Question regarding speed

[jprspereira]

Just some extra ideas

@EiriMatsu

all Kanji lessons

I usually recommend splitting the kanji lessons before the radicals get guru’d because not everyone (if not most) aren’t able to memorize that much information at once (yet).

@EiriMatsu

Doing small amounts of lessons at a time may seem easier, but, ultimately, you’re going to miss review times and they’ll bulk up, especially if you assume lower accuracy.

Sure, but the same could be said by missing a review time of a larger group of items. The difference with big lesson sessions vs small lesson sessions is that the former will be better for people with previous knowledge of what’s being learned while the latter will be better for those learning the content for the first time.

Example:

Person A knows 50% of the vocab already. Does 40 lessons. Learns 20 new words.
Person B knows 0% of the vocab. Learns 20 new words.

Same intensity, but different results in the WK system.

Not to mention that one’s retention of new information goes down the more they study (assuming no or little rest).

[EiriMatsu]

@jprspereira

I usually recommend splitting the kanji lessons before the radicals get guru’d because not everyone (if not most) aren’t able to memorize that much information at once ( yet ).

I could see that. Personally, because I find the kanji more difficult than the vocabulary, I like to group them together and separate them from the vocab so I can better concentrate on them. It’s also true though, that I do know a great deal of the kanji’s meanings in advance though (as its mainly readings I don’t know).

@jprspereira

Sure, but the same could be said by missing a review time of a larger group of items. The difference with big lesson sessions vs small lesson sessions is that the former will be better for people with previous knowledge of what’s being learned while the latter will be better for those learning the content for the first time.

It’s more difficult to have a backup of large reviews, though. During the Apprentice 3 and 4 stages, I usually have hours of time without reviews, sometimes up to over a day’s worth depending on how they line up with the burn reviews. I review in the open time when I arrive early at work, right after lunch during break, and after dinner. Otherwise, I don’t touch WK and still maintain about 7d21h average.

My argument is less about intensity per say, but time efficiency (as I understand that bulk reviews are definitely more intense). If one wants to assume a 90% accuracy rate, they’ll have to spend much more time (or rather, have to dedicate more periods of time during their day) reviewing do the the number of reviews. Bulk reviews all for the least amount of times necessary to visit WK, as all reviews remain on the same SRS queue with little overlap until Guru+, where mistakes no longer inhibit leveling. Even if you miss 10% of items, they will all move to a bulk queue that has little chance of overlapping with the main chunk a second time.

[acm2010]

@EiriMatsu

Doing small amounts of lessons at a time may seem easier, but, ultimately, you’re going to miss review times and they’ll bulk up, especially if you assume lower accuracy.

Your approach is bulking up items from the start, where is the difference?

Another problem I have with your approach is that you put the kanji in a place where failure is guaranteed to lead to delays. When you start with kanji right away you can go easy on half of the kanji and fail a few times.

In general the idea of SRS is to fail items until you remember them, but for going fast you have to learn them well enough to last for a week, and the reviews are just obstacles that potentially make you go slower. I basically skip the SRS until guru and then let them resurface (but I usually remember them well enough).

[EiriMatsu]

@acm2010

Your approach is bulking up items from the start, where is the difference?

The difference is that doing fewer lesson items at a time (smaller lessons) is supposed to lighten the load. If such an approach leads to bulking, one might as well simplify the process and bulk from the get-go. The only way to avoid that is higher accuracy or devoting a huge amount of one’s day to shifting on and off of WK.

@acm2010

Another problem I have with your approach is that you put the kanji in a place where failure is guaranteed to lead to delays. When you start with kanji right away you can go easy on half of the kanji and fail a few times.

Failure is guaranteed to lead to delays regardless though. Even if you 100% the initial kanji, if you mess up the remaining ones, you can’t progress. My approach simply makes it more likely you’ll miss more than 10% of the kanji due to that lack of a “safety net.” Back before I started this, I had a level or two where I made enough typos on the last grouping of kanji to be delayed a few hours after I’d guru’d all of the initial kanji.

As for the idea of SRS, if one has less than 90% accuracy on a maybe 35 item review for items reviewed no more than 2 days ago, they are rushing reviews and/or shouldn’t be progressing to the next level in the first place. The point of OP was not necessary to attempt to level as quickly as possible, but to spend as little time as possible reviewing. And seeing as that is the case, it is my opinion that one is more likely to remember the kanji when they aren’t in the middle of a pool of vocabulary, leading to higher overall accuracy and lower risk of failure.

I haven’t had a kanji leech since I transitioned to this review style, so it’s definitely working for me (and 5/6 of my vocab leeches are typos). I figured I might as well suggest it since I also try to reduce the amount of time/energy I have to spend on WK as much as I can.

[_Marcus]

@jprspereira

For example, at level 30 you’ll be able to recognize 86.48% of the kanji present on Wikipedia

Wow! I didn’t know that! Thank you!
Target - level 30!
That suddenly makes everything much more doable!
Thank you @jprspereira!

[ptb]

I do one lesson session per day (when available) of 14 (includes radicals and/or kanji) or 21 (pure vocab) and it takes me ~15 minutes to do my lessons. I did a batch of ~100 reviews in one sitting last night and it took me the length of the latest episode of Archer, almost to the second, to finish them, so around 24 minutes or so I guess?

I’d say, once you get used to WK and get more confident and faster at answering, it’ll dramatically reduce time you spend actively doing the reviews. jprspereira’s numbers don’t seem too far off for me.

According to wkstats I average just over 10 days per level at my current speed. I have no intention of speeding up but I may slow down in future.

By the current estimate though I’ll hit level 60 the same day I turn 31, so…

[omoide]

8804 items/20 new items a day = 440 days.
8804 items/15 new items a day = 587 days.

Practical question: where can I check how many lessons I already took today? I mean the total of today sessions.

Thanks

[jprspereira]

@marcusp

Wow! I didn’t know that! Thank you!
Target - level 30!
That suddenly makes everything much more doable!
Thank you @jprspereira!

Just to remind you that it doesn’t mean you’ll be able to understand 86% of content But yeah, get to level 30 and you’ll definitely feel that an enormous progress was made.

@souvenir

Practical question: where can I check how many lessons I already took today? I mean the total of today sessions.

Thanks

Well, as far as I know, Wk usually records only the items learned in the last session. There’s not really a place that directly gives you the total. What you can do is pay attention to the amount of lessons you do

[Naphthalene]

@OP:
I’m actually quite curious about your maths. You keep saying they don’t match reality, but could you just tell us what you expected?

Also, I hope you have sent your APIv2 key to @rfindley by now, because they are the most likely to be able to help you figure it out…

[rfindley]

@souvenir

Practical question: where can I check how many lessons I already took today?

If you have the Open Framework installed, you can paste the following in your Javascript console and press Enter:

wkof.include('ItemData');
wkof.ready('ItemData').then(()=>{
	var day = new Date(Date.now());
	day.setHours(0,0,0,0);
	wkof.ItemData.get_items('assignments').then((items) => {
		var lessons = items.filter((item) => {
			if (!item.assignments) return false;
			var start = new Date(item.assignments.started_at);
			start.setHours(0,0,0,0);
			return start.getTime() === day.getTime();
        })
		console.log('Lessons done on '+day.toDateString()+': ' + lessons.length)
    })
});

If you want to check for any other day, change this:

	var day = new Date(Date.now());

to something like this:

	var day = new Date('Feb 12, 2016');

[GrumpyPanda]

@jprspereira

8804 items/24 new items a day = 367 days.

Sorry if you’ve already answered this, but the thread turned out quite long and I didnt want to read through all of it…
With the system you talked about, isn’t that speed impossible simply because WaniKani doesnt allow you to go that fast?
Assuming you always do your lessons at 9am, wouldnt you wait 12 hours between unlocking a new set of Radicals/Kanjis (since you unlock them at at 9pm) and every level would take ~8 days?

[WindFalcon]

Part of the problem I think is that people assume the use of a script to reorder the lessons, or doing the vocabulary parallel with the kanji/skipping it, and not one after the other without much of an overlap.

I think my experience of always having a good number of lessons is because of the problem that both my level 3 and level 4 kanji got unlocked at the same time like @rfindley noticed.

Also, I still don’t quite get some of the suggestions (though I’m very appreciative that people share them) - do people advocate when going for daily efficiently (meaning as little amount of time spent on WK per day as possible) to go for all lessons at once as soon as they become available, or aim for the average number of daily new items that would produce a workload (lessons + reviews) that is manageable within that small amount of time?

Again, lets assume that no reorder script is muddling the equation. I’ll figure out further optimization with such scripts when I’ve figured the vanilla method.

Also I don’t care so much for doing actual calculations in my posts since like I’ve discussed, there are a lot of potential factors, some of them unforeseeable, that we can’t know beforehand (the calculations we do in the tread assume 100% accuracy and optimal routine for example), so the math won’t be accurately reflecting reality unless you are one of the single digit % people that have a lot of time and can strictly stick to an optimal routine and/or are a dedicated speedrunner. That’s why I’m also interested in hearing slower people that reached 60 give their input and share what their daily workload was and how they found out best to handle it for themselves.

I was also hoping that that user that posts simulation graphs would show up and give us a couple ones with 5/10/15/20 items per day and a reasonable margin of error, so we can get an idea of how bad the review count can get for each of those (or just post his excel sheets that I guess he uses as a template).

I just don’t think that for example the ~9.8 days level up time for 15 new items per day that the math produces is actually achievable for most users of WK, and in reality if we had site-wide statistics for everything, I think we would find out that the people that had that daily 15 item goal actually in the end got a lot slower days per level time. This is why I initially thought it’s impossible to get the ~1 year level 60 with 1 hour a day, I assumed that is only possible if people are doing all lessons immediately as they become available and somehow still managing to do all reviews immediately and maintain a near perfect accuracy on them, which means that their workload is the worst possible that the site can produce, and hence needing more than one hour per day. I guess when a person is amazing, they are amazing at absolutely every element at the same time, at any given time, indefinitely, huh?

This thread has proven that it is as least possible for some, yes.

I still think though that for the rest of the people that are struggling with little time/big workloads the idea should be to lighten the load, while most posts so far suggest doing the opposite (if I understand them correctly, but almost no one talks with concrete average per day numbers and what workloads they produce) - get as high a workload possible and just do as much of it as you can, which is not a good long term strategy at all in my opinion (especially from a psychology of learning/teaching methodology perspective), especially when starting out. At levels 30-40 - maybe its the best strategy, but not initially.

[jprspereira]

Not sure if I understood your question, but each normal level can be done in 6d20h. There are also 17 levels that can be done in 3d10h (if I counted them right). Are you having problems in understanding the intervals, is that it?

[Naphthalene]

@WindFalcon

Again, lets assume that no reorder script is muddling the equation. I’ll figure out further optimization with such scripts when I’ve figured the vanilla method.

Well, the vanilla method has more variability than the reorder method, since the number of days until you do the lesson for the radicals will change based on the remaining amount of vocab from the previous level.
There will always be some vocab from the previous level, since, by definition of leveling up, you just guru’d their kanjis.
So, if you do 15 lessons per day on a level with few vocabs or 20 lessons per day on a level with more remaining vocab, your leveling speed (and thus your hability to get more lessons) may be the same, or even slower in the second case.
Again, this is why we keep recommending to use the reorder script, since it means that leveling speed will be directly correlated to the number of lesson per day and your accuracy. It’s not the case with the vanilla method.

To put it another way, with the vanilla method, assuming you can do on average, say 18 lessons a day like rfindley, you will eventually get the same average leveling speed they do, and the same completion time, but with variation of the amount of days per level.

So, your maths may very well be correct, and yet not match your observed leveling time.

@WindFalcon

if I understand them correctly, but almost no one talks with concrete average per day numbers and what workloads they produce

So, I’m on my phone, and I can’t produce complexe simulations, but averaging everything, for a 90% accuracy rate (assuming you get enough lessons per day, etc). Also assuming only one review session per day.
A failed item has to be answered three additional times (technically it depends on its level, but lets go with that), so you have a 130% work load.
There are eight levels for items, then they disappear from your queue (burned), you your average load per day during the equilibrium phase (burning items, still adding items) is:
1.3lessonsperday8 (reviews) + lesson_per_day (the actual lessons).
So for 10 lessons a day, that’s 104 reviews per day on average plus 10 lessons.
Twice as much (obviously) for 20 items a day.

In practice, you would probably be slower, since one review session cannot give you 20 lessons a day on average, though…

[acm2010]

@WindFalcon

I was also hoping that that user that posts simulation graphs would show up and give us a couple ones with 5/10/15/20 items per day and a reasonable margin of error, so we can get an idea of how bad the review count can get for each of those (or just post his excel sheets that I guess he uses as a template).

You are probably talking about me

I did some one-shot simulations, the days needed will vary a bit depending on what kind of errors you did. The errors are independent, so there are no leeches or items that are more problematic than others.

But I don’t really get what you are trying to achieve, ask if you want some other parameters

[Click for graphs]

~90% accuracy, reviews twice per day, 10 lessons/d without reorder => 868 days to level 60, ~100 daily reviews

~90% accuracy, reviews twice per day, 15 lessons/d without reorder => 599 days to level 60, ~140 daily reviews

~90% accuracy, reviews twice per day, 20 lessons/d without reorder => 522 days to level 60, ~170 daily reviews

100% accuracy, reviews twice per day, 20 lessons/d without reorder => 466 days to level 60, ~150 daily reviews

~90% accuracy, reviews twice per day, max. lessons/d without reorder => 495 days to level 60, ~175 daily reviews, but more variance

And no guarantees, I just hacked the daily lesson feature into it.

[With reorder I think you can get everything to around 500, the delays come from radicals and kanji waiting behind large walls of vocab from new guru kanji. If you have a 90% accuracy but do reviews and lessons as often as possible you need ~420 days.]

[Naphthalene]

Wait, so to reach level 60 in one year, you need basically perfect accuracy on radicals, kanji or both?

[acm2010]

At least for the items on the “critical path” you need high accuracy, but you can get the kanji you get along with the radicals wrong until you guru the radicals, and you can get radicals and kanji wrong worth 10% of each level (the radicals themselves don’t matter, but the kanji they unlock), and after guru you can basically fail as you like.

But the margin of error is not so big

[Edit: if you mean the last with 420 days, with 90% you are likely to get 4 kanji wrong and be stuck for a while in each level.]

[WindFalcon]

Thank you both @Naphthalene and @acm2010 for the replies and effort.

Very interesting. It seems that from a min/max perspective, a person should aim for at least 15 new items per day, probably mixing 15 and 20 since there will inevitably be times where he skips a day due to life, if he wants to maintain the speed that is.

One thing to clarify, with 15/20 new items a day, there will be days where you don’t have lessons, or not that many, do the graphs reflect that as well?

But if in the very end the reorder script doesn’t produce a difference with the vanilla method (if I understand correctly), why use it at all? Obviously people don’t care for the statistics on wkstats, but for what actually happens in real life.

Also, does using the reorder script means you do lessons on vocab with kanji you haven’t done the lesson for? Because I would assume that for a person who is slow/lacks time in the first place, this will slow him down even further (because its easier to do kanji->vocab than vocab->kanji lessons).

[Naphthalene]

I’m not sure about the first question, but for the other two:

@WindFalcon

But if in the very end the reorder script doesn’t produce a difference with the vanilla method (if I understand correctly), why use it at all? Obviously people don’t care for the statistics on wkstats, but for what actually happens in real life.

I can’t speak for everyone, but for me at least, the reorder script has two advantages:

• stability: if I do the time critical item first, my leveling will always happen at the same rate (7 days, in my case). I can then spread the vocab items regularly throughout the week.
  Specifically, I (used to) do 30 lessons a day, if available. Upon leveling, I do all radicals + kanji (almost 30 items, usually) + 1-3 vocab items. Then 30 vocabs the next day, then up to 30 vocabs the third day (usually there isn’t that many left). Fourth day (technically 3.5th day), I unlock new lessons, I use the reorder script to do the kanjis first (less than 10) and complete up to 30. Fifth day, I usually finish my lessons. Sixth day: break! Seventh day: level up, rinse and repeat.
• peace of mind! Sometimes, I only have five minutes for review, but I know I have time critical items in the mix. I just use the reorder script to do those items, and then wrap up the session.

Of course, both properties can be (and are) abused by people, but that’s not the fault of the script itself.

@WindFalcon

Also, does using the reorder script means you do lessons on vocab with kanji you haven’t done the lesson for?

Nope. It just allows you to change the order of lessons (or reviews) you currently have. It does not allow you to access locked lessons.

[WindFalcon]

OK, so what exactly do you reorder then? Isn’t the normal vanilla distribution exactly what you talk about (radicals+kanji critical items first, then vocab)?