renshuu does adjust based on the first stroke, actually. So if you make a good stroke but it's a little too far to the left, renshuu does account for that.
There is not really a solution to what you're describing without throwing out the entire system, or building a new system alongside it, which I do not plan on doing. Some of the visual changes (such as toggling the colors on or off after drawing) might be possible, but the system is well received by most such that this is not a high priority (given that I am the only developer, and I have a huge number of things to do)
While this doesn't solve anything, this information may help you better understand how things are set up.
The writing system is layered on top of those svgs - those are the visual images of the kanji strokes for the thousands of kanji that have them. They come from the https://kanjivg.tagaini.net/ project (which is awesome).
So when you look at a single stroke, it is not described in the sense of "ok, draw this kind of shape, curve here, hook here, etc.", but rather it's ultimately a collection of pixels that make up the stroke.
So, when the user draws a stroke, in overly-simplistic terms, it's going to say "how close is their stroke to the expected stroke". However, a computer program (without getting 10x more complex than what I have) cannot do what humans do when we look at the stroke, so it is ultimately looking at a percent overlap between the two strokes. In addition to a percent of overlap, it also has numbers representing how much of the stroke was covered (for example, did you leave off the bottom half), and how much of your stroke went outside of the expected area.
So it takes these three numbers, and tries to bracket the stroke into a grade (the low/medium/high). There are absolutely edge cases that one can create that will make it look like "what is the system even doing", but I believe it works well enough most of the time that trying to address those edge cases would make 1% of it better, but cause issues for the other 99%.
There are many things that renshuu cannot do (without *significant* changes - and much slower processing - if it takes half a second for it to grade each stroke, then users are not going to be too happy).
- Adjust the size of the user strokes. It might seem "obvious" to you (a user) to see a small ム (let's pretend that was one stroke) and say "obviously that is the right shape, but it's 25% too small". From renshuu's perspective, there is no "shape" data that can be used to compare to the expected stroke. So, the only solution that I can think of would be this: "after user takes a stroke, stretch out the height and width of the stroke to match the height and width of the expected stroke". This would cause...a lot of problems, much more than the issue it is trying to fix.
From a teacher's perspective, I would argue that this should not be implemented, even if it is possible. Getting used to keeping all of your characters the same size in relation to the box is essential for legible writing (if you plan on using it in the real world). Legibility and "regularity" in writing style is important both in allowing the reader to quickly and accurately read your text, as well as the handling the reality in judgement that all people make when seeing things in the world. It might be external, or it might not even be something they realize they are thinking, but I would expect that people will treat writing that is clean and legible better than something that is skewed in shape/size, perhaps looks like a young learner wrote it, etc. I'd like to think that doesn't matter, but humans are humans.
- Adjust the angle of the stroke (or part of the stroke). There it is a single stroke (一) or something more complex (like the bottom of 道), trying to take that shape and rotate it along the center point (if a center stroke can be determined for some of these) a few degrees in each direction, then doing another overlap scan would be computationally expensive, and would not even solve a number of the issues (such as one piece of the stroke being correct, but the 2nd part being just a bit off in one direction or another).
I do hope I can add additional customization in the future, but both due to time constraints and the growing user base of renshuu, I try to strike a balance between customization and visual complexity. renshuu is already considered by many to be extremely hard to use, the interface is too packed, etc. A lot of this comes from the level of customization available, so each request has to be considered given the needs and expectations of both "normal" as well as "power" users.
