Ooh, I used to handcraft mazes on the back of my notes when I was bored in uni classes. I never really learned about maze generation algs, but I eventually developped a strategy. Interestingly, my approach somehow looks similar to Wilson’s algorithm, which was a very fun surprise to me.
I usually manually placed the start and end. About half the time, I would first connect them with a somewhat random path (as random as my hand drawing can get anyway). If I didn’t directly placed start or end beforehand, I would usually draw a simple lineart (e.g. a heart), and that was the path from start to end. Until now, no real similarities with Wilson.
But once I had that solution path, I would pick a random point on any of the currently connected “nodes” (a term which may not make sense for a hand drawned maze, since it was not always drawn over a grid or with any underlying structure), pick a random destination somewhere not too far, and connect the two points with a “random” path, forbidding connection to the current maze (picking about just anywhere may force you making huge detour to avoid loops, for instance if the current maze is Y-shaped and you try to connect the bottom branch to somewhere in between the topmost branches).
This kind of is Wilson’s algorithm reversed: walking from the maze to somewhere randomly without loops instead of walking randomly until you hit the maze without loops. I think the nuance makes it biased though, but if you’re smart about how you pick points, you can make pretty interesting mazes, while unbiased maze are, imo, not that interesting to solve (which was my goal, making fun mazes to solve for friends while bored in class). Most of my last mazes trully look random, but then a drawing appear as you solve it; or don’t look random at all but feel like randomly wandering in a field of weird shapes. I was kinda proud of them, especially because this was all just self-teaching and experimentation without theoretical backing.
Ooh, I used to handcraft mazes on the back of my notes when I was bored in uni classes. I never really learned about maze generation algs, but I eventually developped a strategy. Interestingly, my approach somehow looks similar to Wilson’s algorithm, which was a very fun surprise to me.
I usually manually placed the start and end. About half the time, I would first connect them with a somewhat random path (as random as my hand drawing can get anyway). If I didn’t directly placed start or end beforehand, I would usually draw a simple lineart (e.g. a heart), and that was the path from start to end. Until now, no real similarities with Wilson.
But once I had that solution path, I would pick a random point on any of the currently connected “nodes” (a term which may not make sense for a hand drawned maze, since it was not always drawn over a grid or with any underlying structure), pick a random destination somewhere not too far, and connect the two points with a “random” path, forbidding connection to the current maze (picking about just anywhere may force you making huge detour to avoid loops, for instance if the current maze is Y-shaped and you try to connect the bottom branch to somewhere in between the topmost branches).
This kind of is Wilson’s algorithm reversed: walking from the maze to somewhere randomly without loops instead of walking randomly until you hit the maze without loops. I think the nuance makes it biased though, but if you’re smart about how you pick points, you can make pretty interesting mazes, while unbiased maze are, imo, not that interesting to solve (which was my goal, making fun mazes to solve for friends while bored in class). Most of my last mazes trully look random, but then a drawing appear as you solve it; or don’t look random at all but feel like randomly wandering in a field of weird shapes. I was kinda proud of them, especially because this was all just self-teaching and experimentation without theoretical backing.