just review load.
I agree with the part you said that this is over a longer period of time. It’s a game of averages after all. Everyone is bound to score below 67% in a review session from time to time. I know I certainly did. I thought it was obvious that I was talking about long term averages, but maybe I should have been clearer.