Question regarding speed

Levels 1-45 averaged 8d16h (excluding the two long ones where I was on vacation)
Levels 46-59 averaged 4d21h

2 Likes

Well that leaves me pretty confused.

After a hiatus on level 2 (only level with untrue average time) I returned to the site, and started by ramping daily items from 5, to 10, to 15 (and now I’m doing 20). But somehow I spent pretty much the same time on level 4 as I did on 3 (which is not 8-9-10 days), which is kinda confusing, since on level 3 I was going with quite a lower average number of new items per day than for level 4, even accounting for exception days on both levels where I didn’t do new lessons, only reviews.

So I’m kind of baffled by this, Does sticking to a perfect routine lesson-4hours-8hours-12hours review really make up for an almost double difference? (I was doing lessons before bed, first review first thing in the morning if I wasn’t running late, and then if I got the chance at work, but mostly whatever came up as reviews after I got home, 97% average accuracy).

That’s kinda demotivating. I simply don’t have a stretch in which to do a lesson-review in exactly 4 hours routine. Or really do an actual schedule to which I can stick really. At least not one that would have a semblance of efficiency it seems.

I’m looking through your data to see if I can spot the reason for what you describe.
What confuses me is that it looks like you started unlocking Level 3 kanji at the same time as Level 4 kanji (April 11).

Your Level 3 kanji unlocks are spread from Apr 11 to May 7, about half of them on May 7.
Your Level 4 kanji unlocks are spread from Apr 11 to May 26, about 2/3 of them on May 25-26.

Maybe Wanikani’s meaning for ‘unlocked_date’ has changed. I’ll have to look into that.
Would you be willing to send me your APIv2 key, since APIv2 contains information about when reviews actually took place, whereas APIv1 (which the stats site currently uses) only has ‘unlocked_date’?

You can send your APIv2 key at https://www.wkstats.com/help

The main point is not to get kanji or radical reviews wrong, doing that on Apprentice 3 or 4 easily cost you days (and ~3 kanji or 1–2 radicals suffice).

True, but they said 97% accuracy, unless I’m misreading.

You don’t need much for a delay, I fat-fingered two kanji recently and lost two days :slight_smile:

“I needed twice the time but am quite accurate” indicates that leveling up is blocked by only a few items.

Radicals: — 99.14% 99.14%
Kanji: 97.30% 97.52% 97.41%

I was actually confused about new level unlocks as well. From what I gather, it seems that you get the new level as soon as you guru 90% of all previous level kanji? So you can literally skip all the vocab if you liked? I guess maybe this is how people do the crazy speeds per level, or maybe they do all their vocab at 1 or max 2 sittings (guess this is what people mean by doing the vocab while doing the kanji with userscripts, and why people get exponentially bigger review numbers, instead of evening out as they are supposed to if done more spaced out). But if the math is correct, wouldn’t going for the average (like lets say 18 in your case) still give you the same result in the end? Because it seems if you stick to something like 15/18 per day, you would get delayed by at least a few days per level (aif going in the order the lessons come in vanilla WK radicals->kanji->vocab then next level items), so something doesn’t add up.

I don’t really know why my level 3 and level 4 kanji unlocked at the same time though. I also noticed this and thought it was strange, but assumed that’s just how WK works.

Again, am I just making a derp somewhere?

Yeah, the level-up requirement is just 90% of kanji Guru’d, and vocab doesn’t matter. But for normal use (without scripts), vocab gets done anyway because it’s mixed in with the rest of your reviews.

What really matters for level-up is:

  • A kanji unlocks when all of its radicals reach Guru
  • You level up when 90% of your kanji reach Guru.

Typically, there are about 30-35 kanji per level, so you can miss up to about 3 kanji (10%) without delaying your level-up speed. But, if you miss 1 radical that happens to be used in 4 kanji, it would delay those 4 kanji from being unlocked, thus delaying you reaching 90% of your kanji being Guru’d. If you repeatedly miss that same radical, it could set you back several days (possibly several times).

Missing one kanji repeatedly doesn’t matter as much since that only prevents unlocking vocab, and vocab don’t affect level-up.

I’ve already done more reviews than you ever did! :confounded:

5 Likes

OK, so here is another question.

Assuming that this is why my average per level stats are slower than what the math would suggest, is this something that only matters for this particular stat? What I mean is, even if you mess up a radical/kanji at Apprentice 3/4, with a pace of 15-20, you always have new lessons to do every day and of course reviews regardless of that. You won’t really be noticing an effect as you are never going to reach a point where you are waiting for a long time without any lessons, because even if you mess up, by the time you reach the next batch of radicals, a lot of time has passed since you’ve been doing the 100+ vocab at 15-20 per day speed.

Then in the grand scheme of things, with the doubled rate of per level stat because of this, will you still reach the main goal of 60 in x years (based on the above math for 15/20 new items per day)? How does all the above change the formula then? It gets doubled, just like the level up time? This also doesn’t seem quite right based on my assumptions when reading forum posts.

If I understand you correctly…

It’s not that simple. Lessons-per-day is directly correlated to level-up speed. If you take 16 days per level, you’re only unlocking lessons at half the speed of someone taking 8 days per level. So, your lessons-per-day would be capped at 9 or 10 per day (avg).

In other words (assuming you’re taking 16 or so days per level), if you did 18 lessons per day, you’d run out of lessons after about 7 or 8 days, then you’d only have reviews for the next 8 days until you level up.

At 15 new items per day I think I only got once as low as 3 lessons remaining, but that number bumped back up healthily in the evening of the same day.

Though I might be misremembering. Or somehow I’m mistaken about something, because I can’t figure how all these things add up to what I got.

I mean its basic math that average number of new items per day would be directly correlated with average level up speed, but considering that 3 mistakes at the wrong item and time can DOUBLE that time, something seems off… Wouldn’t then average time for getting to 60 not be 2 years for the average Joe that is not speedrunning, but more like 2.5-3-3.5?

With your APIv2 key, I could tell you exactly why you took as long as you did on Levels 3 and 4.
In fact, that’s one feature I really want to build into the new stats site.
If I have time, that is.

4 mistakes (not 3) won’t double the level up time, but it will block you from leveling up for 3 more days. And you won’t get any new lessons while you’re waiting to Guru those kanji.

Well I’m either misremembering, I’m confused about something, or considering I had a healthy amount of reviews at all time, then I guess this is not what happened in my case… Which doesn’t answer the question.

I liked to do 10-20 lessons a day. At that time, I was averaging 12-14 day levels.
I have since slowed WAAAAAAAAAAAAAAAAAAAAAAAAAY down.

I don’t know what you expect out of this thread exactly, as the answer mainly depends on how fast and accurate you can remember the items. The reviews are just a “mechanical” component, I’m doing 5 seconds per answer now, so ~1 hour per day is quite accurate for me as well. I don’t really spend much time on lessons as well, if you know the kanji meanings and readings well you can “easily” come up with most of the vocab answers during reviews. But especially in the beginning you need good time to “learn learning kanji”.

I would say there are three phases in WK. In Level 1–15 you need a lot of time during lessons because everything is new and all kanji use new parts, in 15–30 you can leverage the previous levels and you have something like a “kanji eye” to go faster through lessons and reviews, and 30–60 where in depends on leveraging stuff like phonetic components to go really fast through the lessons, and not messing up similar kanji in the process.

If you are time-limited you should do lessons (fill the apprentice set) until you are approaching your daily time limit (it may be a bit “spiky” with larger review sessions on some days), and stick to that. Over time you can increase your review items because you get better at learning kanji.

Another thing to take into consideration is that the last 10 levels done at max speed can act as a bias to the average of new items/day. You’re easily learning 40+ items a day during that phase.

@acm2010 omg you’re so close to meeee! Great to have you join the lvl 60 reign with me and @GojaButt :slight_smile:

Happy one year WK as well! You were catching up too fast, I didn’t take “going fast” too seriously :wink:

We will all have much more time again starting from the next week …

1 Like

I think probably the number one disconnect that may be occurring the is the issue of 90-95% accuracy. Unless you are getting the same kanji wrong over and over, 90% accuracy in effect will slow you down because your Guru rate of kanji is unlikely to reach 90%. If you get 10% of the total kanji wrong in one review, then get a different 10% wrong in a second review, you’ve now dropped 20% of your kanji’s SRS levels. This could result in a cycling problem that leaves you with constant reviews and eventually no more lessons.

90% accuracy on radicals is an even large problem, especially on low levels, where radicals can number over 20. Those radicals are required to unlock more kanji which contribute to the 90%. As, except in level 26 and 40-something onward, the non-radical kanji pool never reaches 90%, you have to have high and/or consistent accuracy on radicals in order to progress normally. A similar accuracy spiral as described above with kanji could result in blocking enough kanji that even assuming you had 100% kanji accuracy, you would still be unable to level up.

Personally, if you are going for a time-efficient approach to WaniKani, this is my recommendation:

Upon leveling up, only do lessons on the radicals and vocabulary (this can be accomplished more easily using a reorder script). Because most early levels prevent you from advancing until you unlock the kanji behind radicals, there is no real reason to review the immediate kanji if you’re going for time efficiency. Sure, it’s nice to have the safety net of knowing you have 80%+ of the kanji done when you unlock the last kanji, but it diverges your mental resources and splits the vocabulary more.

By doing this, your vocabulary management becomes cleaner. I (unlike what it sounds like you intend to do) perform all of my lessons in bulk, so, after leveling up, I complete the lessons for all of my radicals and vocabulary. By the time I unlock the final kanji for the level, my vocabulary is Guru’d and out of my regular review schedule. This means that my only Apprentice items are now the full list of kanji for the level. This is all assuming 100% accuracy, of course, but any vocabulary items missed with now be priority items as the only apprentice items among a blob of kanji.

Every time you level up, you unlock not only more radicals and kanji, but more vocabulary. By restricting yourself from reviewing the initial kanji upon leveling up, you cycle the new vocabulary from the same level kanji to the point of leveling up again. This focuses your reviews more. Of course, by doing 20 lessons a day, you’ll spread everything out more, but the idea is still there.

In short terms:
Level Start: Radicals and Vocab lessons.
Mid-Level: Radicals and Vocab Guru’d, all Kanji lessons
New Level: Kanji Guru’d, Radicals and Vocab lessons

The biggest concern I’ve always seen with incrementally doing lessons (and anyone who actually does this may rebuke me at any time), especially with the assumption of 90% accuracy, is that your SRS load is going to become excessively hard to manage. Doing small amounts of lessons at a time may seem easier, but, ultimately, you’re going to miss review times and they’ll bulk up, especially if you assume lower accuracy.

Day 0:

  • 20 Lessons [8 AM]
  • 4 hour review: 18 correct (Apprentice 2), 2 wrong (Apprentice 1) [12 PM]
  • 4 hour review: 2 correct (Apprentice 2) [4 PM]
  • 8 hour review 16 correct (Apprentice 3), 2 wrong (Apprentice 1) [8 PM]
  • 4/8 hour review: 4 correct (Apprentice 2/3) [12 AM]

Day 1:

  • 20 Lessons [8 AM]
  • 8 hour review: 2 correct (Apprentice 3) [8 AM]
  • 4 hour review: 18 correct (Apprentice 2), 2 wrong (Apprentice 1) [12 PM]
  • 4 hour review: 2 correct (Apprentice 2) [4 PM]
  • 8 hour/1 day review: 30 correct (Apprentice 4), 3-4 wrong (Apprentice 1/2) [8 PM]
  • 8 hour/1 day review: 4 correct (Apprentice 3/4) [12 AM]

Maybe I’m overthinking things, but it just seems like it takes more time to manage than bulk lessons, and assuming 90% accuracy makes things even more messy. I follow a reorganized variant of @jprspereira’s tactics described above, and WK doesn’t take up much time for me, especially compared to the time I spend on other activities to stretch my Japanese.

1 Like