edit - changed the question from "100 reviews" to "in 1 hour."

Inspired by very helpful discussion in this thread, I wanted to know how many cards people can finish in one hour. Please post the deck you're going through (RTK/core10k/sentence mine/etc), and how you format your anki cards. I'll start!

I did 330 reviews an hour today in Core2k/6k with this format:

FRONT
私

BACK
私（わたし）
"I"

私は大学生です。

I have the audio play for the individual vocab word, and then the example sentence. I usually move on before the sentence can play in favor of speed, but I like hearing newer words pronounced.

On average it takes me about ten minutes, but that's just for young and mature cards.
The most accurate answers would be:
-Over my entire studies (new, young, and mature cards), I average 9.6 cards per minute.
-In the past month, it has taken me approximately ten minutes to complete 100 review of young and mature cards.
-Over the past month, my total daily study time in Anki has been about thirty minutes or less
-Over the past month, my longest study sessions have me doing ~280 answers over ~30 minutes.

My template (the deck is Core 2k/6k/10K Further Optimized PIC/SOUND) I changed it a bit to my liking:
###Front:
<span style="font-size: 28px; "> {{Notes-MyOwnNote-Reading}}</span>
<font size="11"><a href="midori://search?text={{Notes-Kanjified-vocab}}">{{Notes-Kanjified-vocab}}</a></font>
<br/><span style="font-size: 15px; color: #5555ff">{{part-of-speech}}</span><br>
<span style="font-family: ＭＳ ゴシック; font-size: 32px; ">{{kanjientence}}</span>

###Styling:
.card {
font-family: arial;
font-size: 20px;
text-align: center;
color: black;
background-color: white;
}

.card1 { background-color: #ffff7f; }
.card2 { background-color: #ffff7f; }

nokana ruby rt { opacity:0.0; font-size:0}
notext b { background-color:#000000; color:transparent }

###Back:
{{FrontSide}}

<hr id=answer>
<span style="font-size: 28px; "> {{vocab-translation}}</span><br>

<font size="11"><a href="midori://search?text={{Notes-Kanjified-vocab}}">{{Notes-Kanjified-vocab}}</a></font>
{{vocab-furigana}} {{Vocabulary-PitchAccent}}<br><br>
<span style="font-family: ＭＳ ゴシック; ">{{furiganaentence}}</span><br>
<span style="font-size: 15px;">{{sentence-translation}}</span><br><br>

{{furigana:examples}}<br>
<font size = '5'><a href = 'http://jisho.org/words?jap={{Notes-Kanjified-vocab}}'>jisho.org</a> </font>
<font size="5"><a class="rel" rel="nofollow" href="http://www.weblio.jp/content/{{Notes-Kanjified-vocab}}">Weblio</a></font>
<font size="5"><a class="rel" rel="nofollow" href="http://ejje.weblio.jp/sentence/content/{{Notes-Kanjified-vocab}}">Examples</a></font>
<font size="5"><a class="rel" rel="nofollow" href="http://tangorin.com/general/{{Notes-Kanjified-vocab}}">Tangorin</a></font>
<font size="5"><a class="rel" rel="nofollow" href="http://tangorin.com/examples/{{Notes-Kanjified-vocab}}">Examples</a></font><br/>
Freq: {{vocab-frequency}}
<div style='font-family: Arial; font-size: 20px;'>{{sentence-translation}}</div>


Basically, it shows the word and an example sentence on the front like:
例文
noun

この文が例文です

The back has the front, plus the translation, reading, and a ton of other example sentences (very useful for those words in Core that had terrible example sentences).

There's no audio nor pictures.

I'm averaging about 12 sec/card (240/hour), but half of those cards are sentence cards so they take longer. When I was doing vocab only it was close to the speed you are going I wouldn't worry too much about comparing speeds with other people though because you can only go as fast or as slow as what's good for you. I actually had to force myself to go slower because I was failing too many cards and it was talking me longer to learn them by going too fast. However I've noticed that the farther along you are, the faster you can go.

This might not be the best place to ask this question, but I don't think it's worth making a new topic for, but, for anki, what counts as a mature card? Is it, say, cards with a review interval of 2 months or greater? One month? 1 year?

About 4 minutes. And 21 days or greater.

Says that I studied 464 cards in 61 minutes today. Most of it sentences (with the audio playing in the question), but 100 or so were vocab (same format you're using), and 20 were Kanji.

That's about my average. I will sometimes stop reviewing to look up a Kanji or word online though, and the clock keeps running for 30 seconds, when I do. There's no way to prevent that from happening, in Anki, unfortunately.

Vempele Wrote:21 days or greater.

Thanks!

Stansfield123 Wrote:Says that I studied 464 cards in 61 minutes today. Most of it sentences (with the audio playing in the question), but 100 or so were vocab (same format you're using), and 20 were Kanji.

That's about my average. I will sometimes stop reviewing to look up a Kanji or word online though, and the clock keeps running for 30 seconds, when I do. There's no way to prevent that from happening, in Anki, unfortunately.

Dang, nice! That's some decisive answering!

I want to ask not because speed is everything, but if I can get to a place where I do 1,000 reviews in an hour, then I can triple the amount that I'm learning now. Not that 1k/hour is possible or even the goal, but anyway, the more efficiently you review, the more materiel you can cover, the more materiel you can use. My average for reviews is 90%, though on days when I add a bunch more cards than normal, it might get as low as 75%. I've heard 70-80% is the sweet spot, where you're learning enough that it's challenging, but not enough so that your retention really slips. For right now my goal is being able to read, and spamming this kind of practice (in the described format) is perfect for that. If I can trust that I can do 500 reviews in an hour, with my correct rate above 70%, then I can comfortably add more cards than I currently do. So I want to open up this discussion. I've learned so much about card making lately - I was using production cards with full sentences and English definitions until recently! (Not that that format wouldn't be perfect for someone in a certain situation, but in my case, changing my format alone has made me so much quicker with reviews and yet it has no impaired my ability to use the reviewed knowledge at all.)

[RawToast]

This depends heavily on the card layout, so from my Anki stats:

RTK (recog) 6.5 cpm ~ 390 cards an hour, but I doubt I could keep this rate up. My RTK reviews are only ~8 min long!
Vocab (single item recog & prod) 10.2cpm ~ but this averages 77 items in 7.6 minutes, I assume 100 in 10 minutes.
Sentences (mix prod/recog) -- 4.7 cpm. The sentences are small, as these are grammar examples.

My decks are small after deleting a ton of cards due to some Anki sync issues. So the speeds are probably higher as I can finish all my reviews in 24m (19m if I studied every day, but I've been shattered thanks to work recently).

Quote:I was using production cards with full sentences and English definitions until recently! (Not that that format wouldn't be perfect for someone in a certain situation, but in my case, changing my format alone has made me so much quicker with reviews and yet it has no impaired my ability to use the reviewed knowledge at all.)

I am currently doing production with both sentences and single items (I know this has issues with similar words). I have been rushing the vocabulary in the JfBP books (25 a day), since I need to familiarise myself with the vocab used in my evening classes. The classes are output heavy, I can read everything given to use but not produce it -- hence the production cards!

I've done a few E->J sentence cards, which were quite hard and slow to complete; however, I found these more useful than the traditional close delete cards. These sentences were basically the exercises from the books, I guess this approach becomes less effective with complex sentences

Jessem what format have you changed your cards to?

jessem Wrote:Dang, nice! That's some decisive answering!

I want to ask not because speed is everything, but if I can get to a place where I do 1,000 reviews in an hour, then I can triple the amount that I'm learning now. Not that 1k/hour is possible or even the goal, but anyway, the more efficiently you review, the more materiel you can cover, the more materiel you can use. My average for reviews is 90%, though on days when I add a bunch more cards than normal, it might get as low as 75%. I've heard 70-80% is the sweet spot, where you're learning enough that it's challenging, but not enough so that your retention really slips. For right now my goal is being able to read, and spamming this kind of practice (in the described format) is perfect for that. If I can trust that I can do 500 reviews in an hour, with my correct rate above 70%, then I can comfortably add more cards than I currently do. So I want to open up this discussion. I've learned so much about card making lately - I was using production cards with full sentences and English definitions until recently! (Not that that format wouldn't be perfect for someone in a certain situation, but in my case, changing my format alone has made me so much quicker with reviews and yet it has no impaired my ability to use the reviewed knowledge at all.)

The correct rate also depends on how you rate your answers.

It seems to me that 75% is the sweet spot if you're really strict with your answers (if you fail everything that even so much as makes you think). If you're more forgiving when you answer, and give yourself a little time to think on difficult questions, then you should probably aim for the mid 80s.

Personally, I like to take a little time to think about difficult questions, maybe replay the audio if it's a tough sentence, maybe look at the individual Kanji and try and figure out the word based on the English keywords, if it's a vocab item. And aim for 85%. But it's plausible that failing difficult cards right away, and getting a 75% rate as a result, with the same exact material, is more effective. I'm not sure.

What I AM fairly confident about is that doing both (taking your time for hard questions and passing them, and still aiming for 75%), would be too overwhelming (the material would be too difficult).

jessem Wrote:...if I can get to a place where I do 1,000 reviews in an hour, then I can triple the amount that I'm learning now. Not that 1k/hour is possible or even the goal, but anyway, the more efficiently you review, the more materiel you can cover, the more materiel you can use.

It doesn't quite work this way. If you do your reviews faster, your accuracy rate will necessarily go down. Depending on how many learning steps you have, each failure will count for several more repetitions. These extra repetitions can really start adding up. For instance when you do 100 reviews with default anki settings and 90% accuracy, you should end up doing ~124 reps. If you do 100 reviews at 75% accuracy you'll end up doing 196 repetitions(assuming you got 75% accuracy on subsequently failed cards). And if you have more than 2 learning steps, it adds up even faster.

This isn't to say that you shouldn't try going faster, just that it's not as easy to hack memorization as it may seem.

It really depends. On my Optimized Core 6k, which is only reviews (I finished the deck months ago), I usually need one metro ride (15-20 minutes) to do half of my 300 reviews. On my Chinese HSK deck (I'm a beginner), around 70-80 cards (including 10 new ones) require half an hour. My Core 10k - 6k cards are a pain as well.

All my decks are english meaning-> either Japanese kanji and kana or Chinese hanzi and pinyin.

My opinion is that review speed or accuracy are not essential: take the necessary time to remember a flashcard rather than trying to fast-check it, real usage is better for increasing the speed (as far as long-term memorisation is concerned, let the SRS magic do its thing). Also, don't be rushed about new cards, I strongly believe that getting your reviews under control is more important than accuracy (what do you want to do about accuracy anyway? Rote learn words? It would be against the idea of SRS).

Note, in my case, my Japanese Core decks are both late and I couldn't care less. My priority is (i) patiently learning chinese and (ii) not wasting my Japanese in the meantime (so new Japanese words from the Core 10k are presented only if I have enough time).

In any case, I believe 1k reviews is way too much. If you have so much time, use it to read or listen to real life Japanese, or to read a Japanese grammar book (note I say "read", nothing more, I believe actual learning comes from real life exposure once you are aware of the rules).

Well, I just threw the 1k number out there. I hope to never have that many reviews. There was another person on this forum who mentioned they add 100 new cards a day. I was adding 30-50 a day for a complete month, but the reviews stacked up and I had to stop after 40 or so days of that. But I've since changed from production to recognition and my reviews are much faster now, so I'm thinking I can attempt that again. I dedicate up to one hour of every day to anki, and a minimum one hour every day for Japanese media. I'm just trying to figure out how to make the most of that anki hour.

Yogert, you made a good point. I don't really understanding the learning steps thing in anki. I don't think I understand anki very well at all, really. But anyway, fails and what aside, I'm keeping track of how long my reviews take me every day (in real-time, not in anki's measuring). If the time increases but the number of reviews don't increase with it, that will be my warning sign. As-is, for the past 2 weeks, my accuracy has not suffered, my review time has halved, and I'm making much more progress than I have in months.

Here's what 128 reviews (with 6 failures, so total 134), taking 34 minutes, on Sunday afternoon looked like:



(RTK practice, keyword + annotated keywords (+ story in very small text) to kanji. All tested kanji should be on there. Plus some other stuff from non-Anki practice on the lower-left. Note I rotated the sheet of paper from portrait to landscape mode after the first column: desk space ergonomics.)

jessem Wrote:I don't really understanding the learning steps thing in anki. I don't think I understand anki very well at all, really. But anyway, fails and what aside, I'm keeping track of how long my reviews take me every day (in real-time, not in anki's measuring). If the time increases but the number of reviews don't increase with it, that will be my warning sign. As-is, for the past 2 weeks, my accuracy has not suffered, my review time has halved, and I'm making much more progress than I have in months.

That's great! Doing the same thing in half the time is good no matter if it fits the theory or not. Different things work for different people. Anki is prettu good on default settings so knowing the intricacies of anki is definitely not required. If you find yourself failing brand new cards a lot, you may want to adjust your learning steps. It just means that you will get more reps on new cards the first day. Default is '1 10' which means if you pass the card you'll see it again in 1 minute and again 10 minutes later. Some people find that adding another step or two helps remembering the new cards better. My steps are set to '.5 3 10 60' to give me a little more exposure the first day so I get less failures for learning cards.

[RawToast]

yogert909 Wrote:If you find yourself failing brand new cards a lot, you may want to adjust your learning steps. It just means that you will get more reps on new cards the first day. Default is '1 10' which means if you pass the card you'll see it again in 1 minute and again 10 minutes later. Some people find that adding another step or two helps remembering the new cards better. My steps are set to '.5 3 10 60' to give me a little more exposure the first day so I get less failures for learning cards.

I can't believe I didn't know this! I've just updated all my decks to 0.5 3 10 and will see how it goes.

yogert909 Wrote:If you find yourself failing brand new cards a lot, you may want to adjust your learning steps. It just means that you will get more reps on new cards the first day. Default is '1 10' which means if you pass the card you'll see it again in 1 minute and again 10 minutes later. Some people find that adding another step or two helps remembering the new cards better. My steps are set to '.5 3 10 60' to give me a little more exposure the first day so I get less failures for learning cards.

Oh, that's good to know! I only had one 10-minute step in my Options, but I was kind of imposing extra steps on myself manually, because I always fail a new card the first time. Then later on in the day, I do "Review Forgotten Cards," which is all the new cards that day as well as any cards that gave me trouble. So I was kind of manually creating .5 1 10 60 steps on my own!
This should help my fail rate a lot, although I do like being able to do a Review Forgotten Cards cram at the end of the day. It's been a nice catch-all practice tool.

Haha yea it only took me a year and a half to try that out. But I'm happy I did.

If you like the 'review forgotten cards', it's essentially a filtered deck set to 'rated:1:1' so you can play around with that too. For instance, you can make a filtered deck with 'rated:1:2' and set the learning steps to 1. That'll pull in all the cards you got wrong in the past 2 days and you'll only see them once. You can search by deck, interval, # of reps...the possibilities are endless. But I don't use that as much anymore since progressively increasing learning intervals just makes a lot more sense.

Vempele Wrote:About 4 minutes. And 21 days or greater.

Do you mind outlining your routine study method? Or directing me to a link of a previous post if you've explained it before?

I'm currently at the ~5000th card in Core10k optimized i+1 (recognition only). My overall average is about 3s/card (and rising) BUT this is with me rating hard/fail on every card with the initial ease set at 130%. And I generally don't look at the example sentence at all.

I'll describe my current routine for you or anyone else can offer advice on.

My current routine is:
1) Commit the new cards I'm learning that day into my short term memory. When I do learn cards, it's usually the next 33 new cards in the Core deck. This is done by spending about 30-50min to write the vocab, the meaning in English and the reading in hiragana down once to commit it to my short term memory. I then glance through the list I wrote down a few times throughout the day (maybe 3-5 times over 2-6 hours).
2) Towards the end of the day, or when I think the new cards are in my short term memory, I review all the due cards + the new cards mixed together

During reviews, I:
1) Look at the vocab, think/say the reading + think of the meaning in English or simply "visualise" the meaning.
2) Click next, glance at the English translation for the vocab, rate either hard or fail.
The audio for the vocab will finish playing on the next card. Sometimes I realise I got the reading wrong so I have to Ctrl-Z to go back so I can fail the card.

I tried adding two 2 hour steps to my new cards + lapsed cards before but I eventually ditched it and stuck with the default steps of 1-10 for new cards. Recently, I started getting annoyed at cards I was frequently failing (since the cards had vague/abstract meanings or strange readings that I haven't encountered much, if at all) so I changed the lapsed card step from simply 10 to 10 1440 1440 1440 1440 1440 1440 (you can see when I implemented this by the massive increase in the red bars here http://i.imgur.com/kPvF2js.png).

I suspect some flaws in my routine are:
Spending time actually writing down the vocab + meaning + reading (only once though) while learning new cards could be an inefficient time sink. It's just that I tend to remember things better when I write things down/type things up as opposed to just passively reading something (I know this varies widely from individual to individual).

Setting 10 1440 1440 1440 1440 1440 1440 as steps for lapsed cards in an attempt to brute force it into long term memory. Perhaps I should just suspend the leeched cards? My first leech card was 職場 which is quite firmly in my memory now (or so I hope). Before, I kept failing it every now and then because I read it as しょくじょう. Perhaps I'll suspend leech cards with more political terms but for more standard words like 職場, I'll keep trying until it sticks.

Perhaps the starting ease is too low (it's at the minimum lol) but I'd rather keep review times/card short. It's more motivating and feels more relaxing/easier to me this way. Even if it boosts my overall review time.

Am I the only one who doesn't suspend leech cards? Lol.

And yogert, you prompted me to change the thing for lapses. I changed mine to 1 5 10, seems to be working much better for me.

I was a hard convert to to suspending leeches too until I read on supermemo's theory page that if you eliminate the most difficult 10%, you can increase your learning speed by 300%. Then I went through my cards in the browser and noticed that over 68% of my mature cards have zero lapses. I also noticed that the vast majority of my cards with several lapses never make it to become mature. Now I have my leech threshold set to 5 lapses instead of 7 and I'm considering going lower. If you're like me and going for volume, suspend those leeches and come back to them once you've picked the low hanging fruit. And it makes for a more pleasant learning experience when you don't keep running into those frustrating cards.

Also, setting the initial ease to 130 has you seeing the majority of your cards way too often. Of course every time you see the words, you'll remember them a little better, but it's far from the most efficient way to learn volume. You want to get the easy cards out of your way to longer and longer intervals so you have time to learn more new ones.

Setting your easy so low and hitting hard every time is making your accuracy really high which is great, but it's slowing down the quantity of new cards that you can add. According to the supermemo page, peak efficiency is in the 70-80% accuracy range. Above 80% accuracy and your efficiency really starts dropping off fast.

Quote:The greatest overall knowledge acquisition rate is obtained for the forgetting index of about 20-30% (see Figure 3). This results from the trade-off between reducing the repetition workload and increasing the relearning workload as the forgetting index progresses upward.

Interesting stuff.

I'm currently going through Core6k as part of my JLPT2 study (and uh yeah, my vocab has always been very lacking due to my previous inability with kanji), so I feel like I don't want to suspend cards when they might be needed for the test. I am kinda freaking out about it, lol.