The API seems to have no problem with the concept of “level 60 completed.” In fact every level you work on has a level number (between 1 and 60) an “unlocked at” date, “started at” date, and finally a “completed at” date.
Yes, I should probably ignore breaks longer than 1 year. I’ll do it later.
Yes, I could use any of a number of estimates: mean, median, and others. For the time being I’m using two exponential moving averages (EMA) but I might include a median or give the user an option as to which function to use.
Oh, TIL! They probably apply the same calculation of “90% Kanji completed”. Interesting that there does not seem to be any visual cue in e.g. the forums - or maybe I just don’t realize that because I’m not there yet…
Yeah, maybe you get a crown or something when you complete the last level. Anyways, thank you for your feedback! I’m glad that other people enjoy this small thing I made.
I think this would be a nice idea. If adding multiple prediction methods isn’t too complex, you could probably provide the user with the option to enable/disable each of them. I think that the most accurate prediction model might vary a bit depending on whose graph you’re plotting, as there are a lot of users that level at various paces.
One possible algorithm that might be useful for users who took breaks is something like mean/median level-up time since the last break. You could detect a break by considering any level ups that took more than a certain threshold a break (either a fixed number or a multiple of the median level up time). That way you’d probably be able to make a guess as to what their post-break level up speed could be without having to wait for the EMA to move past the break period.
The prediction seems to be pretty spot-on (I am hoping that I can finish this year, but I’m not sure whether I can manage, so a forecast between January and March is really realistic right now).
Yes! That’s what I meant by "a way to encourage yourself not to slack off.” If you check the chart in a few weeks and see that the forecast is slipping by, that will put you in the mood to pick up your pace!
This also gives me confidence that the coefficients I chose are realistic. I just need to handle a few more edge cases like @weaverZ’s long pause and such.
I love stats! As usual, my graph is boring though. There’s no difference between the light/darker grey predictions, but that’s probably because I’m getting so close…
I’ve made the forecast computation ignore “pauses” longer than 3 months and I’ve made the chart display gaps whenever there is a pause longer than 6 months (eg. @weaverZ)
Well, I’m at level 12 and I’m already reading manga every day. Of course I need to stop every few words to look up in the dictionary, but every week I see new words that I recognize, so the progress is real and there’s no way I’d give up.
Maybe you need to apply what you’re studying in a way that makes sense to you.
I absolutely agree. I decided to finally take real classroom lessons, and for me, the structure and being REQUIRED to be there every week is really helping me keep the momentum. Otherwise, life just gets in the way sometimes.
And yes, I also agree that people should start reading as soon as possible. There are lots of other threads on here about that, but I think it is essential to start immersing in the language as soon as possible.
I first tried it quite a while ago, but I didn’t like it then. It just wasn’t the right time for me. I think the program could do with some better onboarding. I might have been more open to it then if it had had that.
I would have been projected to finish around the end of 2018 on my first try. But instead, I have this:
