[olofsj]

The SRS algorithm is based on SM-2, but I'll probably tweak it to handle the "missed a day" problem as you say.

The main reason is the handwriting feature, yes. Maybe it could be implemented on top of Anki, but it seems more work than to write the SRS part myself. The recognition unfortunately doesn't work flawlessly, but the native Japanese who have tried it said it works well.

I haven't done any study on if it's more reliable compared to self-reporting, though it would be very interesting to do so. From my experience, with self reporting there are some subtle mistakes you can make that you won't notice yourself so you can't correct them. For example, when writing 田 it's easy to write the middle two strokes in the wrong order. But when self reporting it's hard to notice this since when written it looks the same.

I'll take a look at the forums you suggested. Thanks!

[dsrguru]

Haha 田 is actually a character whose stroke order differs between China and Japan. I'm so used to the Chinese order of the horizontal middle stroke and then the vertical one that I'd probably deliberately use the "wrong order" for Japanese. But yeah, I'd be really interested to see the effectiveness of your handwriting recognition against self-reported scores. If its innaccuracy is worse than a human's judgment of his or her performance, then it's just a gimmick that will hook people who are too lazy to learn how to use an SRS that doesn't hide its implementation details. But if it really works better than self-reporting, your app would be a real advancement to Japanese and Chinese literacy acquisition.

Also, I agree that handwriting on top of Anki would be more work than its worth, especially if you're trying to market your product as a standalone Japanese-specific program.

[olofsj]

Actually I wasn't aware that the order for 田 is different in China, I just used it as an example since I always have trouble remembering it. =) Generally, I find the stroke order hard to self evaluate, since it takes time to check stroke order diagrams each time and you tend to get lazy after a while (like in 左 and 右 where it's easy to mix up the order of the first two strokes).

I'd love to do a real study on the effectiveness. But realistically, I don't see it happening since I don't have the resources to set up a scientific study. But if you're interested or know someone who would be, let me know. =)

One plus from the algorithmic approach is that you really have to write each time. There's no way to skip it and just look at the answer even for easy ones you think you remember. Therefore you get a lot of practice writing and hopefully writing the characters sticks like "muscle memory". Of course, if you're disciplined you can do the same with for example Anki and pen and paper, but not everyone (for example me...) are that disciplined.

[w1ntermute]

> like in 左 and 右 where it's easy to mix up the order of the first two strokes

That one's easy, just remember it's "opposite" - in 左, the first stroke is to the right, whereas in 右, the first stroke is to the left.