Limiting Anki Reviews

Just let me preface this topic with: I am not currently considering doing this, I just want to start a discussion :smile_cat: After all, there were plenty of interesting Anki threads on the old forums.

With that out of the way, it seems the general consensus (on youtube, koohii, reddit, etc) is to leave the review limit at 9999 and to never limit reviews. Yet, I wonder if that shouldn’t always be the case.

I manually limited my reviews when coming back to Japanese. I had an old iKnow account, with various words already covered. Instead of resetting my account, I decided to tackle the 800+ reviews that had built up.

To take these reviews I limited myself to ~100 reviews per day; although, I did more when travelling. Without adding new cards, the reviews dwindled down to a small number so I could start adding cards again. I did the same when I stopped reviewing my kanji cards after going taking a holiday. I could’ve used Anki’s feature instead and given myself a nice limit to target.

So personally, I think limiting can be useful when coming back to Anki – but what about new cards?

I wondered if the same would work for a smallish deck. We all know the perils of trying 30/50/100 new cards a day, but what if we set a limit to reviews? Yes, there would be some delays to cards and we might have a higher failure rate – but it might be fine. And in the end, once the deck is complete the reviews will slowly come down under the limit.

I honestly wonder if this is as bad as many said. These days I’ve seen people recommend increasing the interval factor, dropping the failure penalties, resetting ease, load balancing, and doing all sorts of tweaks.

Does it really matter if you see a card after 6 days instead of 4, or after 64 days instead of 48? You might still get it correct?

So maybe limiting isn’t so bad after all? Perhaps it could even work well to limit your Anki time and allow for other studies, such as reading.

1 Like

I was perhaps one of the most against this on Koohii so I’ll clarify my position, though I haven’t actually used anki for its intended purpose in a while. I’m not necessarily against it in all circumstances, in particular I think your example of catching up on an old deck while not adding new cards might be a case where it’s better to use it than not, but generally it seems like a bad idea.

The main problem is that all it does is hide information from the user, which is fine when your not adding new cards, but is disastrous when your getting more and more behind without realising it. And that’s mostly how I see it used, people think setting a review limit is a substitute for being considerate of how many new cards they’re adding.

I don’t think it does matter that much if it’s 6 instead of 4, or 64 instead of 48, but the problem is that once you get to the point of cards being 16 days late all your cards are going to be that late, even the ones you were supposed see the next day. I can no longer remember if anki draws randomly from the pool of due cards or does them in the order they were due, but either way if you see 64 instead of 48 you’re also likely to see 17 instead of 1 and 20 instead of 4.

If you want to reduce the review load, for whatever reason, it’s much better to alter the formula to those larger intervals you suggested in a way that is proportionate to the amount you’ve reviewed each card. Your example of a small deck seems like it wouldn’t be too bad, but the reviews are going to get unnecessarily harder as you progress through it. (I’m also not sure how many anki users actually stop after creating small decks.) I think doing a quick first pass through new material is good, but I would treat reviewing new cards as revision in that case. I.e. I’d read about a subject first and treat the anki deck as revision, working my way through at a pace I can handle.

This is all just my opinion, and if someone finds other approaches work better for them I won’t stop them. Learning well seems to be a mixture of very generalisable good practices and entirely individual needs, so it’s hard to know what is and isn’t good advice.

2 Likes

I agree with Splatted that you shouldn’t hide review information from the user. I’m going to look at it from a different angle, though.
When it comes to limiting review-load, I prefer either doing it manually “Crap, a 1000 card backlog… I’ll review until its down to 800.” or via timeboxing.

In either case, it allows you to prescribe a set amount of reviews (or review time), but leaves the total backlog in your face for a bit of added motivation. “I got it down to 800, but I can keep going!”

Yes, you can still see the total backlog if you look at the deck details, and then add more reviews in chunks through custom study, but that’s the problem: you have to actively look for it and then actively prescribe more reviews. By doing what I suggest above, you are passively exposed to your actual backlog, and any extra study can be done until you don’t feel like it anymore.

If you are adding more reviews through custom study, and you only get partway through before fatiguing, you either have to leave reviews on the counter (defeating the point of ‘daily review limit’ feature, IMO) or you have to force yourself past your fatigue point to work through the rest of them (which is bad for motivation, IMO).

Of course, other people may see it differently, but that’s my reasoning behind not limiting reviews through the settings.

I don’t think it does matter that much if it’s 6 instead of 4, or 64 instead of 48, but the problem is that once you get to the point of cards being 16 days late all your cards are going to be that late, even the ones you were supposed see the next day. I can no longer remember if anki draws randomly from the pool of due cards or does them in the order they were due,

That’s interesting, I guess apart from trawling through the code. The only way to find out would be to experiment and I am content with my current Anki load!

I actually use an increased Interval Modifier and failure penalty, which has worked well for sentences. This is probably because you will see words and kanji readings again in other sentences. Whilst for RTK I found the default settings worked well for me.

Your example of a small deck seems like it wouldn’t be too bad, but the reviews are going to get unnecessarily harder as you progress through it. (I’m also not sure how many Anki users actually stop after creating small decks.)

I’d probably class Core 2K as a “small deck”. If you went ‘wild’ with 20-50+ new cards a day and a cap of 100, you’d finish pretty quickly. Then the learner could move on to grammar study or finishing Genki whilst continuing to just review the cards.

One of the things that inspired me to write this post was a video from Steve Kaufmann. He gets asked about Anki a lot despite not being a fan of flashcards.

It was funny to find out that he doesn’t really use the flashcards on his own site. His flashcard “method” is to look at the card: if he’s correct great, but if he’s wrong he still marks it as “good”! If that was your approach, then I don’t see any harm in limiting reviews.

From the Anki manual:

Review cards are always shown in random order. If you wish to see them in a different order, you can use a filtered deck. More specifically, Anki randomizes reviews by grabbing batches of 50 cards in the order that they exist in the database, randomizing each batch, then putting them together. This means that there is a slight bias towards older cards being shown first, but it prevents individual cards from showing up in a predictable order.

The only real benefit I see of putting a hard limit on the reviews (aside from seeing a smaller number in the UI) is that, if you get any cards wrong, and they switch to the “Relearning” state, they won’t be delayed further into the deck than you wanted to go.

For example, if you wanted to do 100 reviews today and you failed the 99th, Anki would show you cards 101-105 or so before you were given another shot at number 99 to get it out of “relearning”. You couls also get around this by setting the review count to zero for all your regular decks, and then using a filtered deck to pull out due cards and set it to only pull a specific number. I use a filtered deck for reviews anyway because I like to have my cards sorted into decks based on various criteria (because the stats are all at the deck level), but like to do my reviews mixed together (a few vocab cards, a few kanji cards, reading, audio, etc) since it keeps me more on my toes.

2 Likes

Ah, that sucks then. It might be interesting to write a plugin to improve that. Shame Anki doesn’t prioritise younger and overdue cards.

I use a filtered deck for reviews anyway because I like to have my cards sorted into decks based on various criteria (because the stats are all at the deck level), but like to do my reviews mixed together (a few vocab cards, a few kanji cards, reading, audio, etc) since it keeps me more on my toes.

Sounds a bit like iKnow!

I also used to separate my decks but ended up moving my sentences cards into one deck once I’ve covered them. Previously, I used sub-decks, but Anki reviews them one by one…

So I have an organised “sentence bank” of new sentences to pull from with sub-decks (e.g. Core, Tango N2, DoJG, Marugoto Master) and one “Default” sentence deck that I move cards into and review from.