[Web App] WaniKani History + Timemachine

Nobody had a WK account in 1970 (for obvious reasons known as the non-existence of the Internet) if it’s showing 19111 days for you. :stuck_out_tongue_winking_eye:

1 Like

There’s something else. I’ve never done a lesson, and had this problem regardless of reset status.

I’m happy to see my projection for Level 60 of 554235 days, which is Monday, October 9, 3539, which is considered infinitely far by these forums!

1 Like

Aww, now I can’t use this as an excuse not to do lessons! Thanks

1 Like

Sorry for the long wait! I’ve been working on improving this website the past few days. Actually I have overhauled many aspects to increase usability and speed. I’m going to update the website later today. Until then I will try to fix as many bugs as I can!

2 Likes

While you’re working on minor bugs, the data points on the graph for me are at midnight UTC. However, I’m in time zone UTC-5 so the “end-of-day time” is the inconvenient 19:00, especially since I do reviews sometimes at this time!

It might be a good idea to make that “end-of-day time” configurable anyways rather than imposing midnight on everyone, but it’s not unheard of for me to do reviews at the 19:00 hour, so I would prefer that not be the time for me :stuck_out_tongue_winking_eye:

1 Like

@saraqael Congratulations on the updates! The site seems to be much improved!

However, my item types counts are still a little off (my key is still c5e0edc2-fb8e-48a7-abae-f79b4ecfcdb4)

Apprentice: actually 182 (site says 76 for the 28th)
Guru: actually 616 (site says 616 for the 28th)
Master: actually 510 (site says 511 for the 28th)
Enlightened: actually 1408 (site says 1411 for the 28th)
Burned: actually 3021 (site says 3133 for the 28th)

The Guru / Master / Enlightened counts look like yesterday’s correct counts, although I don’t understand why I can’t see today’s anymore.

Interestingly, 3133 - 3021 = 112 and 182 - 76 = 106 (super close again!)

1 Like

Yeah, I noticed. I tried for a long time but I don’t seem to find the error today. I’m definitely going to try again tomorrow! I found that the amounts that “Apprentice” has too few “Burned” has too much. There might be a problem with the reset still or something else?

Anyway, I just wanted to say that your account is literally perfect for testing purposes! It has resets (even one with an edge case 1 → 1) and a long time span to work with.

Also, thank you so much for continuing to support my endeavor. There are probably many who could solve the problems better or more quickly than I can but you still always try to help me build this website and sink a lot of time in yourself! I also noticed that you “advertised” the Web App in another thread. I just wanted to say that I really appreciate you helping me so much!

And also everyone else here who always write encouraging comments: Thanks :smiling_face:

P.S. When I’m done with bug fixes I’m going to write a formal update of every new feature and add them to the original post.

1 Like

The 1 → 1 reset was because I knew enough about WK that when I came back in April 2019 (after a brief stunt in June 2018) that resetting would make things better for statistics (didn’t want it to say I took 250 days on level 1 when I wasn’t even thinking about WK most of that time). It also forced the new post-Overhaul mnemonics as a side benefit. :wink:

I went to WK Timemachine (with the expired certificate) and grabbed some photos to help you debug:

WK Timemachine Photos




@LeoG-B Your level progression chart seems to be fixed, you can continue your lessons now! Or did you want me to take a little longer so that you don’t have to do your lessons yet? :grin:

1 Like

Haha, I’ll pretend I didn’t see this so I can leave it a while.

Thanks very much, great work on the site. So useful

2 Likes

By the way, what are you doing exactly with the median calculation? I presume you’re including reset levels in it which is why it’s different than WKStats, but are you including the user’s in-progress current level?

I decided to take a closer look at @jprsjsalkfdlasjd’s token. I think the wall of shame / fame & review info look OK for not having any reviews, although you should probably handle the 0-review case more gracefully to avoid the error thrown in the console :wink:

However, something is clearly wrong with whatever you’re providing in the “word info” section, and level progression is missing its data! Before, there was ~250 days on 1R and a short time on 1, but now they’re both showing as 0, as well as derived calculations like the median.

P.S. In my opinion, the current level probably shouldn’t be factored into the median calculation, but it doesn’t seem unreasonable that all previous levels, even 1R, could be included. It wouldn’t take very long for a real user to work that off.

P.P.S. A user can reset multiple times… there’s no reason this token couldn’t have in the future yet another 1 → 1 reset to have yet another level 1 to put on the chart!

P.P.P.S. Sorry for all the feedback; this site is pretty amazing & I’d love to help you try to figure out the remaining kinks!

1 Like

By the way, can you remove the forced 0 as minimum on the item types graphs and/or remove the option to start with 0?

It doesn’t make sense to have start with 0 as an option if the amount of items in an SRS category is decreasing over an interval of time & not be able to show negative values (e.g. -50 Apprentice items doesn’t make sense in reality, but 50 less Apprentice items on May 1st than March 1st absolutely does!)

P.S. I saw you worked on implementing review caching! Keep up the good work! :muscle:

1 Like

Im quite enjoying this stat page, thanks! although accuracy is currently broken for me (havent seen anyone else mention it yet)
image

It actually starts to go wild at the same time as my first burn reviews appeared (no clue whats that second irregularity), so it might have something to do with that - perhaps number of active items going down by burning not being represented? I would expect even the non 100% parts to be lower, my accuracy has went down sine the beginning.

example accuracy graph from other source

(very smoothened out version)

i assume its some known issue, but if you are interested in my very boring api key (no resets, no extremes), let me know

1 Like

Thank you so much as always for pointing stuff out to me that I didn’t even realize! Without you I’d probably miss half of this stuff.

Here’s what I’m going to work on:

  • fully implement caching
  • remove forced 0 on SRS item graphs
  • “no review data” bug fixing
  • median calculation for resets (and not factor in current level, although median is always a value in the dataset so really high values/anomalies don’t make much of a difference; that’s why I calculated both the average and median; but at least remove it for the average)
  • fix the SRS item counts for resets (still a problem :melting_face:)

I really appreciate the input and hope to make this website a useful as possible! So, into the bug fixing I go.

1 Like

Thank you very much for telling me. I’m glad you’re enjoying the website!

This graph actually shows the accuracy on a specific day (the items that were last reviewed at that day count to this statistic) not the average until that point. But there still might be some error in there. I’ll have to take a closer look at that.

Either way it would probably still be more useful - like you said - to have an average accuracy over time graph. Maybe I could just add that one as well, but that would maybe cramp that section a little. So I might just exchange the current chart for the one you posted.

1 Like

Thanks for looking at this - there definitely has to be some error there, i wish i was good enough to have 100% accuracy for full month hahaha. I think this daily stats are good and there isnt any “average up till that day” needed - my graph is also just daily result, albeit weighted in 30day average (something like youve done with checkbox - Smoothing for “Reviews Per Day” Chart). I still think those burn reviews are somehow incorrectly calculated.

1 Like

You could do a smoothing option like with the reviews / day graph :stuck_out_tongue:

1 Like

I love the new scatterplot of reading/meaning percent correct answers! (Was that English? :thinking:) Also very humbling to look at my wall of shame and think wait, that kanji exists? What IS it?? :joy:

2 Likes

Thank you! I originally wanted to make the same thing but like a bubble chart (like you sometimes see with countries) but there are just too many words in WaniKani you literally couldn’t see anything. Something like this:

But I think that you don’t really need to change the size of the bubbles. Currently the WK level the word appears in is not represented by the size anymore but by how dark/bright the color is.

Honestly… same. There are some Kanji to which I just think “I swear that I have never such a thing in my entire life!” :joy:

But yeah, I should stop rambling. If you’ve got more ideas for charts you would find interesting you can tell me. I’m always open to new ideas!

2 Likes

Yeah, you’ll see leech clouds on the forums sometimes. That would be cool but it’s just icing on the cake! :wink:

1 Like