About that workload graph... Is it correct?

About this workload graph:

Firstly, kudos to whoever made it, I think it makes a useful point very effectively. But I also think it’s wrong. I put the data about how many of each item are available in each level, have the script run through the SRS progression of each item as efficiently as possible, and I get this:

The numbers along the bottom here are week numbers not level numbers (although basically the same I guess). Basically you hit full workload much more quickly than you’d think based on the old graph and the drop off is much sharper because you don’t have any lessons anymore but then low and steady as you burn a handful of things each day.

8 Likes

Maybe I’m misunderstanding something, but how could you hit max work load at 10 weeks, when it would take at least five months before you start to burn things, which is when workload peaks?

6 Likes

And you have included the fact that items come back to reviews at given intervals? So while you are busy adding more and more items from new levels to your review queue, old items that are guru/master/enlightened pop up again and add to your “regular” work load? (Keep in mind that that chart assumes that you do your reviews perfectly; the crabigator may have mercy on those poor souls that fail lots of burn reviews and then have additional items in their review queue!)

I don’t have the numbers to prove it because I never doubted that workload chart but that was also my own experience: things started to get difficult around level 15 and really painful in the mid twenties for me.

(also, where do you get 300+ radicals from in week 0? I think I’m not understanding your chart correctly :D)

5 Likes

Good question. I ran another graph this time broken down by the stage of the items in question.

As you can see the enlightened items come in on time around week 23/24, but it looks like the amount of other items is dropping off. Looking at the number of items total in each level there is a bit of a drop off after the first 15 levels or so from about 175 to about 145. I think that’s the reason.

But where would you get 1000 items from at the beginning? Something seems off here.

4 Likes

Yes, they’re included. My graph in the reply to jaearess probably makes this clearer too.

I could put in fail rates for each SRS stage maybe, model how that works. Perhaps 8% for Apprentice, 4% for Guru, 2%/1%… Would be interesting to see how that affects it.

I’m just starting, it’s true. But I trust the numbers that I have. I can share the code and such if people want to check it.

That’s because of the faster rates of level 1/2 items in apprentice. You just do more review more quickly.

The data is grouped by week, that first week is actually quite busy if you do everything you can because of the shorter SRS timings.

Level 1: 26 radicals + 18 kanji + 40 vocab = 1000??
image

Even if we included lvl 2, it wouldn’t add up to 1000 items.

3 Likes

If you just do a very quick Excel sheet and, for the sake of simplicity,

  • leave out details about any shortened levels (because the first 2 levels are not really relevant in comparison to the big picture and on the other hand, you would have to factor in the short levels after 41 too)
  • assume that every level has 130 items that you unlock immediately when reaching the level

And be a bit fuzzy in general about it because I just wanted to get it done :wink:

Then I get this. But I see what you mean, it is not as drastic as the workload chart makes it seem. However if you assume that you don’t get perfect review percentages then things feel very much like the workload chart.

If you would also put in the short levels, then things would get much worse after level 40 though.

image

Week 1, everything done as soon as possible, contains all items from level 1 and their first five reviews, all radicals and kanji from level two and their first four reviews plus the lessons for level 2 vocab and their first two reviews, and all radical lessons from level 3 and their first review.

(26 + 18 + 40) * 6 = 504
(34 + 37) * 5 = 355
81 * 3 = 243
20 * 2 = 60

Total 1162.

2 Likes

The wording in the initial post and the lack of axis labels was just potentially confusing, the y axis shows reviews, not items.

2 Likes

Yes, this looks overall quite similar, the level to level variation in number of items accounts for much of the difference.

I’ll see if I can model this somehow, see what happens.

1 Like

Ahaa, okay. So that’s the math behind it.

This mislead me, I guess.

I was tricked :0

1 Like

Wait a second, I put burn reviews after 4 weeks instead of 4 months, I’ll redo my chart :wink:

1 Like

So, I think those other charts show “total active, non burned items in the system” on the y axis, and days on the x axis. We have “total reviews per week” on the y axis and weeks on the x axis in our charts, so of course they will look different.

After fixing my chart intervals, it looks like this (something is buggy at the end, didn’t bother looking into it, should go down to 0). There should be another rise after week/level 41 but I was too lazy to show the short intervals there.
image

If you factor in the a couple of failed reviews, it will look much worse.

But no matter what the charts look like: enjoy the first 10 levels, they are called “The Pleasants” for a reason :wink:

2 Likes

Here is a better version of the workload chart btw, with a description what we are seeing on the x and y axis:

image

I do however agree that calling this chart the “workload” chart is misleading. Since it shows number of items in the SRS, not the number of reviews. . We would have to create a chart that is also based on hours instead of weeks to see how different our charts would look to the original workload chart. Everything looks more drastic when it starts and ends at 0.

9 Likes

Ah, so the original graph was just number of items in the system. Thank you, I was not aware of that. When I tried to do it by hours it just looks spiky. Even days looked spiky actually, dropping to weeks was a necessary smoothing adjustment.

2 Likes

spiky

Yeah, I think the chart that we now know as “workload chart” was also simplified to make a point. Probably by someone who was slightly annoyed by the always repeated question of “why is WaniKani so sloooooow in the beginning?” (which I think is better these days because the intervals where not always shortened in the first two levels?).

But it would have been nice to have a legend for the axis at least. Charts without axis labels are evil! :see_no_evil:

5 Likes

At least it’s accurately inversely proportional to a user’s free time

Agreed.

Here’s one last graph then I really should get back to studying. I added a small chance of failure (1% at apprentice decaying to 0.1% for enlightened).

So these are a little random now of course, but it has this crazy dip where the same radical kept failing to get out of apprentice and dropping back by three days. But the overall shape remains similar, more or less. Thank you for the input.