This is part of a very nice larger article by the same author on visualizing algorithms [0]. It covers sampling, shuffling, sorting, and maze generation.
> an unbiased sample of all possible spanning trees
this also makes the mazes generated by the algorithm the most annoying to solve without a computer, but they are the most resistant to be searched in reverse of all the forward building algorithms (without explicit honeypots for trapping reversers)
>> this also makes the mazes generated by the algorithm the most annoying to solve without a computer
Came to say that. I see a lot of what looks like the letter E in there. Many tiny dead ends. Really hard to visually follow a path through, and I can only imagine how not-fun it would be to play a first person game in there.
given the complexity I an see them being useful for decorative garden mazes (as in, very very tiny) - lot of complexity but actually small could give an interesting run without being overwhelming.
Ah, this is very similar to some old maze generation toys I was working on a few years back. I was experimenting with adding a little extra stochasticity to the mix. E.g. selecting from the last ~30 neighbors added creates interesting mazes with "multiple growth fronts" rather than one tendril or a uniform expansion at the edge. You can see a GIF example here: https://raw.githubusercontent.com/j-s-n/mazes/master/example...
(My code uses toroidal topology, which is why you can see things going off the edge and coming back on the other side.)
And things like varying whether new neighbors are added to the head or tail of the list adds a little perceptible chaos. It's too bad it looks like my demo site is down right now.
The "flooding it with colour" link leads to an algorithm which will identify far away parts of the maze. You might, for example, pick a faraway (in travelling distance) point that is also on the other side of the maze physically.
[0]: https://bost.ocks.org/mike/algorithms/