Wanikani Data Analysis

Hi everyone!
Long-time lurker, first time poster.
I was thinking the other day about all the user-data that Tofugu has, and what sort of interesting things they might find by analyzing it. For instance:

  • Users often ask about whether these should reset when they fall behind. WK could probably look at all the profiles that accumulated >1000 and see of those who reset, what fraction remained active compared to those who didn’t reset but attempted to chip away at their reviews.

  • It feels like most of my mistakes are things like confusing しょ with しょう, or じゅ with じゅう. What fraction of user mistakes are of this sort?

  • What kanji meanings are most often misattributed to other kanji?

  • How often is an item is on the verge being burned get forgotten? When an item but makes a mistake, the second time they are in that same situation with the same item what is their success rate?

I could go one with questions about user differences based on geography, pacing, participation in the forums, etc. You get the idea. I suppose this post is primarily directed at Tofugu and company, requesting a blog post or two addressing this, but I am also curious to hear from the community what sort of questions they would find interesting. (Also, maybe this sort of analysis already exists and I don’t know about it.)

5 Likes

one of the staff members recently mentionned a few things about common errors, but i don’t recall the details, not the post it was in.

transitivity pairs were amongst the most frequent errors.

some more data would be nice, but most of what i’ve seen was mined by users…

3 Likes

They do not have this data. They only know whether you passed or failed the review, not what you answered

6 Likes

A fun stats project any WK user could do is to calculate the total number of lessons/reviews/etc. done by all WaniKani users each day/month/etc. When I was looking at the API, it looked like for each object type (e.g., an individual review / lesson / etc.) used by WaniKani has a unique ID number that is generated by adding one each time a new review / lesson / etc. is done by any user (with separate counters for each object type). So I think if you want to calculate say the total number of reviews done by all WaniKani users in a month, you could just look at two IDs for reviews you have done spaced a month apart and take the difference and that should be the answer.

You could also get some interesting info like total level ups per month, total resets per month, etc.

1 Like