Suggestion: Self-adjusting number of radicals per level


#1

DISCLAIMER: THIS ISN’T YET ANOTHER THREAD ABOUT HOW EARLY LEVELS ARE TOO SLOW.
In fact, the current proposal may make levels even slower.

MAIN PROPOSAL: Based on success rate of user during reviews, adjust the number of radicals they will see on the next level, from x0.25 to x2.0.

  • In the case of smaller amount of radicals, additional “intermediate” levels are created: once you have guru’d all your allocated radicals, plus 90% of the appropriate ratio of kanjis (same rate as the radicals or of all available kanjis for the unlocked radicals, whichever is lower), you unlock the extra radicals for that level (and additional available kanjis, if applicable)
  • In the case of larger amount of radicals, the number of kanji/vocab per level does NOT change. The only difference is that some levels may go from being ~7days levels to being ~3.5 days levels.

ALTERNATIVE PROPOSAL: Same as the main proposal, but the rate is adjustable in the danger zone rather than based on an algorithm. It seems easier to program.

BACKGROUND:

  • WaniKani has very nice dependency tree going radicals -> kanji -> vocab.
  • Number of radical per level seems “fair”, but is arbitrary in grand scheme of things
  • This proposal is inspired by this discussion between @koichi and @Vanilla about how to reward fast (correct) reviews. Note that it also applies to the case where the user is making a lot of mistakes.

EXPECTED RESULT: People will get a more appropriate review queue size. Those who are having troubles keeping up will get a lower amount of lessons without having to self-regulate. Those who find things are too slow (and are breezing through their reviews) will get more lessons, but in a controlled way, so that they don’t risk to be overwelhmed later on. (Also, in the algorithmic case, that will also be taken into account, plus resetting a few levels back is always an option too if things really get out of hands)

ADDITIONAL COMMENT: I feel it wouldn’t be to much of a hassle to code, since there’s already an order for radicals and a full dependency tree. If I understand correctly, there’s a (major?) update of WK coming, so I thought it’s the right timing to make that suggestion. :slight_smile:


Wanikani is just not useful enough, but could be
I'm level N3 but I just started using this site. Do I really have to start from the beginning?
#2

You want to detach the radicals from the levels?

Starting from level 14 there are not so many radicals per level anyway, and from level 40 there are only a handful of new radicals. I would be worried about the pacing past level 30.


#3

Yes.
And yes.
That’s the point, really.
I feel like at higher level the speed restriction (making level slower) is the only thing that matters anyway.

Edit: it might not be super clear in the original proposal, but if you are at a below 1 pace, the number of kanjis available at one time is also reduced.


#4

It not so clear to me what you mean with “adjust the number of radicals they will see on the next level, from x0.25 to x2.0.”

I don’t think the number of new radicals is explicitly used, rather chosen to yield the right number of kanji.You mean the old version is 1/4 * number of kanji, or what?


#5

Yes, that’s what I meant.
Edit: wait, no, I’m not sure I get what you mean.

It makes a lot of sense to me to have more radicals early, and less and less as you go on through your study.

The only point is to dial the speed (with a multiplying factor). That will also impact which kanjis and thus which vocabs are available.
Moreover, Mini levels mean that you still get to see radical, kanji and vocab at the right timing (you don’t have to go through all the kanjis for a level before you get to see vocab, the way it would be if you just restricted the number of lessons you are doing in a day)


#6

Did you see this:
https://www.idigtech.com/wanikani/#items.wk.rad

For example on level 40 you only get 3 new radicals but 32 kanji. How many radicals do you want to give? Some later levels have no radicals, they might collapse into a single one with like x*150 lessons?


#8

I think referring to levels makes this more complicated. In essence you’re proposing that users go through the items in the intended order they were meant to come, but rather than items coming in stages (in the current system: radicals, guru’d, kanji, guru’d, then the next level with vocab and new level radicals, repeat) the items come in a measured stream (what I think you proposing: radicals, any items that are guru’d brings the related kanji, any kanji that are guru’d brings the related vocabulary). Is that what you’re saying? The user isn’t waiting for 90% mastery for intralevel items, but get lessons for items that they qualify for?


#9

@acm2010
Yes, I know of the stats website. It feels like I spend maybe a third of my day there…

In the case of level 40, at x2 speed you would get 6 new radicals, if that many are still available (I guess not, if you were consistently at x2 there should be 0 radical left to learn.)
You would get all 32 kanjis immediatly (assuming you had guru’d their radicals already).
For a level with 0 radical, there’s no difference between x1 and x2 (and anything in between, if relevant)

At x0.25 speed, you would get 4 mini-levels: 40-A *with one radical and 8 kanjis, 40-B (1 radical, 8 kanjis), 40-C (1 radical, 8 kanjis), 40-D (0 radical, 8 kanjis). The number of kanjis may be skewed a bit toward C or D depending on what is available. You need 100% guru’d radicals and 90% guru’d kanji for a mini-level to go to the next (as usual)

@LucasDesu
That’s pretty much it, yes.
Well, I like the gamification aspect of levels, so I wouldn’t propose a full stream. Moreover, it would require a deeper transformation of WK to achieve that. Finally, forcing periods of breaks feels like it helps with fixating memory (I have no proof of that, it’s just a feeling; I may check literature on the topic when I have more time) Edit: I mean breaks between periods of lessons, not reviews obviously (breaks between reviews being just the basis for srs)


#10

I like the idea, especially the part where you can go faster. For example, why should I have to wait until 33 to learn 内 and 間 and then have them be the reason the level takes twice as long. The whole system of needing to wait an extra 3.5 days to learn what is in some cases a SINGLE kanji makes no sense.

Reward people with higher acc with the ability to go faster and force those with a lower acc to actually understand the material before proceeding. As it stands, radicals are just used just as a means to slow people down and I don’t understand why. And for people who say “ohh well some of the radicals are used in that levels kanji, so you need to learn the radical before the kanji”, you’re lost. There is no reason those radicals couldn’t have come a level earlier, especially radicals like 内 and 間 which could have been taught before level 10.


#11

What if you always get everything at an increased rate because you’re great at reviews? Do you get more kanji since you get more radicals, essentially decreasing the number of levels? Or do you just get the additional radicals when you speed up?

The first approach could lead to burnout if people don’t pace themselves. Or if they already pace themselves, unlocking additional items sooner will have no practical effect.

The second approach basically gives no reward and could in fact be detrimental since you could learn radicals long before learning any kanji that use them.


#12

From what I’ve seen, Tofugu is a respectable business run by respectable people. But it’s still a business. Slowing down the levels somewhat means monthly and annual subscribers have to subscribe for longer. Additionally, forcing users to slow down decreases the chance of burnout, which in turn results in more people staying subscribed.

While some people may unsubscribe due to this approach, I have to assume Tofugu has done the math and determined that this course is better for their business in the long run.


#13

I dont doubt that there is a business side to this, but its not a pleasant thought that we may be paying for something that intentionally throttles our learning to get more money out of us. I understand it doesn’t apply to everyone, but it does apply to some.


#14

Okay I understand what you mean. I only mentioned disregarding levels because thinking from how WK is currently set up, it’s really difficult to conceptualize what you’re proposing. Essentially the leveling structure would be similar to how a game like Tetris would move through levels. Without interruption, the user graduates into the proceeding level after accomplishing a pre-determined items of that specific level. To clarify, I’m not alluding that WK’s leveling mechanism has interruption in the process, but new items are chunked together more than what you’re proposing. If I’m understanding correctly.


#15

If you always get 100% at your reviews, while taking only a few seconds per answer, then yeah, you are probably fine with more radicals. And that’s it. No additional kanji per level. BUT since you have seen radicals ahead, some levels will go from taking 7 days to 3.5 days.

That can be a pretty daunting speed up, leading to a higher failure rate. WHICH IS GOOD. The system then slows down until it matches the optimal speed for you.
Maybe there are periods in your life when you can’t focus on Japanese and WK. No problem, levels get chunked into two, three, four, whatever sub-levels. You still get to see progress (you are still leveling up), but you are not overwhelmed.
Stuff change, now you have 100% of your time to dedicate to WK. Rates change again and you are back to 3.5 days per level. That’s what I mean by “self-adjusting”.

Obviously, the increase in speed is only valid for lower levels. Once you have seen all radicals, only the slower pace has an effect.

In the mean time, at low-intermediate levels, like @Vanilla said (and like I have seen in other threads) it feels weird to be slowed down waiting for a radical that you ALREADY know to guru so that you can finally get the last kanji to level up.
Again, not saying it’s bad, just that it’s weird. Having a self-adapting system may smooth out that issue.


#16

Well, that’s the logical conclusion of this train of thought, yes. But it would require a lot of changes to the structure of WK.

I feel like my proposal (just doing that with radicals, basically) is a good middle ground. At least that’s what I was aiming for.


#17

Its worth mentioning that the 3.5 level up time would still be optional. You can still be down to 0 lessons by the next level as usual by changing the order you do lessons. Assuming you still kept your accuracy, you could choose anywhere above the 3.5 day level up time minimum.