Hacker News Reader

> Every few years I use Anki heavily for language stuff (German, more recently Latin and Classical Chinese), up to 90 minutes a day, but I can never keep the workload under control, or my habits change, and give it up.

Well... yeah? I guess if you're a full time student or something it's fine but IMO if your review times exceed 5 minutes it's time to stop adding cards for a while.

I had the same problem when studying English. I just capped my review/new cards study time to 30 mins per day. Basically if you have too much cards to review your vocabulary hits a "plateau" until your workload becomes manageable again.

I had the same experience. I invested a lot of time on Anki and it paid back, but I could never stick to it. There is always a point where I have to skip 2 or 3 days because of life and the huge backlog is too much that I give up.

I recently started to use Supermemo because of some comments on HN, and so far it have been going quite well. I feel less overwhelmed when I have to take a break, I think the algorithm (or maybe it is just good defaults?) is much less punishing. I am still new to it, and I have a baggage of knowledge from my past experiences using SRS, so it is unfair to compare it to Anki at the moment. But it renew my passion for memorizing, so far it has been worth the investment.

I think Anki was based on an old version of Supermemo. It would make sense if the latter had a better algorithm.

> There is always a point where I have to skip 2 or 3 days because of life and the huge backlog is too much that I give up.

Some things I do to manage this:

You can put an upper bound on the number of cards reviewed per day. For me it's 50. After that it says I'm done for the day and I wait till the next day.

If missing only 2-3 days leads to such a huge backlog, you really need to take a break in adding new cards. My aim is to spend no more than 5-10 minutes per day reviewing cards. If it seems I always have too many cards to review, I don't add any new cards until I get only a few cards to review per day (when really busy, I don't add a new card until I have a day with 0 cards).

You can only learn so much per day. You need to limit card additions to ensure you don't exceed that amount.

The other thing I noticed is that I had two types of cards: One where the answer can be instantaneous (e.g. facts). And ones where some thinking is involved (proof sketch of a theorem, details of some algorithm, etc). The latter would kill the experience because of the time involved. So I separated the decks: Simple cards (the majority) go in one deck and this is reviewed daily. Hard cards have their own deck(s) and I'm not as committed to them - I work on them only when I know I have time and am committed to it.

Anki algorithm's is extremely punishing, which is awful for language learning.

I think SRS needs to be only a small part of the learning routine, the cherry on the cake. For me that is ~15 minutes / day, and backlog is manageable even if I occasionally take 2-3 days off.

The new Anki FSRS algo lessens this punishment significantly.

People focus way too much on the backlog - don't even look at it, just set your max cards per day to something you can actually DO and let Anki schedule them out as best it can.

It is, after all, just a scheduler.

I have issues keeping up with large language decks, too. In my case the problem usually lies not with Anki but all the other stuff going on in my life — stressors, work, etc. There’s just a cap on what I can be juggling at any point in time and if things aren’t flowing smoothly it’s going to be more difficult to stick to working through a long deck. If I could just hit pause on everything else there would be no problems.

Retention is high so that helps keep progress between periods of deck usage at least.

I also use anki for german. I use a deck with 1850 cards to learn the nouns' gender.

I just pulled up my stats.I have been studiyng for almost a year and have averaged less than 2 minutes of study a day (with the average review lasting less than 2 seconds). In total it's more or less 10 hours.

It has been insanely effective. I read somewhere on HN that making extremely easy cards helps. It for sure helped me to keep the habit.

> Every few years I use Anki heavily for language stuff (German, more recently Latin and Classical Chinese), up to 90 minutes a day, but I can never keep the workload under control, or my habits change, and give it up.

This is a problem, but perhaps not the one you think. The goal here is the habit; how much you get through is way secondary. As a multi-year user of Anki, I'd recommend working on the daily habit as an addition to whatever German, etc. you're already doing.

Anki has easy ways to limit your use; set your max-card-per-day to something silly like 5. You can do 5 cards a day, yeah? That'd take less than a minute, typically.

Once you're ok with that, up it a bit. Don't make this the SOLE vector of practice, use it as an adjunct to your normal routine of learning.

A habit done suboptimally is >>>> missing doing it, but perfectly. Frequency is generally better than quality. Once you get the former, ratchet up the latter.

These two things are related:

> I can never keep the workload under control, or my habits change, and give it up.

> to build up a highly personalised system that doesn't need to work for anyone else.

"Anyone else" includes "future you", because humans change over time. A highly personalized system is usually a highly inflexible system.

Hmm. Let me think about your point.

The problem seems not dependent on the specifics of my efforts for me - it seems inherent to Anki for language-learning. That the flash cards have lots of bells and whistles seems unrelated to my giving up because 90 minutes a day wasn't enough to keep things under control.

I haven't had any time that my circumstances have change so much that my accessorising ended up a negative. (And I'm having trouble thinking of situations where that would be the case).

But sure, in general I can think of cases where it's probably more advisable to go with a general purpose solution (stairs rather than a climbing wall for going upstairs) in case personal circumstances change, temporarily or permanently.

Inessential complexity increases the workload of maintaining a system (like the climbing wall vs. stairs: can't maintain an upstairs floor unless you maintain the ability to use the climbing wall).

Bells and whistles, especially when it comes to making things perfectly tidy, tend to introduce a lot of rules for using and maintaining a system. For example, a color-coding system with too many colors to reasonably recall without daily practice from using the system: any interruption, such as a vacation or even a weekend, will disrupt in some way the ability to use and maintain the color-coding system.

Informal rules also accumulate over time, adding to the complexity of any given action. It doesn't take a major change to topple your flashcard system, but slow, accumulated changes have been enough to do so.

90 minutes is a very big ask if you want to be consistent. Maybe you're simply trying to bite off more than you can chew? I think consistency is king, so whatever you have to do to stay consistent will pay dividends, even if that means shaving your working set down to just 10 minutes a day.

The thing with Anki is that you cant control how much workload you have per day. Anki algorithm decides that. Say, you start by doing it little bit every day and you are fine for first three weeks. But then, Anki starts assigning you more and more repetition and that is it.

If you do less one day, Anki gives you those cards the next day on top of the normal workload of that day. There is no way out.

Worst, the less you retain, the more mistakes you make, the more your workload goes up. Anki algorithm never do the "lower workload because this person does not retain" decision. Instead, it will just give you more and more work in such case.

There is a cap on how many cards will get reviewed from a deck each day. Lower it if the default cap is too high.

If you're finding that you reliably do not recall some card Anki will eventually mark it as a leech. Those cards get suspended automatically after some number of lapses. That will reduce your workload because suspended cards are't part of your workload.

https://docs.ankiweb.net/leeches.html

The leech feature can also be dangerous if you're unaware of it, because AIUI Anki does not ask you whether to remove a difficult card or not, it just goes ahead and does it.

That's true, though the stats screen will tell you how many suspended and buried cards you have. If I track that a few times a week I can usually tell when something has been suspended that I didn't suspend myself.

> Anki can help you identify leeches. Each time a review card 'lapses' (is failed while it is in review mode), a counter increases. When this counter reaches 8, Anki tags the note as a leech and suspends the card. The threshold, and whether to suspend or not, can be adjusted in the deck options.

> If you keep failing that card, Anki will continue to alert you about the leech periodically. These warnings occur at half the initial leech threshold. For example, if you set the warning at 8 lapses, future warnings will happen every 4 lapses (at 12, 16, and so on).

Suspend means you never see the card again. So how could you keep failing it after it's been suspended for being a leech?

You need to mark that card with lowest recall for leeches to happen. If you mark it with second lowest one, it won't become leech.

If it becomes leech, it will be ignored entirely by anki.

All of that is too much expertise about how to mark cards instead of allowing you to do it intuitively.

I really wish something like https://github.com/fasiha/ebisu becomes the norm. That is, the idea of fitting the cards to your time (by prioritising) rather than you having to do everything there software wants.

The only bit missing is some algorithm deciding how often to introduce new cards based on your historical data.

Yeah I've had a similar issue with the Anki-like system Wanikani uses for learning Japanese Kanji. I can keep up for some time, but as soon as I get busy with something else, I get overwhelmed by the piling up reviews. It has been very effective for the ones I learned when I was able to keep up though.

Recently a friend got me into Duolingo with him, I've been managing to maintain a streak on it because if I'm too busy I can atleast just do a quick "lesson" before midnight, but I don't really find it to be very effective for learning unless you're coming from literally zero starting knowledge.

I already know the basic vocabulary (especially phonetically) and sentence structure, so it's too slow. I usually just do the first handful of lessons and have figured out the kanji enough to just jump ahead to the test to unlock the next unit.

Are you using Duolingo for Japanese? If so, please stop. There are better and more efficient ways, as you've probably noticed. Anki (with a Core3k or similar deck, 10-50 new cards a day depending on your daily life) + reading is the way to go in my opinion. You can use jpdb to see your progress relative to the material you want to read.

Yeah, as I said, it's not very effective and I'm only really managing to keep it up due to being able to do a quick 15 minute lesson right before bed. I have some workbooks and do often end up perusing and attempting to converse in japanese to practice, but I don't manage to keep it as consistent due to all the other things I have to balance.

Language learning and art practice both tend to quickly get put on the backburner for weeks whenever I get busy with a project or whenever work gets busy.

> what the ecosystem would look like

There is a third-party paid exchange here:

https://www.ankihub.net/

(I’ve had a few exchanges with the founder, and wish him luck.)

The trigger that helps me is Anki on my phone will put little notification badges on itself if I have cards to review, and I hate to see unresolved notifications, so it forces me to study.

> They have a lot of shared decks hosted online

I have mixed feelings about shared decks. I’ve tried many, yet always find something about them that irritates me - aesthetics, content accuracy, etc. Or more likely that just doesn’t fit contextually with what I’m learning.

However, the low startup energy with shared decks is certainly a selling point. I can’t even begin to estimate how much time I’ve spent over the years creating 50K+ cards…

> highly personalised system that doesn't need to work for anyone else.

My experience exactly.

> I wonder what the ecosystem would look like if things were otherwise.

Shameless plug - I'm building https://github.com/AlexErrant/Pentive which is basically GitHub/Reddit for flashcards. Very much pre-product and a WIP, though the offline client proof of concept is done.

Yeah, I've had the same thing every time I have tried Anki - it's fine for a while, but once you get enough cards added, or just have been memorising a deck for long enough, even missing one session generates a huge unsurmountable backlog. It's bad and consistent enough for me that I just stopped trying to use Anki at all.

Is there no way to skip a day? And simply pause?

You can suspend cards and decks, but also you can skip days.

It'll create a backlog of cards to review, but that's surmountable. The number can become intimidating (I think my worst was around 2k cards), but there are a few ways to clear it. You can "just" get back into the daily habit, it'll take (by the default numbers, if one deck) # cards/100 days to clear out, more or less (some cards may come up for review again in that period so the actual number might be a bit higher). You can also up the daily review limit to clear it faster (I did this when I hit 2k, 200/day was about my personal time limit to spend on it so it took just over 10 days to clear).

I get skipping days is against the algorithm, but wouldn’t it be better to simply freeze the algorithm for a few days opposed to get a backlog at all? A backlog would break the habit for me making me unlikely to return.

Skipping days is not against the algorithm, it accounts for skipped days.

In a simplified system suppose that we just double easy card review times, and reset hard cards back to 1. You have a card up for review today, you last saw it N days ago. Two scenarios:

1. You don't delay. You review the card now. It's either hard and reset to 1 (see it again tomorrow) or easy and you see it again in 2N days.

2. You delay. You review it M days from now. When you finally do it's either hard and reset to 1 (see it again the next day) or easy and you see it again in 2(N + M) days.

That's it. The algorithm has you covered if you delay. It doesn't do something silly like say "This card was supposed to be reviewed after 2 days, but you waited a month. You remembered it, but we're going to show it to you again in 4 days." The algorithms will take the delay into account (maybe not one-for-one) like I illustrated above.

Good point. The real risk in skipping days is that you might forget altogether some of the cards that were due for review. But the Anki default is to review often enough that probability of recall for each card is very high, so if you're only skipping a few days at a time this is not a huge concern. Capping the amount of cards you do per day has a similar effect; Anki will prompt you with the highest-risk cards first, and some will be left unreviewed for the day (hence, practically skipped).

FSRS seems to reduce the amount of reviews required, might be worth enabling it and seeing if it helps.