Maybe it's me, but I'm having real trouble getting skritter to recognize the second stroke of any character using these. With 口 it's only those for which 口 is the outer-most character. Same problem with the third stroke of characters resembling 问.

With 饣, it's any character with the first two strokes. Sorry I don't know my radicals, heh.


This happens even when I trace the blue outline exactly, I'm fairly sure it's not my ineptitude shining through here, though I've been wrong before.

I turned recognition strictness down to 50%, but that's not really ideal as I can now get away with scribbling messily (was enjoying being forced to write the characters properly with strictness at 65%). Is Skritter being extra harsh with its strictness for these characters?

Anything above 50% results in about 5 failed attempts at the aforementioned characters, which means I have to press the "toggle correct" button to tell Skritter that I do actually know this one, honest!

I'd like to know if anyone else has this problem with these sections of characters? Am I alone?

Duh I just noticed that for 口, it's not happening just for cases where 口 is the outermost character. Happens with 烟 as well.

Both of those strokes should have a flick at the end. Even though 口 might look like a square, you should flick the end of the second stroke towards the centre.

The flick in 饣 is also important. Flick it at a sharper angle and skritter will be better at identifying it.

Hope this helps!

I've had a similar problem with 围 - Skritter recognised the second stroke as the top horizontal in the 韦 followed by the bottom horizontal-hook in the 韦, finally recognising it as what it should be. I was also tracing the phantom. Funnily enough I cleared it after writing to see if I could reproduce the problem, but the second time it didn't have the same problem.

Nick's said that this kind of problem will be improved when the fabled component-based recognition is completed, but not sure it would help for the characters you mention. I've turned my strictness right down to avoid these kind of thing happening, but it does mean strokes get recognised out of order much more often.

I think it takes a little while to learn how skritter likes the strokes to be written. When I started I used to get stuck on certain parts (this may be due to improved recognition)

I occasionally write a stroke too messy but I never have to rewrite a stroke more than once. I feel my writing style has slowly changed to become more skritter-friendly.

I have been monitoring my strokes and noticed some features that may be helping:
1. Increase the sharpness of your strokes, for the second stroke of 口 don't draw it flat, the horizontal part should slant upwards (10degrees), and then the downwards part should move down and slightly to the left (5degrees), this creates a more acute angle, which skritter appreciates.
2. Extend the strokes, increasing the tail length of certain strokes can help.
3. Many of the horizontal strokes actually lean upwards, so make sure you draw them with a slight angle too.

@faceleg, I think making your angles tighter will fix the problem.

@ZachH: I'm flick'n flicking the flick, for flick's sake! Hahaha ;) Interestingly, I have more success with 2nd stroke of 口 when I *don't* do the flick. The second stroke of 饣 is troublesome at 65% no matter what.

@ximen: what did your turn your strictness "right down" to? 50%? Lower?

Please note that at 50% everything works well - I just like to push things. I'd rather have higher accuracy requirements placed on my learning, but this issue makes 65% impractical for me at the moment. Again, this could just be happening to me - albeit on a consistent basis (only with these strokes though).

I believe Nick will solve this issue, and am happy to wait. I was just interested to see if anyone else had encountered it.

As it is I'm amazed at the recognition code used in Skritter, I'm sure glad I didn't have to code it ! (while at the same time being jealous).

@ZachH: I agree, it does take time to learn how Skritter does it.

I've been trying to copy the blue outlines exactly, but I don't think I've been moving left 5 degrees. I'll pay special attention to your points during my next study session.

Thanks :)

Nick to confirm?
How strongly linked are the phantom and recognition? I didn't think they were linked at all.

pff well that might explain it.

I use 0% strictness, 100% reliance on stroke order.

0% strictness because if I know a stroke I generally want Skritter to accept it even if I'm writing quickly and don't do it exactly how it wants it. If I don't know a stroke then I mark the character wrong.

100% reliance on stroke order because I want to encourage Skritter to recognise strokes in order, rather than assuming I actually meant to write it in a random order.

The phantom and recognition aren't linked, in the sense that when we resize/rotate a stroke to fit into a character, the recognition doesn't change for that stroke. So we try not to resize things too much.

The recognizers are also tuned not to the phantom, but to the squigs we collected of how people were actually writing each stroke. So in some cases, tracing the phantom may not recognize as well as a more natural writing style. It is true that making sharp angles helps a lot. The most important thing is that the number of segments in your squig match the number of segments in the stroke (although with certain types of box strokes, it's optional whether to put a hook or not).

Faceleg, it might help if you up the stroke order reliance at the same time as you up the strictness, because that'll give a slight boost to the score of the stroke that should come next (while decreasing the score of any strokes that shouldn't). In general, the strictness ranges aren't very well done and won't always lead to workable combinations. But that's been "due for retuning" for several months now.

The recognizers you mention probably just tend to score things a bit lower on average, so when you bump up the strictness, their average recognition scores fall below the threshold.

Hope this helps; I'm probably not going to do too much with the recognition in the short term, although I always have big plans for how to improve things down the road. So, I hope it's not getting in the way too much.

I'll try your suggestion, thanks :)

As ZachH said, it could also be due to the fact that I'm not so familiar with Skritter's nuances as he is.

I'm defnitely not going to have a tantrum over this, I'll wait for something far more petty.

Well I still use 50% when I study. Skritter isn't a handwritting improvement tool for me. I get through my reviews as fast as I can, not bothering to move the mouse and accurately place the strokes within the character.

Using higher strictness would mean carefully placing the strokes, slowing me down.

Although, my chinese characters ARE quite messy when I write them on paper.

I can’t get Skritter to recognize the third dot of the character 憋, the lower left one of the group of four on the top left. I’ve to skip it every time. But this is the only character that I’ve problem with so far.

pts, it's probably because that stroke isn't a very good fit for that character. I've changed it to a different one; let me know if it works any better.

憋 didn't give me any trouble (although it is SO fustrating to learn!)
Are you sure your stroke order was right pts? It is a little strage, (the lower left dot is the 6th stroke). I learn with 50% strictness, so maybe that helped.

Good luck with learning 憋! I hope you can learn it quicker than me.

nick, thank you for correcting the problem. I’ve just checked with the Scratchpad, it’s OK now.

@ZachH: Actually Skritter recognizes the whole character except that little dot, no matter how hard I tried. So, it’s not with the stroke orders. For me, this character is not particularly difficult because I’ve learned the 敝 character before.