Uploaded November 2019 | Updated September 2026, 2 hours ago
A series demonstrating different maze algorithms via interesting visualizations.
The maze generated here is my variation to the Hunt-and-Kill algorithm. To summarize how it operates, and where it differs from Jamis Buck's version:
- The maze is generated by creating paths through unexplored territoriy from a previously visited cell.
- When the tunnelling gets "stuck" (i.e. has nowhere not visited left to explore), it finds a new starting point.
- The traditional version of the algorithm scans the maze in a pre-determined fashion and chooses the first valid location it finds.
- My approach differs as it looks through the entire maze without stopping, counting all the possible starting points that could be selected, and then randomly chooses from that group.
- Although a lot more expensive computationally, it produces a more varied maze generation visualization, which I think looks pretty cool!
This maze type is not as clearly biased as the Binary Tree and Sidewinder mazes - which can be seen as there are no diagonal or vertical trends visible in the flood-fill heatmaps. (I had previously stated that this was completely "unbiased", but there are biases - just not as startlingly obvious!).
The program constantly generates new mazes - makes for a nice screensaver of sorts!
Acknowledgements:
This maze algorithm is a variation on that presented by Jamis Buck in his excellent book "Mazes for Programmers" - a great read!
Other Social Media stuffs:
twitch.tv/s0llygaming
twitter.com/s0lly
instagram.com/s0lly.gaming
github.com/s0lly
patreon.com/s0lly
A series demonstrating different maze algorithms via interesting visualizations.
The maze generated here is my variation to the Hunt-and-Kill algorithm. To summarize how it operates, and where it differs from Jamis Buck's version:
- The maze is generated by creating paths through unexplored territoriy from a previously visited cell.
- When the tunnelling gets "stuck" (i.e. has nowhere not visited left to explore), it finds a new starting point.
- The traditional version of the algorithm scans the maze in a pre-determined fashion and chooses the first valid location it finds.
- My approach differs as it looks through the entire maze without stopping, counting all the possible starting points that could be selected, and then randomly chooses from that group.
- Although a lot more expensive computationally, it produces a more varied maze generation visualization, which I think looks pretty cool!
This maze type is not as clearly biased as the Binary Tree and Sidewinder mazes - which can be seen as there are no diagonal or vertical trends visible in the flood-fill heatmaps. (I had previously stated that this was completely "unbiased", but there are biases - just not as startlingly obvious!).
The program constantly generates new mazes - makes for a nice screensaver of sorts!
Acknowledgements:
This maze algorithm is a variation on that presented by Jamis Buck in his excellent book "Mazes for Programmers" - a great read!
Other Social Media stuffs:
twitch.tv/s0llygaming
twitter.com/s0lly
instagram.com/s0lly.gaming
github.com/s0lly
patreon.com/s0lly
