What is the impact of nonsensical dungeon layouts (topological impossibilities)?

[zouflain]

Newby here, so please be gentle 

I was wondering, does sense-making in dungeon layout matter much for a roguelike?

I was working on a design document and came across an interesting idea. I typically design experiences centralized around the player. How does X influence player autonomy? How does Y satisfy the player?

Well, I was extending this to AI and thinking of where AI might be useful to give the player a better experience. I started thinking about using an AI director (ala L4D) to coordinate the behavior of AIs, then realized that really the AI director could be used for all sorts of things. Like dungeon generation.

Typically, I believe most roguelikes (certainly brogue) build an "interesting" dungeon using a generic algorithmic approach (L-Systems, Cellular Automata, Binary Space Partitions, etc) that's divorced from player feel. Sure, some constraints are put into the generator, like don't put the exit next to entrance, and reject a map that's too small or too large, etc, but it's all in an indirect attempt to create opportunities for interesting experiences. Why not turn that on its head?

Why not have each room generated by an AI director every time you open a new door? That would allow certain guarantees that are almost impossible in the indirect approach. You could use the Suite/Theme/Flavor system from that old GDC talk to give dungeons a general "feel" and then feed that into the AI director to generate rooms that provide an experience suited to that feel combined with a certain "narrative" aspect. For example: "This dungeon is themed around Exploration and Dread. There will be many branches, loops, key-and-door "puzzles" and dead ends. When a sufficient number of rooms have been Explored, spawn a Dread room that spikes up the combat intensity until Dread is satisfied, then go back into Explorer mode."

The problem is, doing all that with a map that fits onto a world topology like a 2D grid is very, very hard. If instead all the rooms are treated more like separate floors are in the conventional roguelike, then it doesn't matter if there's no physical way to fit room #204 adjacent to room #002 - if the AI director says there should be a back-loop here, it can just link the two rooms assuming they have suitable unopened doors to link.

It does mean only displaying each room in isolation instead of letting the dungeon sprawl out in front of you. And it means trying to hand draw a map of the place is impossible (I grew up watching my mother spend hours hand mapping Might and Magic games so I know how important that can be to players). What are your thoughts?

TLDR: How would a topologically nonsensical map effect the roguelike experience in general? How important is a topologically sound map to the game play of a traditional roguelike?

[Krice]

I think developers overthink and spend too much time trying to figure out a "perfect" way to generate dungeons, while they are more like least important part of the gameplay. Fix everything else first, then try to be clever with dungeon generation. Think something like URR, it has I guess very detailed world generation, but as a game it's completely useless.

[jofadda]

Dungeon layout matters but it matters in a way that effects gameplay not specifically logic, nor reasoning. Look at IVAN for example:
as your first dungeon you've got a tunnel connecting two islands that aught to be well and truly impossible by any logical sense and practical layout.

Now lets look at how dungeon layouts change your gameplay:
DCSS has long corridors which allow for intense bottlenecking of hordes of enemies. It also has some wide open spaces that allow for some unique boss fights with interesting setups. The issue with this is that empty corridors give way to the need of an "auto-explore" button. If we were to take away(or even just lessen) the corridors and the wide open spaces then many of the enemy types and amounts would need to be drastically changed by virtue of the games layout.

Dredmor however generally has small-ish rooms connecting eachother with occasional large rooms within the map. This in my opinion would not allow for DCSS's enemy types nor spawning patterns to make for fun and thus good gameplay. However it allows for the idea of the dreaded monster zoo on large scale to be an occasional "fun" encounter and it allows for doorways to cease being a "chokepoint" and start being a "blockpoint" that depends on your capacity to close the door.

Then we've got games like neoscavenger that mess with dungeon layout by messing with the grid dimensions, taking them from squares to hexes, this doesnt seem like much, but it is. Being out in the open and surrounded is no longer 4/8(depending on diagonals) attacks per turn cycle you need to soak, that then jumps to 6/12(depending on if monsters adjacent to the hexes corner count within melee striking range) which then calls for a change of strategy and changes how you need to deal with getting surrounded as well as your prevention methods.