Anki 2 Beta 1 is out now

[querido]

Anki 2 Beta 1 is out now.

(As usual the author warns that it isn't fully stable yet.)

Many of the changes in Anki 2 don't matter to me (at first glance), and learning these takes some time. I've spent part of two days on it.

Here are the features that matter to me enough to switch now:

It has better learning and cramming modes.

New cards and missed cards can be made to follow a schedule with a configurable number of steps of configurable duration, e.g. (in minutes), 1-10, 1-10-60, 3-9-27-81, etc., before being promoted out to 1 or more (configurable) days. This schedule can be escaped at any time by grading Easy, which schedules the card out to 1 or more (configurable) days.

It has a cramming mode with an interesting new feature.

A cram deck behaves a lot like a "smart playlist" (except that it is reloaded upon pressing a button instead of automatically); it is loaded with cards according to a search string, the same string that is used to search for cards in the browser, e.g., "deck:Intermediate is:leach".

The cards here can also be made to follow a schedule as above. When they graduate from this session they just go back to their regular deck; keep cramming until they're all gone. If they still satisfy the search string they can be called back for more cramming by pressing a button.

For example, this string "is:leach" would pull all of your leaches together from all of your decks at any time. (And the number of misses that defines a leach is configurable too.) So, you could try to clean out your leach collection with as many cramming sessions as necessary, very conveniently.

[feng]

Are you sure its not stable?

I was under the impression that versions of software labeled alpha versions were generally not guaranteed to be stable, and beta versions are supposed to be complete stable versions getting some final tests before their final release.

[afritzse]

Alpha versions can be unstable, and they do not include all the features of the final version. Beta versions are supposed to include all the features of the final version, but they aren't guaranteed to be stable. Of course, not even the final version has to be 100% stable. HTH.

[feng]

Of course, not even the final version has to be 100% stable.

Dude! I hope you don't work for my banking institution!

[afritzse]

Rest assured, I don't. But Anki has never crashed on me, I just wanted to correct the misconception that software could be completely error-free.

[feng]

This attitude is something that users learn, software companies are quite happy that users incorrectly believe these sorts of things. Software can be completely error free. Its just a matter of the developer being competent and following proper procedures.

For example, when you go to the hospital and have your medicines drop dosage controlled by a computer, how sure do you think people are that its going to give you the correct dosage?

The difference is (for example) in the medical field, every possible input and every possible output of every piece of code loaded onto that computer is tested to work correctly.

[afritzse]

Well, it's not enough for each individual piece of code to work correctly, you'd also have to check for all possible states the pieces of code together can be in. But that's a little off-topic, and I don't claim to know the latest research, nor do we have to.

[Silent]

This attitude is something that users learn, software companies are quite happy that users incorrectly believe these sorts of things. Software can be completely error free. Its just a matter of the developer being competent and following proper procedures.

This is obviously just theory. Sure every bit of code can be 'perfect' (though you basicly can only 'proof' it for meaningless small parts of code). If you demand a meaningfull peace of software to be 'perfect' the costs are unbearable. And what's the point? Even if your code is perfect and contains no bugs it will still fail as the hardware it runs on is not perfect. And even more, the people operating it are not perfect etc etc. No matter what, it will fail sometime in some way.

You have to find a balance between costs, utility, flexibility , chance of failure etc. Bugs are perfectly acceptable.

[querido]

Y'all are awesome. Get out of my thread. Just kidding.

Setting the steps too close together (1 2 4 ...) seems to clog things up.

But this feels pretty good (for single-word cards): 1 4 16 64 256.

My whole-lesson cards with audio will need wider spacing because I often spend minutes on each card. I haven't tried it yet.

This learning/cramming mode works especially well for cards that are hard because they're close together in meaning.

I feel optimistic; I needed some help!

[Silent]

Ok, back on topic.....

Is there anything changed in the spacing algorithm itself? I tend to be annoyed by the fact that spacings seems to be based on the planned spacing instead of the actual spacing.

E.g. if I review early because I will be busy. It may be that I review the card at a two day interval instead of a 4 day interval. For calculating the new interval anki still seems to use the 4 days while the spacing has been only two days. Same when I'm late with reviewing. Spacing may be intended at two days but if I review it only after four days it still seems to calculate the new spacing based on 2 days while I still remembered it after 4 days.

I'ld say when correct the new spacing should be max (calculated based on actual spacing, previously calculated spacing)

[querido]

I believe Anki handles both of those cases well, but I don't have time to study it again.

The logic is tricky but the author always seems to make the right decision when I look into it.

[realmayo]

Silent: I certainly think in the second case (you delay a review) you are mistaken: if I have a card due in a few days but I leave the deck for a week, the new interval will be more than a week. In the first instance I have no idea, I've never reviewed early.

My concern with the new Anki is that some of the changes will stop me doing what I have been doing happily until now. But I haven't looked in any detail (though moving away from separate decks to just one big one makes me nervous). But as you say querido, the author seems to make the right decisions -- I've been using it for four or five years now and evey time there has been a big change I haven't liked it initially, but ended up liking the programme just fine once I'd got used to it.

[Silent]

Silent: I certainly think in the second case (you delay a review) you are mistaken

I may be mistaken for the early reviews. It has been a long time I've done early reviews. For the late reviews it's absolutely true. I tried with an inactive deck. Haven't used it for 3+ months. If I start reviewing many of the cards have only 2 to 3 months spacing for the hard button. So, less than the actual interval between the reviews. I checked deckproperties and there for button 2 it says 1-1.1 times. If actual spacing was used for the caculation button2 should have at the least well over the 100 somthing days.

[querido]

:-(

[feng]

Haven't used it for 3+ months. If I start reviewing many of the cards have only 2 to 3 months spacing for the hard button.

If you havent seen a card for 3 months, and you can still remember the card, doesn't it deserve to be pushed 2 or 3 months ahead?

I'm just wondering what do you expect it should do?

[realmayo]

Silent, what about Button 3? That is the most relevant one surely.

I have an inactive deck which I've not opened since 17 days ago, and the first card from that was due after only one day, so it's 16 days late.

My options are: B2: 6 days; B3 22 days; B4 1.8 months. Seems correct to me.

Also: when you say "I checked deckproperties and there for button 2 it says 1-1.1 times", how did you do that? I don't have any mention of button 2 in my deck properties, except for the "initial button 2 interval" option, which is not relevant here.

[Silent]

If you havent seen a card for 3 months, and you can still remember the card, doesn't it deserve to be pushed 2 or 3 months ahead?

I'm just wondering what do you expect it should do?

I expect it to be pushed ahead at least the same time as it has been the last time that I saw the card. The last time I saw it was well over 100 days. So I expect it to be pushed ahead at least 100+ days, not 'only' 2-3 months.

Also: when you say "I checked deckproperties and there for button 2 it says 1-1.1 times", how did you do that? I don't have any mention of button 2 in my deck properties, except for the "initial button 2 interval" option, which is not relevant here.

That's the one I meant. Now you say so, you're probably right that it's not too relevant here. However I don't see the relevance of button 3 and 4 here which are indeed much larger. However this is 'always' the case as button 3 tends to be initially 3-5 times and button 4 7-9 times. This changes based on the buttons you've pushed before. IMHO buttons 3 and 4 only become much smaller if you've pushed hard/incorrect several times before. As I never studied the deck very intensively there are no cards where this is the case.

What I've understood from SRS is that it shows the card 'right before you are likely to forget it'. If I remembered it, though hard, after well over 3 months, what would be the ratio to lower the spacing to just over 2 months? This is a decrease of more than a month, one third!!! of the actual spacing. The only reasonable explanation I have is that the spacing is calculated from the spacing stored rather then the actual spacing. If you keep up perfectly there's no issue. Even with a little sloppyness it's not really a big deal though it would save time if actual spacing was used. However when you failed to do your dues for a while and want to catch up it makes it a lot harder. Many cards may show up again far too early and interfere with the cards that really need to be studied (again). Sure, you can compensate a bit by pushing the easy button instead of button 2 or 3 but meddling with the algorithm in that way won't give the optimum result either.

(please note I don't know the exact figures, if I look at the deck graphs the last time I studied the deck was between 100 and 150 (estimate 110-120) days ago. Since about 200 days ago I studied it only 3 or 4 times, so there's a good chance the actual spacing is even larger then half a year. Display order is review cards in order due)

[realmayo]

Silent, what about Button 3? That is the most relevant one surely.

You pressed Button 2. Which meant it was hard for you to remember it. Just guessing, but it looks like if you review a very delayed card and press Button 3 you'll get the result you are looking for (a long interval taking into account the delay). But if you find the answer difficult and press Button 2 you get a much shorter interval. If that's a deliberate finessing of the algorithm by Anki I think it's a very intelligent one. I have an inactive deck which I've not opened since 17 days ago, and the first card from that was due after only one day, so it's 16 days late. My options are: B2: 6 days; B3 22 days; B4 1.8 months. Seems correct to me.

the "initial button 2 interval" option

The clue is in the word "initial": these settings are only relevant for the very first time you see a card (ie the initial review).

[querido]

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. You could post at ankisrs-users. He will listen to coherent reasoning. I can imagine a change to the code below that would accommodate you. Can you?

Mnemosyne does not apply this adjustment, which you could learn by reading the code. You could switch to Memosyne, which is now at beta 9. In addition to his google group(s) he posts about it often on google+. It's nice. Start a thread.

"

if ease == 2:

interval = (interval + delay/4) * 1.2

elif ease == 3:

interval = (interval + delay/2) * factor

elif ease == 4:

interval = (interval + delay) * factor (etc, etc...)

"

feng:

-If it is known to be perfect why would it be labeled beta?

-If it is not known to be perfect then the author feels a duty to warn the user.

I am using the beta myself and wanted to tell you about it, but I felt a duty to pass along the warning.

Also, in this context, "unstable" does not only mean it might have errors; it means the feature-set could change.

Lastly, you asked: "Are you sure its not stable?". Who were you asking? I did not say it wasn't stable. I said the author said it wasn't stable (finished).

This is a great thread. Thanks everybody! :-(

[querido]

realmayo:

I'm using it and I would answer any questions. But the documentation is very good too.

Thanks for helping the thread! ;-)

Edit:

My latest steps are (minutes) 3 15 60 180 360 for new cards and 60 180 360 for lapses.

I have adjusted these just "by feel". It seems to be working well!

Edit:

I repeat: it creates a rep for every one of those steps, but the extra ones are cancelled if the card is graded Easy.

With few exceptions, so far I like just pressing the space bar.