My 9th grade pre-calc teacher showed the class this. I never erred again.
1 Like
“Good thread, 10/10” - A statistics nerd (myself)
I was literally going to comment on how the data followed Zipf’s law until I noticed you already did that. I’m glad I’m not the only one that thought that. Is there no way to do a strikethough? Just pretend this whole thing had a strikethrough
That’s true, I hadn’t considered that!
I overlayed a Zipf model and an exponential decay model over the results:
This is where the Zipf model is expressed by y = y_max * (1 / level reached)
and the exponential model is expressed by y = y_max exp(-(level reached))
I know some of my assumptions are probably wrong though. Like you said, Zipf is a power law, and there’s no dependence between the levels. I’m just a physicist, so my statistical analysis knowledge only goes so far (and I’m very tired). Have fun correcting me!
5 Likes
Hey colleague!! (physicist here as well xD)
So… our human eyes really suck at comparing shapes, unless… the shapes are straight lines! That is why, in order to discriminate between power law and exponential law you need to check out, respectively, a log-log plot and a semi-log plot.
Wherever you see a linear trend, that’s the law! In the first post our hero showed us a semi-log plot, and it’s fairly linear, which means it a clear exponential law 
5 Likes
@viet It would be cool to get official statistics at some point. 
11 Likes
I forgot about the log plots (rather embarrassingly)! And yes, discarding the first 3 levels:
Then, taking the log of the number of users:
I just hope none of my professors use WaniKani…
3 Likes
It would be awesome if we could know our level percentile.
I think that’s the right name. Don’t kill me, I know about food, not statistics D:
9 Likes
I have plotted both in log-log and semi-log to see the difference and… Actually the log-log is not doing too bad.
Here is a comparison:
As you can see (also from your plots) the semi-log plot is “more linear” than the log-log.
BUT: the log-log is doing not so bad if you consider only the levels from 20 upward. But those are also the levels with fewer users, so less statistically relevant…
The best way would be to do some chi-squared test
2 Likes
It took me 17 days from creating an account to actually start using WK. At first, I created the account because people told me about this helping learn Kanji. The problem was that I used the unofficial android app so it didn’t come with instructions. Everything seemed so strange… so I opened the app, looked and closed it after 2 minutes. 17 days after, I came back but I don’t remember how xD
So I bet that a lot of those lvl 1 simply don’t know the logic… or maybe it’s because the level 1 is “slow”. Idk.
3 Likes
To be fair, even Japanese people have trouble reading those. I’ve witnesed it, and it’s not pretty.
4 Likes
This is starting to bring back memories! Performing a chi-squared test on levels 4-60 and 20-60, I got the p-values 0.2484 and 0.2519 respectively. As far as the significance of these results goes, I don’t remember that stuff!
2 Likes
I can’t remember what it’s like to not be level 60. 
8 Likes
This was addressed in the original post.
Ah, I’m on my phone and didn’t see the other 15 graphs.
thanks for this, i enjoyed this ALOT. lol…
Please make it happen! 
Considering it’ll take at least half a year to burn the items you learn at lv60, it makes sense for there to be a bunch of still paying lv60s kicking about
5 Likes
Very interesting. I’m surprised more people didn’t give up at level 3 (ie. completed the free levels then quit) and that so many gave up on the first several premium levels (seriously, why do so many people decide they like Wanikani enough to pay for at least a month’s subscription and then quit without completing a single level?). It also sort of makes me want to quit on level 59 just so that I can be the first
3 Likes
@bertoncelj1, this is a great piece of procrastination!
off the top of my head, people quit early because:
- Free account for the first three levels. Many people opened an account and never used it.
- they didn’t know they had to be familiar with hiragana first
- they were annoyed about waiting four hours to be tested on the first bunch of radicals in level one.
- they thought they could learn Japanese with WaniKani alone. They didn’t know they had to learn grammar.
- Ran out of money to pay for subscription.
- Cannot commit the time to use WaniKani.
- Japanese is a lot harder than they expected.
- Unwilling to submit themselves to the ways of the Crabigator.
6 Likes
I am one of them. My level 1 is 1232 days, that is, 3.3 years.
8 Likes
9 . Get distracted watching Giada on Food Network
5 Likes