Good morning everyone.
I've been using Anki for reviewing my 750 cards. I've been using it for a month but with standard Anki settings.
Could you please give your own Anki settings that you use and are happy with, or write some recommendation on what should I keep eye on?
I know what some settings may be more of a mater of opinion but I'd like to get more of a optimal experience (intervals - 4 days is default and so on).\
Thanks.

The only thing I can say is I change the "easy interval" based on how many reviews I have in the next days. I have 3 as a default value, but if in the statistics I see in that day I have already 60 reps to do, and on the 4th day I have only 20 reviews, I change that value to 4 so the new cards will be reviewed on the 4th day from today.

On the starting ease I have "229" instead of "250", I don't remember how I came with that value but I'm sure it works better for me.

The Load Balancer plugin will handle doing that for you cophnia61. It's here:
https://ankiweb.net/shared/info/1417170896

I would suggest not bothering too much with the card lapse intervals or getting too crazy about setting super granular 5, 10, 30, 180 minute intervals. I actually have them disabled now so cards I miss just come up again tomorrow. I don't even bother looking at them again today. It cuts my review time down a bit, and I haven't noticed any decrease in overall card success rate.

If you want to look over the cards again later it's easy to see those failed cards again in the Anki Browser or make a Filtered Deck to look them over again if you have time and feel the need to.

Also, after only a month, you probably haven't seen tons of cards become mature yet. The number of cards due per day will subside a lot as they get pushed further into the future. It seems like you're doing fine if you're managing to keep up with it every day.

erlog Wrote:The Load Balancer plugin will handle doing that for you cophnia61. It's here:
https://ankiweb.net/shared/info/1417170896

I would suggest not bothering too much with the card lapse intervals or getting too crazy about setting super granular 5, 10, 30, 180 minute intervals. I actually have them disabled now so cards I miss just come up again tomorrow. I don't even bother looking at them again today. It cuts my review time down a bit, and I haven't noticed any decrease in overall card success rate.

If you want to look over the cards again later it's easy to see those failed cards again in the Anki Browser or make a Filtered Deck to look them over again if you have time and feel the need to.

Also, after only a month, you probably haven't seen tons of cards become mature yet. The number of cards due per day will subside a lot as they get pushed further into the future. It seems like you're doing fine if you're managing to keep up with it every day.

Installed now, thank you for the suggestion! This is going to save me much effort!!

under the lapses section i set my "new interval" to 20%... have been doing this for about 3 years now.

basically it nerfs your fail button so that you don't go all the way back to 1, 3, 6 days after failing a 3-month card. saves a lot of time.

the potential downside is that occasionally you might actually need to start from scratch for a card that you forgot after 3 months. my reply would be that, if you're in such dire straits re this card, you'll probably fail it again after two weeks, and then you will be back to 1 day again... so it's kinda the same thing.

personally i would really prefer just to have a 6-button system like supermemo... with a weak fail AND complete fail option. anki doesn't give us that... only complete fail, or various levels of interval growth.

damian is determined to give us 4 buttons (or fewer sometimes for some weird reason) so we have to be creative.

dtcamero Wrote:personally i would really prefer just to have a 6-button system like supermemo... with a weak fail AND complete fail option. anki doesn't give us that... only complete fail, or various levels of interval growth.

This would be perfect :/

I find that setting my 'easy' interval to 2 days makes me more inclined to use the easy button, though 2 days is too short for some cards.

A selection of easy intervals would be ideal ...

cophnia61 Wrote:

dtcamero Wrote:personally i would really prefer just to have a 6-button system like supermemo... with a weak fail AND complete fail option. anki doesn't give us that... only complete fail, or various levels of interval growth.

This would be perfect :/

anotherjohn Wrote:I find that setting my 'easy' interval to 2 days makes me more inclined to use the easy button, though 2 days is too short for some cards.

A selection of easy intervals would be ideal ...

Op, sorry for asking a somewhat-related question: could @anotherjohn, @cophnia61, @dtcamero, and others comment on other aspects of Anki they find less than ideal? I ask because there's been a lot of research done over the years in the technology-aided learning branch of experimental psychology that suggests ways to complement or replace Anki/Supermemo's approaches. Certainly auto-evaluation is something Anki should provide more than The One Way, but that's something I doubt will be easy to get Damien to modify unless we write a whole new scheduling engine plug-in that can consume more-finely-discretized (or less-finely! binary?) auto-evaluations. From these three quotes I am guessing there is some demand for such a thing, hint, hint, plug-in writers.

Another item on my Anki/SRS wishlist is interfacing Rikaichan/WWWJDIC/etc. to SRS so the latter knows what you've been reading, and more importantly, what you haven't been reading (creepy, right).

[s0apgun]

Anyone know if that plugins functionality will translate to Ankimobile?

No, only reviews you do on the computer will still result in more balanced scheduling in the future.

What the addon does: when you review a card, it gets scheduled to the "easiest" day within acceptable distance (percentage and absolute number of days) of the ideal interval. The "ease" of a day is a function of the number of cards scheduled for that day and their average ease.

Overall I think anki's defaults are pretty sensible and don't need to be tweeked too much. However here are a few things I do that make it work a little more to my liking:

dtcamero Wrote:under the lapses section i set my "new interval" to 20%... have been doing this for about 3 years now.

What he said. It makes me less hesitant to fail a card with a long interval. And cards with an intervals like 6 months really shouldn't go back to one day just because you (temporarily) had a brain lapese.

After a few months you can see what your deck's average ease is. If you change the remaining cards' starting ease to this number, learning will be a little more efficient.

Some people like to add and extra learning step. It just gets those new vocabulary words into shot term memory a little better.

I use a filtered deck "rated:1:1" to review the cards I got wrong today. Sometimes I use "rated:2:1" for failed in the past 2 days. This is to get more practice with the cards that are giving me trouble.

As of this past month, I just put my settings back to the original ones, with one exception: I un-ticked the "Show review times above answer buttons" option. Highly recommend doing that, my review sessions feel faster and better focused now.

dtcamero Wrote:personally i would really prefer just to have a 6-button system like supermemo... with a weak fail AND complete fail option. anki doesn't give us that... only complete fail, or various levels of interval growth.

damian is determined to give us 4 buttons (or fewer sometimes for some weird reason) so we have to be creative.

I've used 6 button systems and I'm glad anki has only 4 buttons. I found when using 6 buttons, I ended up spending an extra second thinking about which button to press. Those seconds add up over hundreds of cards per day. Probably more than offsetting any small increase in efficiency from better informed algorithm.

aldebrn Wrote:...comment on other aspects of Anki they find less than ideal? I ask because there's been a lot of research done over the years in the technology-aided learning branch of experimental psychology that suggests ways to complement or replace Anki/Supermemo's approaches. Certainly auto-evaluation is something Anki should provide more than The One Way, but that's something I doubt will be easy to get Damien to modify unless we write a whole new scheduling engine plug-in that can consume more-finely-discretized (or less-finely! binary?) auto-evaluations. From these three quotes I am guessing there is some demand for such a thing, hint, hint, plug-in writers.

I would love it if Damien would provide an interface to roll your own algorithm. He's expressed to me little interest in developing the algorithm, so why not encourage others to innovate while he focuses on other functionality.

Another thing I would love to see is a way to export learning data to excel. The graphs in anki are ok, but there are a few metrics that I would love to be able to analyze myself. So, why not give us access to the raw data in an easily accessible format like xls or csv and let us go to town with it.

As mentioned above, I wold love to see load balancer functionality in anki mobile. I do 99% of my studying on mobile and my workload gets pretty lumpy sometimes.

yogert909 Wrote:What he said. It makes me less hesitant to fail a card with a long interval. And cards with an intervals like 6 months really shouldn't go back to one day just because you (temporarily) had a brain lapese.

I think they should. You don't have brain lapses about reading words you know in a language you can speak, do you? If you have a brain lapse about something now, the next time you see it shouldn't be in a month.

Also, if you untick the show intervals option, you won't be hesitant about failing cards because you won't know their intervals. You'll be able to fully focus on the task at hand, which is reviewing Japanese words.

yogert909 Wrote:I've used 6 button systems and I'm glad anki has only 4 buttons.

I've found even using 4 is unrealistic. I can pretty much only use 2 without thinking (fail with one, pass with space). Every time something's too easy, I have to pay attention to what I'm pressing.

I shouldn't have to be paying attention to that. Part of the problem is that the easy button is either 3 or 4, depending on the situation. If it was always 4, it'd be a little better. Any chance there's a plugin to fix that? (it would probably be trivial to make one, for someone who wrote Anki plugins before)

Stansfield123 Wrote:I think they should. You don't have brain lapses about reading words you know in a language you can speak, do you? If you have a brain lapse about something now, the next time you see it shouldn't be in a month.

I actually do forget words in my native tongue almost every day. Not recognition, but producing them in conversation, sure.

Japanese words I've failed with an interval of several months, I could probably pass them anyway and get the majority of them right several months down the line. Of the long interval cards I've had set back to 15%, I probably have close to perfect accuracy on subsequent reviews, so there's not reason to go back to 1 day. It works for my brain, maybe not for others and that's fine.

Stansfield123 Wrote:Also, if you untick the show intervals option, you won't be hesitant about failing cards because you won't know their intervals. You'll be able to fully focus on the task at hand, which is reviewing Japanese words.

I like your thinking on this, but I just can't bring myself to do it. I like seeing the intervals too much. It's probably the same reason why I spend too much time looking at the anki charts instead of actually studying. It gives me a sense of progress or something.

Stansfield123 Wrote:Any chance there's a plugin to fix that? (it would probably be trivial to make one, for someone who wrote Anki plugins before)

from anki.hooks import wrap
from aqt.reviewer import Reviewer

def learnOn4(self, ease, _old):
ease = min(self.mw.col.sched.answerButtons(self.card), ease)
return _old(self, ease)

Reviewer._answerCard = wrap(Reviewer._answerCard, learnOn4, "around")

Edit: basically identical to https://ankiweb.net/shared/info/992946134

Second on nixing "show intervals"...I hate to say this, but I felt hesitant to mark a card wrong when the next due date was 2 months or whatever.
If I take a break in the middle of reviewing, when I come back, I cover the first two screens so I don't know how many cards I have left.
Generally, I find it easier to concentrate when there's no clocks or ways to check your progress around.

Sorry to go off-topic, but:

yogert909 Wrote:Another thing I would love to see is a way to export learning data to excel. The graphs in anki are ok, but there are a few metrics that I would love to be able to analyze myself. So, why not give us access to the raw data in an easily accessible format like xls or csv and let us go to town with it.

I haven't forgotten about this!

Today's mini-project has been an experimental web interface to extract all your review information from your collection.anki Anki database into a CSV file.

Try it out at http://fasiha.github.io/fuzzy-anki/ using the second set of inputs, titled "View Reviews", to set a couple of options and upload your Anki collection database.

Please note: this is a client-side Javascript application, so you're not uploading your Anki database to anyone (other than yourself...), and your privacy is not at risk.

I recommend putting a limit until you really want to export everything because for my collection of 12'000 reviews, Chrome took ~30 seconds.

I included the card's underlying facts (its "note") as a JSON string in the last column even though that significantly adds size/noise to the file because with that, you have a complete review database you can import into Python/Excel/whatever (if you insist on not keeping it in Javascript ) and not need anything else from the Anki collection.

The file you download will have a weird name and no extension because I'm using Blobs, since I want this to work for power-users with many more reviews than me. It's just a UTF-encoded CSV file. I expect Mac/Linux to be able to open it automatically in your spreadsheet app, but in Windows/Excel, I found it was insufficient to add a ".csv" extension because Excel wouldn't auto-detect UTF-8 encoding and mess up my kanji. I found I needed to start with an empty sheet, go to Import Text Data, select this file, set it to be "Delimited" mode with the "comma" delimiter, and instruct it to use UTF-8 (65001, near the bottom of its list)---yes, the year is still 2014 Microsoft .

I wanted to add some visualizations (I made these in Python/Matplotlib last year https://twitter.com/fasihsignal/status/3...1304741889 and with D3.js one can make way cooler ones) but getting just the CSV going has taken enough of the night.

Sorry I don't have more documentation, I'll update the Github readme and fill things out a bit more with this reviews browser later.

aldebrn, this is seriously awesome! This is something that I've been dreaming of for the last year and one reason I even started learning python to try something myself. I don't know why, but I really like measuring my progress and this little tool you've built will allow me to see all kinds of things like many new cards I'm adding over time as well as figure out my forgetting curve for each deck among other things.

Quote:I recommend putting a limit until you really want to export everything because for my collection of 12'000 reviews, Chrome took ~30 seconds.

It takes about 2 seconds here but only finds 28072 out of 131751 reviews. And the Download CSV button is a no-op, but copypasting the blob: url to the addressbar worked. (Chrome and Firefox on Windows)

yogert909, thanks for your kind words. Please stay tuned, I have some very interesting plots and charts in the works.

Vempele Wrote:

Quote:I recommend putting a limit until you really want to export everything because for my collection of 12'000 reviews, Chrome took ~30 seconds.

It takes about 2 seconds here but only finds 28072 out of 131751 reviews. And the Download CSV button is a no-op, but copypasting the blob: url to the addressbar worked. (Chrome and Firefox on Windows)

Thanks for the bugreports!

1- I fixed the damned "Download CSV" button this morning and forgot to push it... it's now a link, and that should work

2- The 30 seconds was when it would automatically make the table in the browser, and generate the CSV. I split that up into two manual steps and as you saw, the CSV-generation takes no time. Good work, nice-json2csv author!

And most importantly,
3- Edit: bug all fixed!