Anki 2 Beta 1 is out now

[Silent]

Two or so years ago Anki's author adjusted the algorithm so that for delayed cards, grade 4 would get full "credit" for the delay, grade 3 one half credit, and grade 2 only one quarter. I remember studying the reasoning for this and accepting it; maybe you would not.

Ok, so apparently it's considered a feature. I really don't see the benefit of it, in contrary. I just checked a few cards. button 2, 2.2-2.7 months (much shorter then the actual spacing). button 3 8.3-9.9 months which is longer than the actual spacing (my guess from the deck graphs is that I started to study the deck about 280 days ago while 9.9 months is about 297 days.) AFAIK I'm unable to see the actual xeduled spacing. So can choose between much shorter (at least 35%) then the actual spacing or fair bit at least 6% longer than the actual spacing. In reality these figures are much larger as I only know that actual spacing is somewhere between 110 and 280 days which is a huge margin.

The deck I have here is quite small, 600 cards and the delay is huge so if I wanted to catch up the issue wouldn't be too bad. Button 2 cards would be delayed long enough to be able to catch up before resurfacing. However if you have a large deck with a young population and only a week or two delay things look quite differently. Cards you answered, though with difficulty, correctly, may resurface well before the backlog is gone and in less time than the actual spacing.

As said I don't see the ratio for calculating a spacing well within the actual spacing when you know the correct answer. I'll see or I can find the motivation without too much hassle. For me it's not a real big issue, as long as one keeps up with the deck it's even irrelevant. So I surely won't sign up somewhere only to start a thread and debate this issue.

[Silent]

I googled and found two explanations that contradict here an imho (mostly) correct approach is given.

Here the approach as given by Querido is mentioned including an explanation why this approach is so rediculous. It says:

When people return to their deck after weeks or months of no study, they’re often surprised by the length intervals have grown to. <cut> This is a good thing. If you have successfully remembered a card after a one month wait, chances are you’ll remember it again after a longer wait, too. The same principles which make SRS effective in normal use apply when you’re studying after a delay, too. It also makes little sense to schedule a card for 10 days in the future if you were able to easily answer it after a whole month’s wait - you’d be going backwards.

Imho it should be quite simple. Every button should have its own multiplier. At the moment you review a card the newly calculated interval should be max (actual interval*factor, stored interval).

What the actual multipliers of the buttons should be, or how they should be calculated is open for debate. I don't know, though I think that adjustment based on buttons chosen previously is a fair choice. Button 1 I feel is very tricky. Why go back from x months to zero just because of a short blackout? And if that's corrected by some button 4 pushing isn't that messing up the factors if they are based on the button pushing? If it's not brought back to zero and you really forgot it, will you see it back quick enough to relearn it or would it take ages?

To me it would at least be clear that the factor for button1<1. button2 should be 1 or slightly above, and button 3 and 4 should be well above 1.

[Guest realmayo]

If it's a card you know well and you have a short blackout then simply: fail the card! The next time you see it press Button 4 (ie easy), the time after that press Button 4 (assuming you really did know it well in the first place and it is easy), the time after that press Button 4, the time after that press Button 4. Next time you see the card will be a whole year from when you failed if (if my maths is right!) The "cost" of failing it because of that short blackout was four clicks of an easy card spread out over one year. Not a big cost at all.

Remember that in SRS the "R" stands for repetition. If you don't repeat stuff you won't learn it. So: say you have a delay of one month. There's a card in there that you've answered once or twice only. By your rules, if you just manage to remember what that card means, it still goes forward by two or three months. For a card you've only seen a couple of times and could only just remember. For me, that's clearly silly. Memory benefits from repetition: your suggestion would mean a big delay from doing your reviews mean certain cards would not go into your long-term memory. For other cards that you remember easily it's no problem. Hence the flexibility of the buttons.

I suggest you don't overthink about the process: just trust in the algorithm. If it doesn't work for you then quit or find an alternative. All the things in your post were discussed extremely thoroughly years ago and the solutions that were devised seem to please most people.

[querido]

You can write an algorithm plugin and make it do anything you want.

Until now I've always run my own algorithm in Mnemosyne and Anki, going back three or four years.

See that Python, above? Just edit it with a text editor.

Write a plugin!

[Silent]

The "cost" of failing it because of that short blackout was four clicks of an easy card spread out over one year.

I guess it depends on the value of the multiplier. I don't know how exactly it's calculated, but after enough misses the multiplier might be quite small and a fair number of repetitions may be needed. And if the multiplier jumps up big time with a couple of eases you may overshoot and it might become an oscillating card.

I suggest you don't overthink about the process: just trust in the algorithm.

I think this is a very bad advice. Don't think, just have blind faith......

You can write an algorithm plugin and make it do anything you want.

Without any doubt, but I don't think that the time and effort of learning how to build a plugin and the programming language needed to do so is a good investment for only a quite small benefit.

[querido]

Silent, which program are you using now?

[Guest realmayo]

... well if you want to think about every card you enter then you need a programme where, after answering the card, you tell it when you next want to see the card -- i.e. you type in the number of days. Seems like a waste of time to me.

[feng]

So now that its been out for a while, whats the verdict, I am starting a complete new / fresh deck, so it might be time for me to make the switch myself

(ill keep my old deck in the current anki for now)

[querido]

(Please don't post here unless it's about Anki 2 Beta)

Hello feng, I didn't see your post. *My* verdict is- I have switched to the beta.

Beta 2 is out with some bugs fixed. Let me tell you, I'm less disoriented with these bugs fixed!

With regard to the steps in learning and cramming modes, with experience I've had to increase the first step and decrease the number of steps. Otherwise, as I said, it gets too clogged with reps due. The steps I was using yesterday, 3 15 60 180 360, caused me to never actually finish my cards. All day! (But with just ten or twenty new words only, it works well.)

Anki 1 had one step; I almost always had it set to two hours.

So, having steps like this for new cards 15 60 240 is still a major help.

In fact, repeatedly missing a larger first step actually seems effective.

A group of new cards should probably be crammed aside from the daily reviews, and not mixed in.

You can just tag them "new" and create the cram deck with the search "tag:new".

It's unfortunate that this program is so hard to understand at first. But its flexibility is really nice, and as I said in the first post, it can now better serve as a place to learn new words (not just retain them).

(Please don't post here unless it's about Anki 2 Beta)

[Guest realmayo]

querido： given that I never use the cramming feature, would switching to Anki2 change how I actually do my daily reviews (the reviews themselves, not how my decks are organised etc)?

[querido]

No, not really. And there would be probably be several days of adjustment. There was for me.

The most important difference with regard to reviewing, is that you can set these steps as you like.

The starting factor (the basic interval multiplier, which has always been 2.5) is now adjustable. I had always used a plugin for that, and have always set it lower than 2.5, but after some years I don't really know where I should set it. Finally having the adjustment available, eh, it's set on 2.5 at the moment! (I hope the learning/relearning steps will give the degree of assistance I need.)

It now has a "Forgetting Index" you can set. (It adjusts the aggressiveness of changes to the multiplier so that over the long term your %right on mature cards moves toward X. This matters less than some people think.)

Its syncing is much improved. I use dropbox, myself.

The browser is very nice, but a lot of terminology has changed (even "deck" does not really mean exactly what it did). It's quite a bother at first. As I said above, I gradually accept the author's way of thinking, but I'm not sure how the masses will handle it!

[Guest realmayo]

I'm interested that people are confident enough to adjust some of these settings themselves: I think I'd be a bit scared to! I get around 87% of mature cards correct and have averaged that for years, and it's not too bad -- if I fiddled the forgetting index (assuming it does what I think it does) I could force that % number into the 90s, but I'd worry it would mean my daily time spent on reviews got longer.

I need to check out what any changes might mean for Ankidoid. Ankidroid takes far too long between cards on my large deck (30 seconds or so) so I can't use it for that.

I know what you mean about terminology changes and other changes, I'm sure it will feel like a nasty shock at first. But yes, I remember years ago when Damien made the first big-seeming changes to Anki and although I hated them at the start I've now either got used to them or much more likely have realised that they were for the best. So I'm trusting that will happen with Anki2 because -- one example -- I can't see at the moment how no longer having separate decks will makemy Anki use any better.

[querido]

"have averaged that for years, and it's not too bad"

Yeah, I wouldn't change that.

"no longer having separate decks"

I believe it was for the sake of syncing and portable devices.

I think some compromises in the GUI were made for the same reason. Fortunately, the pressure has been to keep the GUI simple and fast.

[feng]

I believe it was for the sake of syncing and portable devices.

I am curious what makes you think this would be the reason? Its not like syncing one file or multiple files is really that different.

The concept of multiple decks for me has never really made sense to me at all. There are just "things" you need to know. tagging and "decking" facts are just organisational constructs. One file, or many, what difference does it make ultimately?

In fact multiple decks to me doesn't make a lot of sense at all, you could download 3 different decks that have overlapping facts, learning the overlapping facts multiple decks is technically inefficient.

In fact, thats my primary problem with Anki two, all my duplicated obviously and clearly conflict with each other, and its hard to merge them back together.

[querido]

"I am curious what makes you think this would be the reason?"

I think he said so. I said "I believe" because I didn't care to google it or argue about it.

"The concept of multiple decks for me has never really made sense to me at all."

It's left-over from the time when people made their own decks with text editors, etc.

"you could download 3 different decks that have overlapping facts"

This is why it's good to have software, like Anki 1, that can automatically remove duplicates.

"learning the overlapping facts multiple decks is technically inefficient."

"its hard to merge them back together"

So why didn't you eliminate the duplicates in Anki 1, already? This isn't a problem with Anki 2.

[excler]

"learning the overlapping facts multiple decks is technically inefficient."

"its hard to merge them back together"

So why didn't you eliminate the duplicates in Anki 1, already? This isn't a problem with Anki 2.

what if you have different lessons with different answers, you don't need to place all the meanings of one word if they aren't used in this particular lesson.

and if you add audio you might want to have mandarin voice from particular mp3 file too

[querido]

If you really want to do that you can.

Anki 2 is at Beta 10 now. The bugs are pretty small.

The steps I'm using now are 30 120 480 for new cards (I'm cramming about 30 new words at a time; it's pretty hard!) and 120 480 for lapses.

With experience I've learned to use fewer steps and to make them longer. I've found that repeatedly missing a card at a too-long first step gives just the right kind of "oh I knew that" annoyance.

[excler]

If you really want to do that you can.

of course I really want, I used to store antonym, synonym, translation and explanation in separete desks and I store different accents sounds separately too, if I store all this information in one desk it is hard to learn them. Anki 2 seems doesn't allow dublicate facts or I haven't found this option yet as it was in Anki 1. I guess I need to create multiply profiles to be able to do so.

[querido]

Note type something like this:

1) question

2) antonym

3) synonym

4) accent1.mp3

5) accent2.mp3

etc.

Cards like this for example:

Antonym

1)

2)

Synonym

1)

3)

Accent1

1)

4)

etc.

Make an Antonym deck, a Synonym deck, Accent1 deck etc.

You could just use a different program that still uses 1-file=1-deck, but Anki makes it easer to rearrange all of those fields if you ever want to.

I put "really" in italics because I wanted to give you some advice, but good luck with whatever system works for you! :-)