Guru 1 interval too long?

Does anyone else think the Guru 1 interval (1 week) is too long? I’ve long felt that Guru 1 is the hardest review stage (apart from Apprentice 1 of course), but after doing some calculations using the WK api, I now have the numbers to prove it.

For vocabulary, my accuracy rates on Apprentice 4, Guru 1, and Guru 2 are currently 85.8%, 71.7%, and 80% respectively. I knew it felt like I was missing a lot of Guru reviews, but it was still surprising to see such a sharp drop.

For kanji, they are 97.9%, 90.5%, and 93.2%, which also shows a big drop at Guru 1, although the Kanji numbers are biased somewhat because Apprentice reviews are required to level up so I try a lot harder on them.

P.S. These numbers may also be biased by the fact that I’ve only been on WK for a month so the only vocabulary coming up for Guru 2 reviews are the relatively easy ones from the early levels. Perhaps I should wait another month and redo the calculations to get more convincing numbers.


Not really. The hardest stage for me has been Enlightened to Burned unless it’s just a really easy word that I encounter all the time. The vast majority of my leeches are items that dropped from Enlightened. I’ve seen very few leeches due to Guru I review. :man_shrugging:


Completely agree with enlighten to burn being most difficult… and not even for like scratching my head on what a word’s meaning or reading are (though there are certainly some that I will trip up and think a reading is しょう instead of しょ or something).

I feel like by the time you get to a burn review, it’s been so long since you’ve seen the word, and in English you might know a half dozen synonyms for the meaning. But if it isn’t on the list or a synonym you entered yourself, even if you know the meaning… you’re boned.


i dont get it. why do you think that way ?

1 Like

I feel like it’s the hardest for me, too. I have the most leeches in Guru 1 by far. Would you be willing to share the stats tool you used to check the numbers? Would love to see mine.


what’s a leech ? :face_with_raised_eyebrow:

1 Like

Leech is a forum slang for an item that you always always get wrong… and it keeps popping up again and again sucking away your energy and time.

As for me Enlightened → Burned is the hardest.


Same here. I find that for more “obscure” words, if I do end up remembering it and getting it correct, it took too long for me to remember that I end up unburning it anyway and starting over.

1 Like

Yes, I agree! I’ve been using Kitsun to learn vocab lately, and the default intervals are the same as WaniKani’s. But since they can be customized I changed 7 days to 5 and 14 days to 10.


I think there will always be a gap between what you can have in your short-term memory and what makes it into long-term memory. At the border SRS is more likely to have you repeat the items.

1 Like

You might be right. I’ve only been doing WK for a month, so I haven’t gotten to Enlightened or Burned yet.

I did it by querying the WK api by hand, though I do hope to turn it into a proper tool and publish it as a website like wkstats.


I cheat in these cases. I open another Window, paste whatever is on screen on the WK search. To me, it’s not about remembering the exact word. If I get a synonym (not provided by WK or manually added by me) or that “feeling” that I know the Kanji or Vocab, but can’t put into words, then I allow it pass. If I was off, I fail it.


i believe wk could do with a few more reviews, and i think it’s learning process isn’t enough either. it could have you drill the items a number of times before it lets you off the hook, to ensure you actually learned the stuff and didn’t just take a look at it.

this has been a pet peeve of mine for a long time.


Yes absolutely! No, wait, abort. No, Guru 1 is not the one that feels difficult to me. On the other hand Apprentice 2—> 3? Don’t ask… Master ----> Enlightened? That’s not good either.
I think you reached the right conclusion, but I don’t agree with what you are saying, i.e. that WK should redefine that time interval just because it’s harder for you. Different brains work differently.

I agree that our mighty Crabgator should do something. I believe Tofugu has three options:

  1. Do what you did and see which stage has the lowest percentages on average, i.e. they should do what you did with all WK users [I think they can do that (legally), and it would improve the product much more than a redesign of the dashboard…unless they’ve done it already. I don’t work at Tofugu so who knows…]
  2. Give users the option to customize the experience i.e. to change the time intervals. Kitsun seems to give that option, why WK shouldn’t? [Is it hard to implement??]
    [Personally I would give the users the ability to change the intervals but with one condition i.e. the sum of all intervals should be the same for all users. That way no one would choose the fastest route, since they all take the same amount of time sorry coelecanths… ]
  3. Ignore the issue. They offer a service and they have no obligation to further improve WK.

I was wondering how you got those results too. If we have numbers to back it up, maybe we can petition for a change of the time intervals. I’ll be waiting for that tool! I’ve always been curious about it and I too believe there’s margin for improvement in the way WK has defined those intervals.


You could just use the userscripts (Wanikani Double Check or Wanikani Override). If you just inputed a synonym, you can just override it, add the synonym and mark it correct using one of the default answers. No need to waste time with that long process (unless you are worried you can’t trust your own judgment with the scripts and you would abuse them).

Also I’m not sure what you mean by “feeling” but the way I consider it, it’s a degraded version of knowing it. Always check on when you are in doubt, and when the doubt is strong mark it wrong, it won’t hurt and if you know it then it’ll go easily the next time.

I’m saying this because I too sometimes am in doubt whether to mark something right or wrong. I decided on a rule: the farther up the SRS process the word is and the harsher the judgment
(i.e. considering the same mistake, if it’s master —> Enlighted I would mark it wrong, if it’s Apprentice 3 —> Apprentice 4 I would mark it right hoping to redefine and realign the meaning). If it’s totally off, obviously no doubt there :sweat_smile:


I don’t use scripts. My WK is as vanilla as it comes, and I like it that way.

As for the “feeling”, well, I can’t explain it. I know it when I feel it. Just like when I read a sentence in Japanese. I know the words and the grammar, but I don’t understand it at first go, but I “feel” the meaning of what I read. Only a slower and more analytical read through brings things into crystal clarity. I don’t always do this though, I just keep reading.

The “feeling” is just the feeling. You feel it or you don’t.


I wonder what the workload graph would look like if there was, say, two more reviews for each item in WK. Also could be a good script for someone to make for people who want further reviews, if thats possible :smiley:


So, I have just been looking at my stats using the api, to see how it compares to yours. I’ve been here long enough to have gotten some items up to enlightened, but I don’t have any burns yet. Still, I can look at the chart a bit further than you can.

For vocabulary, I see a huge dip at Guru I, from ~85% (true for App 2-4) to ~75%, but it continues to drop to 72% at Guru II, then recovers a bit to 75% at Master. Those numbers hurt, tbh, but aren’t really unexpected, knowing my brain as I do (ie, my long term memory generally sucks)

Kanji do better, with 89-91% for Apprentice through Guru I, then dips to 85% for Guru II and Master.

Like you, I see a big dip at around the Guru point, but it’s not the smoking gun you see, since my kanji success continues to Guru II. This might be because I long ago did RTK, and relearning kanji meanings is basically trivial for me. So it might just be that helps me past the hump a bit. Not really sure.

In any case, I’m not really surprised by these numbers. That transition from short term to long term is a pain. I know my leeches cluster at the A4 G1 boundary for vocab especially.

As for changing the WK intervals, while I’d like more flexibility, I’m not sure how to implement it effectively. The cycle times they’ve picked mostly seem to be based upon easy to understand and plan for intervals (days and weeks) rather than any statistical reality. They’re more about being understandable than being rigorous.

This has been interesting to read (for a newbie). I’m surprised to see that at the burned level people are still fighting with しょう instead of しょ. OMG. hahaha… I have such a long way to go, and still I’ll be fighting with that!


I felt like this was the killer zone when I started out but I’ve changed my style to be a bit more ruthless. For the last three levels (eight, nine and ten) I have given myself a zero tolerance policy on wrong answers and it is starting to pay off. In the past if I was 75% sure on a kanji answer I’d look it up to check before typing. I’d know it was one of two similar words but not be sure on the right one and had figured that I didn’t want to slow down my levelling up progress by getting knocked back on words I felt like I knew. I was also guessing readings I had forgotten and quickly checking on a possible renduku or a little tsu in vocabulary items if I wasn’t sure. Now I have stopped the checking and guessing get more apprentices wrong and I am taking a few extra days to level up but I’m not stumbling on guru to master. If I could I would still change the intervals though, they’re too far apart, even if I’m closing the gap on my own little by little.

I haven’t been using the program for long enough to know how I’ll progress from enlightened to burned but I can’t help but be worried. Why there isn’t an option for an additional review at an interval chosen by the user is a mystifying one. Other issues with the site like the quality of the mnemonics and sample sentences don’t harm your Japanese progress the same way the one size fits all review structure can.