It only got rid of the first 500 odd from my queue, leaving me another 275 or so. Is this on purpose?

I found myself skipping through a lot of variants on a theme this time. Various compounds ending with 边, 法语， 法文， 方法, variations on 去, 意思 and 没意思, 声音 and 大声 and 大夫. It would be better these repeated characters out more rather than me skipping through a lot - skipping is bad because if you didn't actually know something there's no way to tell you would have got it wrong, especially when it's the last character of a word and Skritter doesn't tell you.

Even with my skipping characters and the save me, I've got 150 odd characters still to review after an hour and a half or so's work. I personally think the scheduling needs tweaking to schedule things further in the future. I've used save me several times, but am still getting characters that I know very well coming up, and not many new ones.

My retention rates are a bit funny for the last four weeks. In order:

77.5%, 82.6%, 97.7%, 76.7%

Seems pretty random to me.

p.s. I'm using the appspot address, no loading problems as yet.

I disagree, I like the new system. It not only tests individual character knowledge but it also tests word knowledge.
for instance the meanings and pronunciation of 义 and 议 are very similar.
Using the new system I can actually learn which characters are used for the different words. (议题 or 义题×) (父母 or 夫姆×）

The overload you are having is caused by the change in system, I am also experiencing it. I recommend you don't use 'save me', all it does is move backup into the future. Its best just to keep working down until 0, even if this takes you several days.

I advise the developers develop a "Character learning mode" for those who want to focus on characters. I think many people want this as individual character writing is most people's weakness. After character knowledge is improved all will switch to word learning.

I'm also a big fan of the new system. One trick is to not recklessly add new words and characters - you have to be realistic about what kind of time you're going to spend, and spend every day or close to it. And as far as skipping characters within words, I think a good rule of them is that if I've written a character once in a given day, and I feel very comfortable with it, I feel free to skip it if I feel that'll make my study time more efficient.

I like the words! It just doesn't feel like the scheduling takes into account the fact that many words have the same characters in them and that it might be better to spread out words with the same characters a bit more, for efficiencies sake.

At this rate I'm certainly not "recklessly" adding new words and characters - it takes several hours to get through a review queue, many of which I feel I'm over-practising. Not that I've got a better idea, I'm just getting a bit frustrated by repeating the same characters. Even with "save me" I'm still overloaded, but it's often with characters I know already, that's why I keep using it: to get onto characters I don't know so well.

Why don't you simply skip the characters you know? I used to go through 150/200 reviews in 40min, and now I tend to skip the ones I know, so it took me 8min to go through 69 reviews this morning.

I am skipping them, but it's not much fun just going through a whole bunch of characters skipping half of them or doing them roughly because I know most of them for an hour. Didn't feel so much learning was going on during the process as when I first started with Skritter.

It's not that I don't like it, just today's practise was a bit boring...

>>Retention Rates: 77.5%, 82.6%, 97.7%, 76.7%

>>I feel that I'm over practising
>>I'm over practising
>>over practising

Not exactly sure your point ZachH. I'm making two points here. First: the Save Me button previously seemed to remove all items from the queue, yesterday it left a couple of 100.

Second: I didn't feel like yesterday's practise was as effective as it could have been, in part due to similar characters coming up repeatedly in different words.

When I have added new words, I feel like I under-practise those, when I'm reviewing I often like I'm over-practising the review words. There are definitely a bunch of characters I don't know so well, but these are getting lost in a big chunk of characters I'm pretty comfortable with.

Also I do want to keep adding characters even if my recall on every character is not 90%. I like to be familiar with a lot of characters, even if I can't write them perfectly. And I like seeing new characters, in part for the connections with old characters.

Maybe by using Save Me I'm not using Skritter as it was designed, but going through a queue of 几百 words, many of which I know pretty well (say a dot missing) is hard work. I'm not saying it's not important to get the dot, just that it's getting lost in the sea of characters I want to put off for later.

Ah, I think I see the bug with the Save Me button. I'll upload the fix soon and that should make the number of things to study go all the way to zero.

As for the scheduling, it's pretty straightforward right now; we haven't even really begun to do the sorts of things we really want to do with it. It does need to be improved. New things should be coming up a lot more often than the old things, though, and if you know something it should only take a few reviews before it's showing up pretty infrequently. The best thing to do if you want to see something more often is to mark it as incorrect, and you should do that whenever you don't feel you know it enough, and it's up to you to decide if you know something well enough (though we want the program to do well at judging that for you, so you can focus more on just practicing).

We do hope the Save Me button will be used sparingly. It should be used for special occasions, like when you've been gone for 2 weeks and your list of things to review is enormous. Ideally you would never use it, but that's in a perfect world where you're using the site all the time and our scheduling system is a lot tighter.

Also, note that whether you get something right the first time has a huge effect on how often you'll see it. So if you're given a word you know, make sure it's marked as having gotten it right, and then you won't see it for about several days. If you get it wrong, it will be scheduled in a matter of minutes or hours.

Another thing to keep in mind is an addition that came with the new system. The scheduling system relies on values which get adjusted based on how well you're doing. For example there's one value that's for how long to schedule something that is new that you got right. Its default is a week, and it goes down if items that use that value are not remembered during the following review and it goes up of items are remembered. Looking at your factor history, ximeng, that value dropped to a day and has been rising since, so for the last while items you got right on the first try were not getting scheduled as far ahead as you may have liked, but that seems to be because you got them wrong too often when they were scheduled farther ahead in the beginning. That may be why they tended to show up often.

I hope all this helps you to get the kind of scheduling you want out of Skritter. We want it to be all automatic eventually but for now hopefully knowing how things work under the hood a bit more will help make it easier to use.

Ximeng, your pain is felt. We don't have a way to space out words with the same characters in them, as the spacing algorithm has to run very fast (during one web request) and can't search arbitrarily for words of yours that share characters. So it can only space out other parts (currently tone or characters) of the word or the characters inside. (It won't be very visible that it's doing this until we add the definition and reading practice, since it usually just combines the characters and tones).

I think the best way to solve the problem of tons of easy words that have common characters in them will have to come once we build the Kitness Center and crunch the data for tuning the Kitten on individual items. Then it'll turn out that 上去 and 法文 are really easy and will get scheduled further much faster, where 议题 and 义题 probably frequently get confused, 父母 is also easy but 夫姆 gets confused with 父母 and thus becomes hard.

That's not going to happen for a few months, though, since we need to build a data collection and analysis system and work item-specific difficulties into the scheduling.

Until then, make sure when you get something you don't want to see often that it's marked correct, and try to work on clearing that backlog--soon after you clear it, things should return to normal. Save Me will make it take longer to clear the backlog, since it'll be more spread out, but the later you do a scheduled review, the longer the next interval will be, so at least for the easy ones that you're going to get right, the total amount of practices will be fewer. (The harder ones that you forget because they're pushed back will take more time overall, though).

I'm not sure why Save Me isn't pushing everything back, will ask Scott. You can just hit it again if it hasn't moved all your items, though.

The number one thing to do is practice regularly. The kitten isn't going to be able to space things out well if you wait several days between practices. It'll still work reasonably well except for some really hard ones, but it'll be pretty frustrating because you might see the same 3-day easy ones as you did last time but forget all the < 1-day hard ones by that time.

Hope this helps. I do think that finally clearing the review backlog will put things aright.

First post!

Chloe distracted me.

I think Scott explained why ximeng is having problems.

On a fundamental level I don't want my review times changing intelligently.
It won't work. Big review piles and characters learnt without skritter mess data up too much.

If you believe in the intelligent spacing system would you at least let us see the modelling variables and manually alter them?

I feel I'm having the same problem as ximeng - in reverse, I'm learning too many new characters, and not reviewing enough old characters. This is because my numbers are affected by characters I learnt before skritter.

Please change it back to the old system, where the spacing didn't try to be so smart, or limit the system within a narrow range, or let us change the numbers ourselves.

I like cementing knowledge, 95% retention rate suits me best, others want to quickly expand their knowledge with new words, 60% is fine (although imo unwise).

Kitten isn't working, it needs some deep meditation you guys.

Ximeng,
I was originally ridiculing your statement of "Over practising" when your retention rate was only 75%.

I now see my mistake. Your chinese character goals are completely different to mine. You want exposure to as many characters as possible in order to stumble through a difficult piece of text, completely understandable.

I want complete mastery of all the characters in my lists so I can speed write easy sentences. I'm happy to spend 1hr rewriting a single new character over and over again to make certain I know it perfectly (if this were efficient).

BIG QUESTION: How can skritter accommodate both of our methods? The kitten keeps pushing my reviews further and further away :( and keeps grabbing your reviews and forcing them upon you :(

Am I making any sense? Do you guys understand what I'm getting at?

I think perhaps the best thing to do is to allow the users to choose their target retention rate, so that you could choose whether you want to try to get things right, say 60% or 95% of the time, or somewhere in the middle. Then you've got an easy way to choose what style of study you prefer. Right now, it's essentially set to 90% for everyone, so if you get things right less often than 90% things start to be scheduled more frequently, and vice versa. But it wouldn't be hard to make it possible for people to set their own values. I can add that feature soon. I think that will accommodate both methods while remaining simple to use.

In building the system, I have though tried to take into account what you mentioned, Zach. The farther away you study something from the scheduled time, the less the value is changed, so if you study something too late or too early that time is essentially not counted. Also, because it's a pretty untweaked system I have limited how far it can go, and I've been making the limits a bit tighter. I think with some time and focus, it can help to make the scheduling more efficient. We just need to spend some time on it that we don't have to give it right now. You're right, though, it does need work.

If you want, though, I could provide you with the tools to view and change the values specifically, Zach. That actually won't take long; I just need to update and clean up the page I made for myself. I don't want this to be what people use normally; I think a simple option in the vocabulary options page would be enough for the vast majority of people. But I'd be interested in what your thoughts are if you toyed around with it for a bit, or anyone else who is interested.

Can you please set my retention to default? I feel that it has been influenced by all the characters I had already learnt before skritter.
I will see how the system handles it when I am only learning new characters.
I could end up with 2,000 reviews, that is fine.
Thanks Scott

Some of the values were higher than others, some lower. I left the low ones low and brought the high ones down to default. It will continue to adjust to try and get you 90%, though, until I can set up a system to allow users to set that number. Right now we're focused on squashing the bugs that remain so that the site is as stable as possible, but once those clear up that feature should be added pretty soon after.

Thanks for the explanations Scott / Nick. Particularly interesting is the info on how new characters are scheduled. Quite often with new characters there are two sorts - ones I really don't know, and ones I just about know but am not used to writing.

I imagine this combination would make it hard to have a universal system for predicting how long to schedule characters when they first appear. I think the reason it dropped low is because sometimes it takes me a couple of goes to figure out the character, sometimes it's going to take significantly longer. It might have gone up when I got a whole bunch of 上去 style variants on simple characters.

I'm still thinking that a way of Skrittering without the expectation of logging in every day would be good. I don't have much time in the week with work, and often the time I do spend on Chinese will go to alternative study methods apart from Skritter. Perhaps just using Save Me a lot to hide from the big pile will work well for me.

I think you nailed it ZachH when you said I'm after recognition of characters, not necessarily perfect recall for writing them. Recall's good, but I'd like my emphasis to be on recognition. Being able to write a character's really good for recognition, but if I get it 90% right (miss a dot or something), then it's often not going to be essential for me to keep practising until I get the dot. I don't mind waiting a bit to get things perfect if I can get it good enough and see more characters. So the temptation is to just mark it right, but that's in tension with wanting to actually get it right 100% one day. Perhaps just stretching the schedule would work for me, or reducing the target retention level. I don't know.

Still Skritter's a great tool, and I think with the save me I have a rather clumsy method to study new characters every now and then and spread out the schedule. I'll also try and rejig my schedule to allow a bit of Skrittering each day, but I'm not sure how that will work out.