Auto Advance Speed

#1

Skritter Android 336

The setting for Auto Advance Speed allows for two options to be ticked, without any explanation of what that does. It would be nice if the fastest of the two were for correct cards and the slower one for wrong cards - but that doesn’t seem to be the case.
What did I overlook?

Other random remarks:

  • Help files say I should go to Learn mode to add words. But in-app I don’t see any link/reference to Learn mode. Wish I could add words (1/5/10/20) straight from the Due cards
  • I second other members saying it doesn’t make much/any sense to place cards due because they were answered wrong at the beginning of a new session. User either has to restart session manually multiple times, or just keep forgetting the new items. A bit pointless either way.
  • Information on the homescreen isn’t the most pertinent I think. I would rather have it show the number of cards due, and a direct link to study due cards instead of requiring an extra click every time.
  • Why does the (gimmick) Time Attack keep flashing yellow?

Gripes aside, Beta looks a lot better than it did a year ago. To the point that having to redownload the old version might be the greater evil.

edit:

  • App has become a lot more power hungry, I estimate by a factor 5 to 10. Consumption level is given as 47%, even more than the screen. Going from never even coming close to having to recharge during the day, to having to recharge probably more than one per day. On a tablet that is used solely for Skritter. That would definitely be a reason to switch back to non-beta. Can the user do anything to make it behave better?
  • Have had to do the same ~100 reviews 3 times in a row (in the same order too). Hope that was a switch-over pain that won’t recur

edit2:

  • Going into any Recent Deck huge or tiny brings me Settings with a checkmark, and an endless spinning wheel next to Words. No idea what it’s doing and when it will stop. Same behaviour when accessed through the Decks tab. This also means I cannot add new words?
  • Recent Decks is so outsized with text and huge images that it fits only 2 of them in view, and only 6 on the main screen. Compare with a not-so-old web interface where it showed one deck per two lines, including relevant information on list status, total size and percentage studied.
  • Progress tab is all empty (Characters Learned: 0, Words Learned: 0, Total Reviews: 0 Time Spent: 0. All time and Today. The only information present is an activity feed has all somewhat relevant information hidden behind a drop-down

edit3:

  • No new cards are becoming due, while there would normally be around 10-15 per hour. Not sure if this a bug or if SRS has gone out of fashion.
  • Battery drain even worse than previously mentioned: Skritter is draining the battery almost as fast as I can charge it when it is only running in the background.

I will stop these edits. The app updated to beta by accident, and the experiment failed. Now I pray I can still switch back without having to do hundreds of reviews for the 4th or 5th time.

2 Likes
#2

When auto advance is disabled, there’s no additional option to set, however when auto advance is enabled, an additional option of “auto advance speed” will show, allowing you to select 6 different speed settings from 0.5 seconds to 3 seconds. This is how fast the prompt will auto advance after having answered a card-- correct cards will advance at the rate indicated, and failed cards will advance more slowly. (It scales throughout all the options). If this still doesn’t make sense or if this isn’t what you mean, would you be able to take a screenshot of what you’re seeing?

The Learn mode is accessed from a deck directly, by going into the deck (by tapping it’s name), and then by using the “Learn” button up top-- this will guide you through the parts of the word and add it directly to your review queue. You can skip this to just add it directly without going through the learning mode guide. The ability to add new words during a review session is something we’re mulling over and it is mentioned often!

The continuous mode works how you are expecting, failed cards will recycle in the same session more like how previous clients worked. The review mode will show failed cards at the beginning of the next session which is a logical system-- you’ll always see failed cards at the beginning of a session, and you can go through the reviews until you reach Zero. (Something not really before, which is pretty crazy!)

It does show the number of cards due! The review button will say “DUE CARDS (XX)”, where XX is the number of items due for review. We’ve talked about ways for the session configuration menu to not always pop up when tapping on review if you don’t like (keeping the previous settings), something we’re also thinking about. :slight_smile:

Well, I wish they weren’t both considered evils, but we’re glad to hear you’re liking the direction of the new version.

Thanks for calling this to attention. We’re looking into optimizations

Regarding the other things you mention, I hope they’re addressed in the latest update (3.3.7). If it’s not too much trouble, would you be able to followup after you’ve updated to let us know how things seem to be working?

Thanks for your feedback! :slight_smile:

1 Like
#3

Auto-Advance: I saw the list with six speed settings. And I could place tickmarks next to any two options at the same time.

Learn Mode: That wasn’t obvious to me, thanks for clearing that up. Maybe it didn’t become obvious to me because tapping a list (any list, any way) just got me stuck with spinning wheels.

Continuous Mode: I didn’t try continuous mode because I thought it would not necessarily study my due items, much like studying after the counter hit 0 in the nonbeta.
Disagree with your statement you couldn’t reach zero before, that’s simply not true. If you had wrong answers or items became due during your session that would increase the number a little bit during that session, as it should. Item due, item placed in queue, study item, repeat until learned. SRS was calculating due times in minutes, hours and days if I’m not wrong. If now the minimum interval is a day that would drastically change things. Note that it’s not just about failed cards but also cards that became due naturally. I did not see any new items in my queue for over an hour, that normally was around 10 per hour.

Due Cards on homescreen: that’s true. Other suggestions for homescreen still stand though.

I already switched back and can’t post screenshots anymore. That switchback took some time downloading and syncing, but was completed without having to rerereview anything: a great relief.

I will give it another try in Q3.

1 Like
#4

Ah hah. I see what you mean now, however I’m unfortunately not able to reproduce the same behavior. I’ll stay on the lookout for this though!

Ok, we appreciate it! An update was just released (3.3.7) which addresses some of this, but I know you said it took a long time to switch back to 2.1.2, and we don’t want to waste your time.

#5

@bezdomny

Agree about cards due regularly hitting zero in Legacy Skritter.

I study by time each day but I almost always manage to get each deck I review down to 0 cards to review within my study period anyway, even though I add new cards every day.

#6

I think what Jeremy meant was that hitting zero has been possible, but the way Skritter calculates dueness on the legacy app meant that things could become due again later in the day after you’re done studying. The only way to check that would be open the app again later and studying more (potentially).

In the latest apps the items due are getting fetched for the entire day in one batch. So, regardless of what time you study during the day you’ll be sure to have a chance to review all the cards that could be due, which you can study down to zero. The number isn’t suddenly going to change unless you add more cards for the day or do a Continuous Review session and get a bunch of things incorrect or something like that.

1 Like
#7

Thanks for clearing this up for me, that’s what I meant! There wasn’t really a way to get to a stopping point for the day since reviews keep popping up-- personally I would always have some more items due right after I would get my queue to 0, so it felt like there wasn’t a finish for the day.

#8

Is it fair to say that Legacy Skritter (Web version 1.0) and the current web version (2.0) are both on a continuous SRS mode, whereas some or all of the apps are now on an “overnight” mode, the same as ANKI? Also, is there any research or pedagogical basis for favoring one mode over the other? For example, are we supposed to “sleep on it” before reviewing some items again?

My personal feeling is that I get a sense of satisfaction from a lot of repetition during one session in order to get something really cemented in short term memory. I know that there is a theory (which I think is probably true) that the best way to get something into long term memory is to make the effort to recall it when it is almost-but-not-quite forgotten. I guess maybe an ANKI style overnight mode is closer to that but, I wonder if having had something well learnt in short-term memory doesn’t also have some benefit for adding something to long term memory through the review of nearly-forgotten items over a period of days and weeks.

2 Likes
#9

Yes, you’re mostly right. The main difference is that when you’re in a normal review session, things don’t become due in that same session, whereas in continuous mode, they do. That means that if you have a large number of items in your account, they will come due scattered throughout the day. Regardless what people say on the forum, this has been a major source of confusion for users over the years. In addition, things you get wrong also become due during the same session, whereas in normal review mode, they become due in the next session (but still the same day), so if you have 100 reviews to go through, finish all of them and get 97% right, you’ll have 3 more cards to review if you start another session immediately.

We’re working on better ways of handling things that you have forgotten or that you have learnt recently, but I don’t think that needs to impact the scheduling very much. Being able to go through recently missed cards, drill down, check example sentences, add mnemonics and so on, would be very useful, along with leech handling (i.e. dealing with cards that you get wrong time and time again).

The other big difference is that normal review mode just ends when no cards are due, whereas continuous mode lets you go on for as long as you like.

Regarding your question about sleeping before reviewing again, sleep certainly plays a role in the formation of long-term memories and there might be reason to believe that reviewing something late one day and then again the following morning is better than reviewing it once in the morning and then again later the same day, but my guess is that any such effects are greatly outweighed by other practical factors such as the student’s study habits in general (most people might not even study every day, so discussing when to study within one day is obviously of secondary importance ).

I think there’s no support for believing that it’s better to wait until something is just about to be forgotten before you review. More reviews lead to better recall. If you take a normal schedule for reviewing vocabulary, such as one that doubles the interval every time (just to use a simple, concrete example), so 1 day, 2 days, 4, 8, 16, 32, etc., we would expect better results if someone used another algorithm that increased with only 50% each time.

But that’s not the point, because it does not take opportunity cost into account. It looks at effectiveness and not efficiency. The reason we want to delay reviewing as much as possible (but not too much) is that it is the most efficient way, meaning that you learn as much as possible per unit of time spent. You could review a set of 1000 words every day and you would get really good at them after a while, but you would have no time to learn other words since doing 1000 reviews a day is quite a lot. The SRS algorithms, most of which are based on SuperMemo in one way or another, attempt to balance the wasted effort that goes into reviewing things you don’t actually need to review, and freeing up time to learn new things.

If you want to read more about scheduling and expected forgetting rates in general, this article over at SuperMemo goes into great detail in pinpointing how much you should be allowed to forget in order to learn as much as possible.

Most of the research I have looked at when it comes to over learning (i.e. when you keep focusing on the same material even though you already learnt it) says that it has little effect on long term recall. In short, studying something a few times extra in the short term does not really lead to better results in the long run, which is bad because it’s also a waste of time. This is part of the reason why writing a character on paper 100 times in a row does not guarantee that you know it a week later if you didn’t review it in the meantime.

For some areas which are not merely dependent on memory retrieval (such an pronunciation or handwriting), the story might be a little bit different, of course, but there’s ample evidence that shows that the spacing effect works for all kinds of learning, not just recalling the meaning of words or abstract facts.

Okay, so back to your question: Provided that there are ways to work on things you’ve learnt recently and that there are ways to handle recently forgotten items, both which are being worked on at the moment, I see little reason to prefer continuous mode from a pedagogical point of view. I understand that some people like it, but for the average user, I would recommend the normal review mode (that’s why it is the normal review mode, of course).

Studying things in continuous mode that are not due decreases the efficiency of the algorithm, and you basically waste time by being slightly better at what you already know at the cost of time you could have used to learn new things instead. Or read a book, listened to a podcast, and so on.

The fact that all cards that would have become due during the day are due at the beginning of the day in normal review mode is very unlikely to have any negative effect at all. The algorithm doesn’t know when you will forget something, and moving the time around within the same day is such a small difference that it can’t really matter. I would argue the other way, actually, that having them all due immediately at the start of the day makes it easier to plan and make sure you get through all of them that day, so it has a positive impact on study habits, which is more important than minuscule changes in when an item is due.

Okay, that turned out to be a little bit longer than I anticipated, but that’s what happens when someone asks good questions! I’ll see if I can’t turn this into a blog post or something more permanent. I don’t think many people will read it if it’s only available far down in this thread. :slight_smile:

1 Like
#10

Good to make this a blog post, especially if there is opportunity for comments. It’s a thoughtful review of the worthiness of different study styles.

Personally I do prefer the legacy study mode which repeats reviews of newly learned characters multiple times in a session, dropped in amongst older words needing review, over getting a whole batch of new words all at once.

But having both types of learning modes in the new app will be helpful for many people.

I think it may have to do with what level you are at; longer-term users seem to prefer the legacy (continuous?) mode, as it is faster and more interesting (being surprised by newly-adding words every once in a while while reviewing older ones) for people who already have a substantial vocabulary.

Newer learners will benefit from the drill-downs, given that so many character components and even radicals will be encountered for the first time. It’s a slower method but this is the stage at which character definitions and mnemonics can be added.

(Just waiting for those individual character mnemonics within words to be available before giving the new app another serious try. Timeline for that?)

1 Like
#11

We’re actively working on it as we speak and it’ll be ready for the 3.4.0. Barring any crazy patch updates it should be one of the next builds we release!

1 Like