Questions regarding vanilla dungeon design

[Cazliostro]

Hello everyone. I'm a long time player / lurker but haven't posted much. I have a couple of questions / observations about vanilla that I thought I'd throw out there and see if they struck a chord with anyone.

I've been playing *bands off and on for years. I really enjoy the challenge of "one life / one game" and the old school feel of 1st edition AD&D which permeates the various iterations. I also enjoyed a game called Adom which I'm sure you're familiar with. my questions here are mostly about the design of the dungeon levels themselves which, for me, are the critical component around which the rest of the game revolves and depends upon for it's character.

The exploration of and combat in a dungeon is the heart of the matter. How to use corners, long hallways, doors, traps, etc. to greatest advantage. The tactical interaction between melee, ranged combat & magic & all that. The thing is, I've always been a little disappointed in vanilla's random dungeon generation. Let me cover a few points:

-------------------------

(1) Why is each level randomly generated every time it is reentered? I know at one time there was an issue with memory. It takes a lot of kb to store each level in the save file but really... modern computers have more than ample storage now.

The disadvatage to having a random level each time is that we are not "invested" in it or rewarded by carefully exploring the level and establishing "safe rooms", equipment caches or mapping the fastest routes between stairs. Permanent maps would give a sense of continuity which I feel is lacking with the "you enter a maze of stairways" scenario currently used.

I know that the random artifacts and named monsters currently depend on the random levels right now, and they can still be randomly generated upon first entering a level and then retained there until found / killed - surely it would be more fun and interesting to know that you've got "named monster X" to deal with on that next level and probably something nice to find there as well even if it takes a few trips and some careful mapping... it would turn the average random "run and gun" into more of a cat and mouse game of wits with boss monsters that stayed close to the "down" staircase of their level, perhaps.

(2) The size of each level is also a little large for my taste. I've played a few variants where one can select "small dungeons" and I always use that option. It seems to me that dungeons in vanilla tend towards "sprawl" without any interesting features that make it worth exploring (or getting too far from the stairs unless you have a proper feeling!)

I wonder if it would be difficult to implement different sizes in the options? Smaller levels with the same amount of monsters / traps & treasure tend to "condense" the whole level experience into a smaller, faster more dangerous run which some players may prefer.

(3) Unusual rooms & passages abound in some variants. These are always fun and interesting tactically. I wonder why they have not been used more in vanilla? I know there are vaults and such every so often but they seem few and far between.

-------------------------

I know that probably the first thing you want to tell my is try variant X and not worry about vanilla but let me expound further:

I wonder why Vanilla has gone through so many changes over the years but never addressed the central theme of dungeon design itself? The copious options menu allows for a wide variety of experiences when playing, why not give players the following options (for instance):

Dungeon Permanance (warning: large save files!) Yes/No
Dungeon Level Size: Small / Medium / Large / Random S/M/L/R
Unusual room configurations: Common / Uncommon / Random C/U/R

I don't understand, and maybe someone can explain it to me, the reluctance to change vanilla Angband's design to allow for more options? Is it because the variants exist and vanilla is seen as a sort of "tournament" for the ultra hardcore? I don't mean this as a flame or anything, I swear - I am just genuinly curious why vanilla itself has changed so little over the years when technology has come so far and there are so many interesting ideas in the variants to borrow from, especially when it comes to dungeon level design, which is the heart and soul of the game.

[Derakon]

Hey there, welcome!

Vanilla doesn't change much these days in terms of core gameplay, and dungeon design is definitely a part of that core gameplay. At the moment the focus seems to be largely on UI improvements and refining the power curve (i.e. the capabilities of the player at any given depth). That said, for your three main items:

Persistent levels -- they're "not really Angband", in many peoples' minds. Most roguelikes have persistent levels; one of the defining characteristics of Angband is that it does not. There are definitely some tradeoffs in this design decision, as you noted, but consider that with non-persistent levels, you are always free to try a different iteration of the same level instead of the one you got. Would you like it if you managed to render a level uninhabitable (e.g. due to a running battle with a summoner that got out of control) and were thereafter effectively barred from returning to that level? Note also that with persistent levels, the size of the dungeon is effectively capped; if you don't find what you need in the 99 levels before Morgoth, then you're flat out of luck.

In general, it seems that many players come to Angband expecting to clear every level. Persistent levels would definitely support such an approach. However, many of us find Angband to be much more fun when we drop that approach to playing in favor of one where we rapidly dive (skipping levels) until we reach our competence limit, and only there start (very carefully!) exploring again.

Level size -- I'll agree that smaller levels can be fun sometimes. I enjoy playing with small levels in variants that support the option. However, they do result in having to punt on a level more often; with less space to work with, it's easier for a level to get taken over by enemies. Would adding support for small levels to Vanilla ruin the balance? Probably not, but nobody's taking the initiative to try it and report on their findings, so it's not getting any attention.

Special rooms -- I suspect this is largely a matter of implementation. Nobody's bothered to incorporate more special rooms into Angband in a long time, not since a bunch of vaults got ported over. I bet that if you were to code them in and submit a patch, it'd be accepted.

As for options, there's a lot to be said for keeping the options list as short as possible. It's less intimidating for newbies and it keeps the codebase simpler. Of course, we're nowhere near achieving a newbie-friendly options list as it stands, but we really do need to think carefully before deciding to just chuck something in as an option.

[Adley]

having played ADOM a LOT, it may be interesting to you to make the following simile : think of Angband as a HuGe ID. this is, in my opinion, one of the most interesting feature, because it is what IS the principle of angband. The difference with Angband and other roguelikes is : You can just go to interesting places. In adom, once you have done a level, it's done. nothing interesting anymore. Same in Nethack.
Why is it kept in those games then? in Nethack, it is because you have largely enough levels to put yourself into a powerful situation. In Adom, it is the same reason.
In angband, it's really hard getting enough Power to kill Morgoth without regenerating levels.
This is balanced by extreme variety, and extremely severe inventory management issues.
So imo, it is a part of the game that should NEVER change.

[EpicMan]

The closest you can come to persistent levels in Angband is IronMan mode (hit '=' during character creation, go to 'Birth Options' and enable 'Restrict use of recall/upstairs'. This prevents you from going back up the stairs, requiring you to make the most of each level before you get to Sauron/Morgoth. It's *very* hard, though.

[Magnate]

Hey there, welcome!

Vanilla doesn't change much these days in terms of core gameplay, and dungeon design is definitely a part of that core gameplay. At the moment the focus seems to be largely on UI improvements and refining the power curve (i.e. the capabilities of the player at any given depth). That said, for your three main items:

Persistent levels -- they're "not really Angband", in many peoples' minds. Most roguelikes have persistent levels; one of the defining characteristics of Angband is that it does not. There are definitely some tradeoffs in this design decision, as you noted, but consider that with non-persistent levels, you are always free to try a different iteration of the same level instead of the one you got. Would you like it if you managed to render a level uninhabitable (e.g. due to a running battle with a summoner that got out of control) and were thereafter effectively barred from returning to that level? Note also that with persistent levels, the size of the dungeon is effectively capped; if you don't find what you need in the 99 levels before Morgoth, then you're flat out of luck.

In general, it seems that many players come to Angband expecting to clear every level. Persistent levels would definitely support such an approach. However, many of us find Angband to be much more fun when we drop that approach to playing in favor of one where we rapidly dive (skipping levels) until we reach our competence limit, and only there start (very carefully!) exploring again.

Level size -- I'll agree that smaller levels can be fun sometimes. I enjoy playing with small levels in variants that support the option. However, they do result in having to punt on a level more often; with less space to work with, it's easier for a level to get taken over by enemies. Would adding support for small levels to Vanilla ruin the balance? Probably not, but nobody's taking the initiative to try it and report on their findings, so it's not getting any attention.

Special rooms -- I suspect this is largely a matter of implementation. Nobody's bothered to incorporate more special rooms into Angband in a long time, not since a bunch of vaults got ported over. I bet that if you were to code them in and submit a patch, it'd be accepted.

As for options, there's a lot to be said for keeping the options list as short as possible. It's less intimidating for newbies and it keeps the codebase simpler. Of course, we're nowhere near achieving a newbie-friendly options list as it stands, but we really do need to think carefully before deciding to just chuck something in as an option.

I will add a third voice that says that persistent levels is very unlikely to make it into V - it's not even in many variants AFAIK (Heng maybe?).

Derakon is absolutely right on special rooms: the only reason I know of for the lack of development is that it's not yet been a pet topic for any active developer. If you'd like to design more special rooms and/or mini-vaults, please do go ahead. #angband-dev on IRC (freenode) is the place to come for advice/discussion on development stuff.

Derakon is I think a little inaccurate about small levels: IIRC d_m coded them up a while back (last year? Early this year?) and then reversed the change after testing of nightlies revealed some unanticipated problems (invariant monster detection radius being the big problem I think). This is a topic I think V would like to return to, as there's no issue of policy or dogma which says that all levels have to be the same size. It's just that V isn't yet ready for variable level sizes - again, it needs someone to do the coding for people to test (d_m has since moved on to other projects, like the quiver, though rumour has it that he may return to dungeon generation at some point).

Oh, and welcome aboard.

[Therem Harth]

Regarding special rooms... Well, it does make the game more interesting. If it's done I think it should heavily favor lesser/simple vaults and the occasional pit, as opposed to greater vaults, which become a PITA pretty fast.

Personally though what I'm most in favor of is just more monsters on the low levels. Early game right now is pretty slow and boring.

[zaimoni]

I will add a third voice that says that persistent levels is very unlikely to make it into V - it's not even in many variants AFAIK (Heng maybe?).

Hengband has them, but they're not very persistent (word-of-recall and changing dungeons both cause non-persistence).

I don't anticipate them in V because doing them "right" (whole-game) is grossly save-file breaking. Not to mention the issues with game balance.

[Hajo]

As noted before, the re-creatable levels are something that defines Angband. Changing to persistent levels would make it a different game, with different strategies and requiring a complete rework of the monster list.

Having said that, I once tried to make a patch to keep the layout of each level, but re-populate it each time one enters (new monsters and new items). I couldn't make it work, but I could imagine that to be a compromise between persistent and freshly generated levels.

[fizzix]

Regarding special rooms... Well, it does make the game more interesting. If it's done I think it should heavily favor lesser/simple vaults and the occasional pit, as opposed to greater vaults, which become a PITA pretty fast.

I'm a little confused by this. I find maybe 1 or 2 greater vaults a game. I certainly would like to see them appear more often on dlevels 90-100 than they do currently instead of the useless orc and troll pits that seem to arise.

It's important to consider 90-100 because that's where you are scumming for the endgame equipment. I think dragon, demon and undead pits are about the right frequency, but orc, troll, giant, jelly and animal pits are way to common in the late game. GVs are too rare, and LVs are too common. What do other people feel about this?

What I think we could easily do are add a bunch of 'unspecial' vaults. Probably on order 100 or so. These would be variable terrain rooms, kind of like the current special rooms, although they would be hardcoded in size. I know DaJ has some that we can poach, and I'm sure other variants do as well. A lot of these could function as mini vaults with a single possibly OoD monster and an item. They'd essentially be lairs. You could also have several iterations of the same room, some with monsters in it and some that are empty. This is a project that any of us could do, as it just involves messing with vault.txt

Another option is to add more 'special' rooms into dungeon generation. Right now these are the moated rooms, cross rooms, double-rooms and pillared rooms. This is a bit harder, unless it was easy to code the rooms. You could see something like spherical rooms being codable, but harder to connect. This is of course more ideal because it allows for more variation than the hard-coded vault rooms.

One thing I definitely would like to see is more rubble appearing in rooms. Rubble seems to only be generated in corridors.

[Cazliostro]

Thanks for the informative responses, everyone. I am interested to hear that the inclusion of more interesting rooms may be something worth looking into.

I understand the reluctance to altar the randomized dungeon levels. As mentioned it is the very fact that levels are generated randomly every time that dictates the strategy of the game in regards to the appearance of monsters, named monsters and artifacts. It would seriously affect the current balance as coded.

I think I assumed that the inclusion of permanently saved levels was something that was just "going to happen" eventually as computers got ubiquitously more powerful and saving a hundred levels of 2d mazes would be trivial in terms of storage space. It seemed to me that the reason levels were random in the first place was entirely because of 1980's era technology and the origin of the coding being from that era of hardware.

My thinking was that, like any game system, mountains of rules were made subsequent to the inclusion of a few poor, but perhaps necessary, fundamentals which then spiraled out of control as the years went by until someone got up the nerve to question those very fundamentals.

As an example I would give the "negative is better" armor class rules and the "spell cast = erasure from memory" fundamentals from 1st edition AD&D. These were frankly terrible ideas that I didn't like at the time (even when I was 12 years old and D&D was the only RPG I knew about!) They're just so obviously clunky. There are of course those who like them, but since then there have been countless subsequent game designs which abandoned the notions altogether (Angband being one of them!) It just makes sense: "Higher armor numbers are better" and "Mana points" make more sense. Yet AD&D persisted for what, 25 years or so, stubbornly clinging to -2 platemail and helpless 1st level magic users... until someone stepped up and made the changes necessary to improve the game. (As an aside I should mention I liked 3.5 better than 4.0 even with the magic system. 4.0 has the merit of introducing a solution to spell erasure but I think they went a little too far...)

So, to summarize my point: I just thought it was a matter of time until someone took vanilla to the next level and produced more "realistic" dungeon levels with permanence, variety, interesting features and perhaps thematically generated monsters, traps and treasures to reflect the 35 years (crap! has it really been 35 years!?) of thought and development concerning the art and science of dungeon delving. Dungeons themselves make little enough sense but can be improved dramatically by the inclusion of even the simplest of themes and consideration for their layout as a series of "undergorund defensive structures" as opposed to just a "random maze".

I guess this is why the variants exist, so that eventually someday someone may decide to do just such a thing. Personally, I think that if that were to happen, that variant might turn out to be the "new" vanilla!

[Derakon]

You're right that the initial decision to go with randomly-generated non-persistent dungeons was made because of hardware constraints. But just because we don't have those limitations any more doesn't mean that the design decision was a bad one.

To pick an alternate example, there are plenty of games made today that have crute 8-bit graphics. They aren't using that style because of limitations in how much space they have for graphics or how many bits they can use to represent colors; they're using it because they like the style. Of course that style initially came about because of games that couldn't do better than 8-bit graphics with chunky pixels. I wouldn't be surprised if, in an alternate history where we'd never had graphical limitations on videogames, we wouldn't have a "pixel art" aesthetic. But we did have those limitations, which lead to the creation of that aesthetic, and now there's plenty of games around intentionally invoking that aesthetic.

In Angband's case, what was borne out of necessity turned out to be an entirely reasonable alternative to the standard of persistent levels. It might seem archaic to you; I'd see it as "intentionally retro". It's been a long time since persistent levels were technically infeasible -- heck; I wouldn't be surprised if some cleverness would have allowed it when Angband first split off from Moria in 1990. But just because it's feasible doesn't mean it should be done. We could replace the ASCII graphics with beautiful polygonal 3D graphics (given a decent amount of effort from our more artistic users), and I guarantee that there'd be plenty of people still playing with the old ASCII instead. Not because their computers couldn't handle the 3D, but because that's how they want to play the game.

[Tiburon Silverflame]

I actually doubt that hardware improvements were that big a factor. The dungeon itself could be stored very efficiently; monsters and items (especially if one did NOT save squelched items) shouldn't take up a great deal of room, simply because, on any one level, there aren't that many. At a guess, I'd say 50Kb would probably be all it'd take per level. OK, for RAM, that's an issue back in the day...but not if you serialize it to disk. By the late 80's, 500 Mb drives were out there, so something like 250 Mb drives wouldn't have been unusual. And certainly by the time Windows 95 was released, Gb drives were becoming very common. So, there's been plenty of time to make this transition, without worrying that it might be infeasible for some.

I don't think it's a natural transition, because that suggests an evolutionary process. This wouldn't be evolutionary; it's a fundamental paradigm shift.

Mind you, it could be interesting. One of the points being discussed is the no-town option, or at least no-selling option, leaving the town available to provide key items (ID, CCW, WoR, basic spellbooks) only. Persistent levels might be one way to do this; as you noted earlier, this would allow @ to create supply caches along the way. But, removing town is itself a paradigm shift.

Look at the effect of persistent levels. Consider item conservation. Why do we hate fire attackers...hounds, the bigger hydras, balrogs? Because they trash spellbooks and scrolls. Cold attacks threaten potions, which are critical. Consider side effects. When we see time hounds, we know we're gonna get nailed for XP and stats, in VERY damaging and potentially risky ways. There will be times when you can take them out, but it's frequently a poor idea. Consider, simply, EXTREMELY high risk such as a greater undead pit with Black Reavers. What's the response? RUN AWAY!!!!

Note that if the level's persistent, we can run away...but we may never be able to go *back* to that level, if the threats are also persistent.

[Magnate]

I actually doubt that hardware improvements were that big a factor. The dungeon itself could be stored very efficiently; monsters and items (especially if one did NOT save squelched items) shouldn't take up a great deal of room, simply because, on any one level, there aren't that many. At a guess, I'd say 50Kb would probably be all it'd take per level. OK, for RAM, that's an issue back in the day...but not if you serialize it to disk. By the late 80's, 500 Mb drives were out there, so something like 250 Mb drives wouldn't have been unusual. And certainly by the time Windows 95 was released, Gb drives were becoming very common. So, there's been plenty of time to make this transition, without worrying that it might be infeasible for some.

I don't think it's a natural transition, because that suggests an evolutionary process. This wouldn't be evolutionary; it's a fundamental paradigm shift.

Mind you, it could be interesting. One of the points being discussed is the no-town option, or at least no-selling option, leaving the town available to provide key items (ID, CCW, WoR, basic spellbooks) only. Persistent levels might be one way to do this; as you noted earlier, this would allow @ to create supply caches along the way. But, removing town is itself a paradigm shift.

Look at the effect of persistent levels. Consider item conservation. Why do we hate fire attackers...hounds, the bigger hydras, balrogs? Because they trash spellbooks and scrolls. Cold attacks threaten potions, which are critical. Consider side effects. When we see time hounds, we know we're gonna get nailed for XP and stats, in VERY damaging and potentially risky ways. There will be times when you can take them out, but it's frequently a poor idea. Consider, simply, EXTREMELY high risk such as a greater undead pit with Black Reavers. What's the response? RUN AWAY!!!!

Note that if the level's persistent, we can run away...but we may never be able to go *back* to that level, if the threats are also persistent.

And this, ladies and gents, is the peerless roguelike known as Dungeon Crawl Stone Soup. The main reason I see no point in shifting Angband to persistent levels is because DCSS already exists and has done it brilliantly. Making Angband an inferior version of Crawl seem pointless: keeping it as a completely different game seems the way to go.

[Adley]

And this, ladies and gents, is the peerless roguelike known as Dungeon Crawl Stone Soup. The main reason I see no point in shifting Angband to persistent levels is because DCSS already exists and has done it brilliantly. Making Angband an inferior version of Crawl seem pointless: keeping it as a completely different game seems the way to go.

Wait. didn't get your point.

[nullfame]

You could see something like spherical rooms being codable, but harder to connect.

I thought all rooms in Angband were already round

So, to summarize my point: I just thought it was a matter of time until someone took vanilla to the next level and produced more "realistic" dungeon levels with permanence... and consideration for their layout as a series of "undergorund defensive structures" as opposed to just a "random maze".

Some people look at stairs in Angband like stairs in their house: one-way in, one-way out, you always know where you're going to end up. I think of them more like a maze of twisty passage all alike. Basically so many interweaving tunnels that there is no practical way you could retrace your steps. If you find yourself in said maze be glad you got out.

Furthermore I think this makes for a pretty effective series of underground defensive structures. The reason no map of Angband exists is because it is impossible to map. If I'm holing myself up 100 levels down I would rather be in an un-mappable series of caves than in a place where @ can pull a blueprint.