User:FIQ/Dungeon Overhaul Proposal

This was my contribution to the Dungeon Overhaul project some time ago (2-3 years?). It is not in a wiki format at this time.

DUNGEON OVERHAUL PROPOSAL -- FIQ EDITION
This document aims at describing my take on the so-called
"dungeon overhaul". It describes my personal view on how
the current game progress of NetHack works, what it does
well, and the issues it has, along with providing personal
solutions on the issues. It aims primarily to describe
how the general dungeon should be designed in my view, but
will also touch on so-called "balance issues" at times if
relevant.

THE PURPOSE OF ALL THIS
Several years ago, in 2014, a group of people mostly
connected to NetHack4 development, wanted to brainstorm
major personal takes on how they would personally overhaul
NetHack's dungeon structure. This was to be done in isolated
circumstances, without cross-referencing ideas (at first),
to allow each one to come up with their own take. After
this was done, the proposals would be looked through and
discussed, the idea being that ultimately a combination of
them all can lead to a better dungeon and game. The timeline
for when this was to be done was on the release of the
stable version of NetHack4 4.3.
    This release never happened, and the Dungeon Overhaul
project ended up thrown in the backburner. Recently, there
has been attempts at restoring this idea, this time without
waiting 4.3-stable.

PERSONAL THOUGHTS OF CURRENT DESIGN
I think that, for the most parts, the earlier parts of the
game works rather well. The earlygame has a nice variety of
places of interest, and this continues for a large part of
the game. I will raise my comments about what I think about
the various places of interest. I personally don't have
strong opinions on major redesign of the pre-Castle part
of the game, the Dungeons of Doom, and feel like it has
*mostly* just minor issues that are salvageable.

Gnomish Mines -- Well-designed, alternative earlygame path.
These levels have a "cavernous" generation instead of the
normal room-and-corridor levels. Has a fair share of gnomes
and dwarves. The levels can be either bright, or dark.
(Minetown, the central Mines level, is generally always
bright except for one of its potential level variants).
The brightness state of a level is random, there is no
logic in which ones are bright and dark, except the deeper
you go, the higher the chance of the level is to be dark.
Dark levels pose problems for people for which gnomes and
dwarves aren't peaceful to, which makes the mines far more
dangerous. I don't really like the randomness of this, and
would like more consistency. Also, I have a problem with
one of the "Mines End" levels, Mimic of the Mines:
* It is noticeably more dangerous, but less rewarding,
  than the alternatives.
* It is rather "plain" -- it is rather similar to a normal
  mines level except it has a few guranteed "closets" with
  mimics or gray stones (one luckstone, one loadstone).
* It has a place where you need to squeeze through
  diagonally. This part in particular is rather easy to
  address. Also note that, should you pickup a loadstone
  inside this area, you cannot easily get out of it, which
  I find rather cruel.

The "more dangerous but less rewarding" and the squeezing
issue alone would make the level salvageable (have more
rewards, or tone down the difficulty by introducing bright
spots or similar, or cut down on traps). However, I find
that the level is simply too plain in comparision to the
other Mines End variants, which can't really be addressed
easily. I personally find "Khor's Temple" a suitable
replacement, allthough its rewards might need to be toned
down a little. I would also personally suggest to change
the Mines brightness to make "upper Mines" (pre-Minetown)
always bright, and "lower Mines" (after) always dark.
I would also move Vlad's Tower branch entrance to the
penultimate Mines level, but add a +10 depth count to make
it more dangerous than the adjacant Mines. The idea being
that visiting the tower is risky, but rewarding, since its
guranteed items matter a lot more at this point of the
game. It also balances the fact that Vlad, for when he
appears currently, is laughably weak.
Also, while I wrote this overhaul with 3.4.3 in mind, I
want to address Orc-town, one of the Minetown variants in
3.6.0. Frankly, I don't like this level at all in its
current form. It is basically the RNG saying "OK, you have
a 86% chance of getting a nice town level, and 14% of
getting a guranteed convertible altar and nothing else of
value. Oh, and by the way, the level is much harder than
the others. Have fun!". I don't really have a problem with
this level *per se*, but feel like it makes for a horrible
Minetown replacement. One twist I came up with that I want
to try, is to include Orctown, but if it generates,
gurantee a special Mines End variant, a "Minetown II" if
you like, with survivors who fled the orc-infested town.
This could make for interesting added variety with the
game basically switching the general content of Minetown
and Mines End.

The Oracle level is probably one of the game's strongest
levels. It's not all that special, but I see no flaws with
it. The Oracle can help new players with some spoilers, and
the guranteed fountains and statues makes for added variety
and can prove useful. This is probably the only level I'd
leave as-is with no changes.

Sokoban is easily the dungeon branch in the earlygame
dungeon that I like the least. My issue with this is
twofold. First, I personally just plain don't like Sokoban
(the game). Second, I feel like the completely different
mechanics of it are rather out of place for NetHack.
However, I know that some love this place. The main issue
is that this place felt as if it was intended as an
optional side branch (and it probably is in SLASH'EM where
it originated), but the rewards for completing it are much
too rewarding to pass up (an amulet of reflection and
a bag of holding are both very good items). In addition,
the branch poses an issue for people who actually like
Sokoban, because it has very few level alternatives, so
after just a few games of clearing Sokoban, you have seen
them all. My suggestions for fixing this:
* Make many, many more Sokoban levels. Several variants
  have already done this, so there is a nice library of
  levels you can use. This makees it very hard to see them
  all, which makes people who like Sokoban the game happy.
  - Alternatively, make a generator. This means you'd have
    to code one, however.
* Replace the rewards with 3 chests (one for each prize
  closet in the final level) with general random chest
  content.
* Move the existing Sokoban rewards to the Quest. More info
  on that in the Quest part.

Big Room. A rather nice twist on the general dungeon layout
(and different takes on it is now common in roguelikes).
My only real issue with this place is that it isn't
guranteed -- it adds variety, and not having it sometimes
is weird. It could also have some minor improvements to the
more plain variants, but it's not a major concern.

The Quest. How well the Quest works mostly depend on the
role you play (and thus which Quest you get). I will not
comment on individual quests, because then it would be
half of this document or more. My main issue with the
individual Quests themselves is that a lot of the Quests'
filler levels are plain room-and-corridor (isn't the ones
in the Dungeons of Doom enough?). I feel like each Quest'
filler level types should be unique to that one Quest.
Also, *several* of the quest levels are rather mediocre,
or have serious issues otherwise, mostly with their layout
(not balance or rewards), the obvious one that springs to
mind being the Ranger Quest's forced long-winded "maze"
corridor with a lot of quest friendlies you have to deal
with.
For the Quest in general, I like the idea and execution
(minus mentioned level issues) mostly. Here is my main
issues and solutions:
* The Quest demands you to be XL14. I feel like this is
  too high, because it usually means that doing the Quest
  before Castle isn't an option. I want this to be a real
  choice the player has. Solution: decrease to XL10
* Angering the quest leader in any way will make him/her
  kick you out and seal the portal, making the game
  unwinnable. Solution: don't seal the portal, and make
  the Quest Leader the holder of the Silver Bell, so you
  can kill him/her (or pacify, if you didn't convert) to
  get the needed things to win the game, but still making
  mistakes hurt (no quest artifact!)
I would also move the sokoban prizes here. The idea is:
make the Quest Home an actual challenge you have to fight
through (some of the home levels aren't) to get to the
quest leader. As thanks for (temporarily) ridding the Quest
of menaces, he/she hands you a choice of a bag of holding,
an amulet of reflection or a cloak of magic resistance.
This doesn't require being able to enter the Quest already.

The Rogue level is something I find rather "meh" at the
moment. Not really all that special, but I personally don't
dislike it. However, I have one significant suggestion to
improve variety -- have tributes to things other than Rogue
like for example Crawl, Angband or ADOM, with different
symsets, level generation and (toggleable) monster
appearances. Other than that, it's fine as-is.

I find Fort Ludios mostly fine as-is. I don't consider it
one of NetHack's strongest levels, but have no specific
issues with it per se. I feel like the game should try
harder to generate it (generate it on the first vault that
isn't the Quest entrance level after level 10 and before
Medusa), though.

Medusa is generally well-designed, but has 2 major issues:
* One of the levels is much harder than the other (due to
  the guranteed titan, which is ironically the far bigger
  threat compared to Medusa herself)
* Medusa is far too easy to 'cheese' and instakill.
I propose the following changes:
* Give the non-titan level many more sea monsters. Those,
  unlike the titan, aren't directly dangerous, but pose
  a risk of drowning. Thus, both levels end up being non-
  trivial to "solve", but in different ways.
* Medusa should no longer fear reflection or mirrors.
  Reflection shouldn't work at all to defend against her,
  but a mirror can be wielded to defend against her
  petrifying effect. A blindfold should work as per usual.
Also, the shield of reflection should be made guranteed to
appear. It should always spawn cursed, possibly with
negative enchantment, but is intended to give players a
source of reflection for the future if they lack it.

Castle is fine as-is, with the exception that the reward
will change. The reason for this is that a wand of wishing
is too much at once to give. I propose to spread out wishes
more. The way I intend to do this is part of my general
remarks below. Instead of a wand of wishing, the castle
will feature 4 chests with a single item each, being:
* a scroll of genocide
* a spellbook of identify
* a 1:75 magic marker
* a wand of polymorph
In addition, the Castle storerooms will contain a guranteed
artifact, which can be any randomly generated artifact in
the game.

GEHENNOM
Gehennom is boring. There is several reasons for this:
* Mazes. Mazes are closed, takes forever to explore unless
  you are a wizard (magic mapping + object detection +
  teleport at will + teleport control), are generally
  *easier* since monsters can only come from 2 paths, and
  are very monotonous.
* The branch covers over half of the game, but has very
  little variety to offer for it. It has several special
  levels, but they are the same each game, with the same
  boss, and the special part of them are (apart from
  Jubilex and Orcus) very small
* The one branch here -- Vlad's Tower -- is small,
  comprises of a very anticlimatic boss for when you face
  it, and have inferior rewards. Vlad was moved to the
  Mines as per above, so now Gehennom has even less to its
  name.
* I think Orcus' and Jubilex' lair is fine (but Jubilex is
  too weak).
* I think the way you enter wiztower currently is fine,
  but would replace the mazes with aiscav.
* I would keep one single fullmaze level -- the vibrating
  square level. The entire idea of it is to find the hidden
  square, and mazes work better for this. I don't consider
  this frustrating as long as the game keeps track of where
  you have stepped (NH4 does this).
* Sanctum has a graveyard when its inhabitants for this
  point of the game are complete pushovers and only serves
  to waste time in a tedious manner.
* Insects are harmless at this point of the game. It only
  wastes time.
My Gehennom overhaul would be as follows:
* I find Valley of the Dead to make for a fine "border"
  between the Dungeons of Doom and Gehennom. It needs no
  changes. The alignmed priest should no longer be able to
  give you protection, because...
* When you go past the Valley, your acquired protection
  from priest donations, or from your god, will immediately
  vanish under Moloch's influence. This makes the lategame
  harder and prevents you from being able to raise AC to
  ridiculous levels.
* Replace mazes in general with aiscav with a random open-
  ness factor for each level. This makes for a much better
  experience, and adds variety by randomizing the openness.
* Replace the guranteed demon lord lairs, except for Orcus,
  with randomly generated demon lords and a large selection
  of dungeon lairs, and monster generation biased towards
  the type of demon lord. Below the proposed changes is a
  list of how I envision a Gehennom overview to be, for
  clarification's sake. Random demon lords make the
  Gehennom run much more unpredictable, and the player has
  to be able to deal with any potential threat. Part of
  this depends on tweaks to how resistances work (see my
  general notes). The demon lords will sit on a throne, but
  the downstair may not require fighting the lord directly.
  Thrones are more notable in this proposal, see general
  notes. The 3rd random lord (replacing Jubilex) will not
  have a throne.
* Remove the bribery feature from demon lords completely
* Add a 5-level branch consisting of full-level mazes with
  2 guranteed minotaurs on each level one of them with a
  guranteed wand of digging. These can work for players who
  actually like mazes, and as a digging wand backup for
  those who desire. The final level should have one larger
  room in the middle with several minotaurs, a minotaur
  emperor sitting on a throne, as well as a guranteed
  blessed +5 dwarvish mattock.
* The Wizard Tower works mostly fine as-is. However, I feel
  that its zoo should be replaced with 12 strong player
  monsters, each except for your own role, with kit and
  AI behaviour depending on role. Trying to sneak past this
  entourage, by luring the Wizard from outside, or similar,
  should be prevented by giving said player monsters
  access to controlled level teleportation means, and using
  it if they sense that you have the Amulet. They will
  level teleport to a random dungeon level 1-3 levels above
  wiztower and ambush you at the downstairs. If you somehow
  keep fleeing, they will continue warping 2-4 levels up
  until killed, or they run out of cursed scrolls of
  teleportation.
* Fakewiz is fine, except that it should use aiscav as
  filler instead of a fullsize maze.
* The Sanctum should be made much more open. This gives
  priests a much easier time to spot, and swarm, you. This
  makes them much more dangerous to deal with.
* The Amulet will confer aggravate monster. This should
  block invisibility, and make monsters aware of your
  location at all times. It will also blocked all forms
  (instead of 33%) of intra-level teleport. To make up for
  this, the mysterious force is removed.
DUNGEON LAYOUT
* Valley of the Dead
* 2-4 filler
* Wizard's Tower
* Random demon lair #1 -- replaces Asmodeus
* 1-2 generic levels, one is fakewiz
* Random demon lair #2 -- replaces Baalzebub
* 1-2 generic levels, one has an entrance to the Minotaur
  Mazes.
* Random demon lair #3 -- replaces Jubilex. Has a lot of
  water, akin to Jubilex in vanilla.
* 1-2 generic levels, one being the second fakewiz level
* Orcus' Lair
* 1-3 filler
* The vibrating square level
* Moloch's Sanctum

THE ELEMENTAL PLANES
I find the Planes a sound concept and feel like its
major shortcomings is in the execution of some of its
planes. One general change I want to do is to make
portal detection using cursed or confused gold detection
fail. Instead, the Amulet should alert whenever the
player is within 5, and 1, tiles, of the portal, with
different messages, if worn. This is similar, but not
identical, to how portal detection with the Amulet works
right now. Also, Plane generation should come at random.
Wizard of Yendor will always show up within your first
plane, carrying a spellbook of dig. The Elvenking and
minotaur will remain on Earth.

Earth:
I feel that Earth mostly work fine, apart from 2 issues:
* Wands of digging trivializes this Plane by digging
  several tiles in one go.
* Its major inhabitants don't hit that hard, except for
  the guranteed Wizard of Yendor, if this happens to be
  your first plane, and the minotaur (which appears
  right in the beginning).
Solutions:
* Any time you dig something, be it with a wand or pick,
  it will "dig" 1 tile. Following this, several tiles
  which were previously rock, should crash down into
  floor, around 33% of them summoning earth elementals.
* Earth elementals should be slower (speed 4), but hit
  twice for 8d6. The idea being to avoid getting swarmed
  and carefully dealing with the elementals as revealed
  with digtools. Or teleport them, polymorph them, or
  similar, this is NetHack after all.

Air:
I only have a single concern with Air, and that is that
currently, levitation gives you full control over your
motions. I feel that levitation should reduce, not kill,
your loss of control here. Instead, assuming the player
doesn't polyself into a flyer, the player should have to
navigate "manually", by hitching rides, "walking" or
throwing objects to move.

FIre:
My issue with the Plane of Fire is mostly 2 things:
* Unlike other planes, this one doesn't really stand
  out as unique. It has a lot of lava, sure, but other
  places also does. It is otherwise just normal floor
  with a billion fir traps, which players are probably
  immune to by now.
* The fact that it has a lot of fire traps, and the fact
  that it spawns resisting monsters, means that this
  place tends to create a lot of message spam. This
  doesn't *technically* have to do with dungeon overhaul,
  but is something I feel that my solution addresses, so
  I figured I'd bring it up regardless.
Solutions:
* Being in close proximity to lava and very hot
  temperatures will hurt the player progressively over
  time. d8 HP per turn if the player has no resistance,
  1 HP per turn if intrinsic fire resistance, nothing
  with extrinsic fire resistance/immunity. This stacks,
  the more lava tiles next to you, the more it hurts.
* Levitation over it burns your inventory and does even
  more damage.
* Remove the fire traps. This removes the spam, and they
  serve little purpose as-is anyway.
* Create an open "entry" plain. This is to highlight
  the next feature without shafting nonspoiled play
* This plane will have a timer. When it ticks to zero,
  the lava will boil and cover nearby floor with more
  lava. The plane is designed to be (mostly) traversable
  without having to traverse tiles exposed to this issue.
  Shortly after, the lava recedes, revealing regular
  floor again.
* Rig the timer so that this happens right as the player
  enters it for the first time. This alerts the player
  to the danger without being unfair on them.

Water:
Water is IMO the worst plane at the moment, due to being
easy but very tedious. This is due to the little vision
the player has, and the little control they have over
how the bubbles work. I have the following solutions.
Those suggestions originate as proposals to the dNetHack
author, and I feel like it works rather well, mostly.
* All the bubbles should be visible at all times
* It should be possible to "swim" between bubbles,
  requiring the player to hold their breath (which they
  can do for 5 turns, similar to strangulation).

Astral:
Astral works fine mostly. However, I find that some of
its inhabitants are currently a bit too easy to deal
with. I have the following proposals:
* Player monsters should have intrinsics expected to
  the players to have at this point
* Player monsters should have a stronger inventory, by
  having generally more useful magic armor, a nearly-
  guranteed chance for an unicorn horn, and bags with
  emergency/utility/buff items.
* Angels' magic missiles should only be partially
  resistible by magic resistance.
* Riders should have the HP they'd normally have for
  their level, not much less.
* Pestilence and Famine shouldn't be vulnerable to
  death rays.
* Using opposite alignment to ascend "wrongly" should
  lead to your god at first starting to attempt to
  disintegrate the helmet, and then summon monsters
  in anger if worn. The helm destruction should be
  avoidable by wearing extrinsic disintegration
  resistance. Note that this will only happen if you
  wear the helm *in* Astral -- already wearing it
  since before Astral will not carry any special
  penalty.

GENERAL
There are 2 things I want to address as part of this
proposal that arguably isn't directly related to the
proposal itself, because it has a significant effect
on how things play out with the changes.

Resistances
Making unique, random, demon lords doesn't help if
the player can resist everything you throw at them
anyway. In addition to the demon lords, this is
also a major contributor to latergame being far less
interesting -- a lot of monsters simply can't harm
players at all later on. The solution here is rather
simple. I suggest to replace "resistances" so that
intrinsic resistances only halves damage, while
extrinsic resistance ("immunity"), nullifies it.
I would also make certain monsters innately immune
to things, getting the extrinsic benefit. This makes
several latergame monsters actual potential threats,
and makes it more interesting and tougher on the
player.

Wishes
Vanilla gives players too many wishes at once.
My solution to this is to rework how thrones work:
Throne rooms have a much tougher-kitted king/queen,
being much harder to kill. This monster will prefer
sitting on the throne and using its magic to summon
an audience (throne monsters) when it feels
potentially threatened.  As long as the monster
remains alive on the level, sitting on the throne
will merely shock you. Otherwise, it will give a
wish (Luck>=0) or "change your luck" (increase it),
and then disappear.  The Fort Ludios throne is
special, it will instead summon a hostile Vault
Guard.  This is to keep Fort Ludios optional.
Thrones will be at the following places guranteed:
* Vlad the Impaler
* Fort Ludios (no wish)
* Castle
* Random Demon Lord #1
* Random Demon Lord #2
* Minotaur Mazes
This gives 5 wishes, equal to a 0:1 vanilla Castle
wand. There is also generally at least 1 throne
room in the Dungeons of Doom. but can be more or
less.