Could someone clarify what the "Learned" progress bar in the Vocabulary List and the individual character pages represents?

Is this just the number of times you have gotten the character right? Or the percentage of attempts that were correct? Or some other metric?

Also, what is the threshold for when a character is considered "Learned" as measured on the Progress page? Is it a certain value of the "Learned" parameter?

It's a measure of the current interval length for that item, on a logarithmic scale. It maxes out at 42 days, currently. When you get something wrong, the interval drops a lot, probably near 0 on the learnedness scale.

When the interval is more than half a day, it's marked as learned on the progress page. This corresponds to about one-third progress on the learnedness bars. This usually happens pretty quickly after you get something right after a day.

We're open to feedback on how best to represent learnedness. Are the progress bars helpful to you? Does the learnedness metric correspond to what you'd expect?

That corresponds fairly closely with my intuition on how to represent learnedness.

The special case that's of most interest to me are the certain characters that just won't "stick", and that repeatedly give me trouble. So just because I've gotten of of these right a couple times today, I wouldn't necessarily consider it "learned". A count of the total number of times you've gotten a character wrong would be of interest to me.

I think you're not going to be able to come up with a single representation that meets everyone's needs, so providing alternate formats would be helpful. The existing data on the vocab page (last scheduled, next scheduled, time spent) goes a long way in this direction (in particular, characters with a long "time spent" tend to be the ones that I most need to study, because a long time spent means that I've both seen it a lot and spent a lot of time thinking about it each time).

The enhancements I'd like to see on the vocab page are as follows:

1. Include the total number of times you've seen the character, and the total number of times you've gotten it right (or wrong).

2. Included the "learnedness" metric as an actual number, rather than a graphic, to facilitate loading the data into a spreadsheet or other representation.

3. Provide an option to see or download the whole data set, rather than just 20 characters at a time.

Enhancements like these would allow the more analytical folks to fine-tune their learning environment and focus on the areas that they need to improve.

It's probably not going to count as "learned" when you've studied it during the same day, since the interval won't be long enough.

Scott may correct me on this, but I think that the current architecture doesn't allow us to query the whole review history like that, to provide counts of right and wrong reviews. We used to do it, but it didn't scale. I think he has some other idea for how to do it, though.

We'll do data export and it'll have all that good stuff in it, and more. It's not coming immediately, though, as it's a bit lower priority.

We could do all of those things. It would require some reorganization of our data, though, so it would take some time. Storing the number of times you've gotten an item right and wrong could be useful when scheduling and relatively cheap, though, so it might be a good idea to change things around so that information is readily available.

The table is a bit cramped as it is, so I wouldn't want to add another column for the number. Maybe I could add a button to click that would alternate between the learned bar and the actual RTD (review time difference) number. That wouldn't take long to do at all.