Last two weeks or so I've been having a really tough time to get the recognizer to recognize the second corner stroke of components like 用角月.

I can draw the exact shape over the shadow and maybe get a recognition rate of under 10%. kind if frustrating when you're in the middle of a thousand character review.

Curiously, I no longer have trouble writing the second stroke of mouth char 口. A few weeks ago, that stroke gave me all the trouble in the world.

I use a high resolution tablet (9" screen, 1280x768) with a passive touch digitizer over the screen. I'm wondering if I'm just hitting a rough spot in the recognizer, or if because of the length of strokes the raw digitized stroke has too many fine jaggies.

Yeah, the second one in characters like this 角 is an utter pain to write. Not so much noticed it with the larger ones though.

The handwriting recognition algorithms should be smoothing over the input and recording points at normalized intervals. More likely those recognizers are just finicky with the beginnings/ends of these strokes. Try not to have any quaver at the start of the stroke and let me know if you still see problems on quaverless strokes.

I will be doing some tuning on these in the next round, and hopefully I can improve the quaver tolerance then.

For the record, I have trouble with those too and it has been in the last couple of weeks. I believe I am "quaverless" but the more I must repeat it the more likely I could be writing with a less-natural, attack-like (and quavered) beginning to the stroke. I was guessing my corner wasn't sharp enough. The other challenge has been the "L" shaped component of 山elements, say for 微 or 端. Even still, the recognition is amazing.