dtsund’s NetHack Overhaul

My good pal and roguelike afficianado dtsund (who sat in with me during a few episodes of Shiren the Wanderer) recently compiled some notes about how he would overhaul the class system in NetHack. He has a lot of good ideas, and a few questionable ones, and this blog post is intended to offer my comments and critiques on the changes he’d make.

I’m quoting some of the notes here, so I can better comment on them, but you can find the full text here: dtsund’s Class Overhaul Notes

NetHack’s class (or “role”) system is in dire need of an overhaul. Too often, there is little distinguishing one class from another; the differences between Valkyrie, Barbarian, Caveperson, and Samurai, for instance, are not so much huge differences in playstyle so much as they are relatively minor technical details. Little thought seems to have been put into determining specific, distinct playstyles and designing classes around these. Rather, the game’s class system seems to have accreted somewhat haphazardly over time.

I still haven’t played all of the classes in NetHack, but I generally agree with dtsund’s sentiment. I think the root of the problem is that basic melee is just too good — or, at least, “good enough” — that any class can resort to melee with the proper gear. This means that classes without exceptional bonuses or starting gear end up playing like “worse Valkyrie” or “worse Barbarian”. There’s a lot of fat to trim.

My first thought upon reading dtsund’s notes were, oh cool, I wonder which classes he’s going to cut? He ended up not cutting any, and while I disagree I suppose it makes sense; his stated goal is to overhaul the classes by making the available ones more distinct, and not by doing any trimming.

Before getting down to business, dtsund proposes a few overall changes to the core mechanics of NetHack, starting with daggers.

Thrown daggers are currently altogether too powerful; they should serve as a comparitively weak ranged option for classes that have nothing else, inferior to dedicated ammo-based ranged attacks like darts and bows. The current state in which daggerstorms can be one of the game’s most powerful attacks, in some cases even more powerful than a volley of arrows, should be rectified. One possible fix would be to remove multishot from thrown daggers.

Magicbane should be effectively useless as a damaging weapon and probably shouldn’t be an endless source of cheap, clean Elbereth either; it’d still be an excellent wielded artifact without these (perhaps it could be turned into a knife).

I’ve never actually played a character that relied on throwing daggers, or on Magicbane, but I have enough working knowledge of the game to understand what these changes mean and what reprecussions they have for classes that do rely on them. I think his assessment here is bang on. The characters best served by these very general strategies tend to have class-specific things that get overshadowed because the easier catch-all option exists. It’s dumb that Rangers can do more damage with daggers than a bow, and it’s dumb that Wizards can walk mindlessly into melee because they have such a good weapon. Weakening strategies that are universally good is a clever way to emphasize the strengths individual classes already possess.

The split between Int casting and Wis casting should be removed. It removes mechanical consistency for the sake of flavor and at the expense of gameplay to boot. Additionally, Wisdom should have a vastly greater effect on Pw regeneration, allowing some classes to restore their magic much more quickly than others at the outset). Maxed-out Wisdom should confer Pw regeneration roughly equivalent to one-quarter to one-half of an Eye of the Aethiopica. Elevating Wisdom should be as difficult as raising Intelligence currently is. Allowing classes to have spellcasting prowess decoupled from Pw regeneration at the start would be a good thing, helping distinguish classes who can occasionally use potent magic from those who can spam, but only with lower-level spells.

My personal feeling is that spellcasting is redundant in a game that already has wands and rings and scrolls and who knows what else, but if you’re dead-set on keeping the mechanic, I suppose this is a good way to go about it. A lot of classes in NetHack already start with spells the designers felt they should be naturally good with, and dtsund continues this trend as he overhauls the classes, but to me it feels artificial. A lot of the starting spells could just be simmered down into class abilities with cooldowns.

While there is some merit to varying the starting inventory, all instances of “you have a 1/5 chance of getting this useful thing, and no compensation if you don’t” should be removed.

This is a good example of updating a game mechanic that made sense in the 80s but not so much today. I could go into detail about why, but hopefully the reasons are self-evident.

Pure Fighter: Valkyrie. Relatively little needs to be changed in this case, save the aforementioned blanket nerf to thrown daggers (which would reduce their ranged combat potential noticeably). As-is, the Valkyrie, from start to finish, already epitomizes the melee game. They may now achieve Expert in any melee weapon type, but aside from the boosts to Two Weapon Combat and Saber, this is unlikely to matter much in practice.

The pure melee class is unavoidable in a game where melee is so good, and any attempts to make melee worse would make the game too frustrating (and, more importantly, make it No Longer NetHack), so taking a very light touch with the Valkyrie is the correct move.

Pure Ranged: Ranger.

Starting inventory:
a +1 dagger
4 +0 daggers (new)
a +1 bow (always, even for gnomes)
50-59 +2 arrows (always, even for gnomes; never initially poisoned)
30-39 +0 arrows (always, even for gnomes; never initially poisoned)
a +2 cloak of displacement (always, even for elves)
5-7 cram rations (always, even for elves; range narrowed somewhat)
a blessed spellbook of slow monster (new)

God gifts: First sacrifice gift guaranteed to be the Bottomless Quiver, a magical tool which provides a number of arrows when applied if it has nonzero charges. The Bottomless Quiver will not be given as a sacrifice gift to non-Rangers; it may be wished for, but gives non-rangers only half as many arrows. The arrows produced will be ordinary arrows 80% of the time, silver arrows otherwise. It can be recharged indefinitely.

Quest artifact: The Longbow of Diana, when invoked, now provides the same effect as a blessed scroll of enchant weapon. +1 multishot over an ordinary longbow. No other changes.

I don’t see why the Ranger needs to start with any +0 arrows. Just give him 80-99 +2 arrows and call it good. The Slow Monster spell here is a good example of something I think could just be a class ability. Perhaps something akin to the way Hunters can mark targets in World of Warcraft to get bonuses against those targets.

I think what dtsund was shooting for re: Bottomless Quiver and Longbow of Diana was a way to ensure the Ranger has a free way to generate and enchant arrows. I don’t think this is necessary. Heavily-enchanted blessed arrows already have a low chance of breaking, and even if they didn’t, the multishot bonus afforded to the class still makes it a powerful weapon. I think what players would do here is just use the Quiver to enchant an artifact melee weapon, and use that instead. In practical terms I can’t see a Ranger needing to re-enchant stacks of arrows very often. One stack of maybe 100 blessed +5 poisoned arrows should be enough to last the game, once the player has the resources to make them.

Pure Magic: Wizard. Much of the Wizard’s design is cruft left over from the days before the Wizard Patch was incorporated, introducing the modern spellcasting system. The result is a class which, despite ostensibly being the game’s magic specialist class, is optimally played with rather little magic early on. The changes here are intended to give focus to the class, and in particular both allow and require it to lean harder on magic in the early stages of the game.

I don’t much like the Wizard class. What dtsund has done is removed most of the random magical trinkets they start with and replaced it with increased spellcasting ability. This would make the class more distinct, but I’m still not sure I would like it, and having never played a Wizard I’m not overly qualified to comment on it anyway.

It is my understanding that a lot of the play this class sees is from players doing challenge runs, because the starting trinkets allow for specially-crafted challenges and conducts. Removing the starting gear changes that aspect of the game a lot, but dtsund fixes that problem later.

HYBRID CLASSES: These classes combine the elements of the pure classes in various ways.

This is where dtsund loses me a lot, because I don’t think many of NetHack’s hybrid classes can really be salvaged. For example, he classifies Samurai as a hybrid melee/missile class, and does his best to really try an emphasize those two points… but every class is already a hybrid melee/something. It’s a problem that the current Samurai is just a crappy Valkyrie, but I’m afraid the overhauled Samurai would just turn out to be a crappy Ranger.

The best hybrid classes in NetHack are the one with interesting or helpful gameplay mechanics that are totally removed from melee or spellcasting. The Priest’s knowledge of beautitude, for example, or the Monk’s aversion to weapons and armor. From this angle, dtsund’s best work was taking some hybrid classes and making them “specialist classes”, as we’ll see below. Unfortunately I think Samurai and Barbarian are just hopeless and redundant, unless you’re going to introduce a cool game mechanic for them to work with.

Ranged/Magic Trickster Hybrid: Rogue.

Quest: In addition to being difficult to reach, the Master Assassin should be nearly impossible to kill via conventional means – not because he’ll kill you first, because he certainly won’t, but because he’s too durable and tends to spam self-healing magic. His stoning resistance should be supplemented by poison resistance. The level’s leprechauns should be replaced by nymphs and moved closer to the nemesis, for reasons that will be clear shortly. Rogues are expected to take the Bell of Opening and the Master Key of Thievery via more underhanded means. Unlike most quest nemeses, the Master Assassin should be generated asleep, rather than meditating, giving players the opportunity to teleport him off of the items (especially since Rogues have intrinsic stealth). Failing this, polymorphing into a nymph, taming some nearby nymphs, or putting the nymphs under the influence of conflict will allow the player to steal the key items from an awake nemesis. If the player must resort to direct combat, it is essential to disable the nemesis with sleep or paralysis to prevent him from healing. If nothing else, death rays should still work against him.

I haven’t played a Rogue, and while I think dtsund’s idea for a quest overhaul is really cool I am skeptical that it is a compelling reason to play the class. The quest is a very small part of the game, after all.

Avoidance: Archeologist. Archeologists are poor fighters, but excel at avoiding combat.

Alignment: Lawful, Neutral.

Race: Human, Dwarf, Gnome.

Starting inventory:
a +2 bullwhip
a +0 leather jacket
a +0 fedora
3 to 6 uncursed food rations
a +0 pick-axe
an uncursed wand of digging
an uncursed touchstone
an uncursed sack
an uncursed oil lamp
12 to 15 +0 grappling hooks (new; see below)

Other details: Grappling hooks are now much lighter and may stack with one another. Archeologists specifically may apply them to nearby holes to create “climbing shafts”: two-way passages between floors, effectively serving as alternate staircases. The destination square for the climbing shaft, and the square from which the Arc can climb back up, is the closest square on the next floor down in X/Y coordinates to the shaft. Shops are ineligible to be destinations. Archeologists may dig through walls using a pick-axe in a single turn, and through a floor in three. When breaking through the floor, an Archeologist with a nonzero number of grappling hooks in his or her main inventory is given a y/n prompt to throw a grappling hook up to the edge of the pit. TODO: Give Archeologists some means of gaining experience aside from fighting.

Archeologist is already one of my favorite classes, and I think the changes dtsund makes here really do them good. I love the idea of a class that can make his own paths through the dungeon at will. My thinking is this is would be a good place to scavenge some of the attributes from Rogue: a class that can move unhindered, carefully pick his battles, and use the element of surprise to his advantage.

Item User: Tinker. Wizards, above, lost much of their domination over the item game. As no other class fits such an item-heavy theme either, the game has room for an additional class to fill the niche. They have basic melee skills, since they can’t kill everything with wands.

Alignment: Lawful, Neutral.

Race: Human, Gnome.

Starting inventory:
a +0 quarterstaff
a +0 leather armor
four random scrolls (not fire, amnesia, identify, or blank paper; all of different types)
a blessed scroll of identify
five random potions (not hallucination or acid; all of different types)
three random rings (not aggravate monster, hunger, levitation, or +0 chargeable rings)
a random powerful attack wand (fire, lightning, or cold)
a random other wand (not fire, lightning, cold, wishing, polymorph, or nothing)
an uncursed magic marker with 70-90 charges
a sack

Initial Basic skills: Quarterstaff.

Skill caps:
Restricted: Attack, Defense, Enchantment, Pick-Axe, Saber, Club, Mace, Morning Star, Flail, Polearms, Spear, Javelin, Trident, Lance, Bow, Shuriken, Whip.
Basic: Escape, Knife, Axe, Broadsword, Two-Handed Sword, Scimitar, Sling, Crossbow, Dart, Two Weapon Combat, Riding.
Skilled: Divination, Short Sword, Long Sword, Hammer, Boomerang, Unicorn Horn, Bare Hands.
Expert: Matter, Quarterstaff, Dagger.

Intrinsics: Shock resistance (level 1), Fire resistance (level 7), Polymorph control (?) (level 20).

Stats: Low Strength; high Constitution to help them carry all their equipment. Good Int, decent Dex, lousy Wis.

God gifts: First sacrifice gift guaranteed to be a very heavily-charged wand of polymorph (think 15-20 charges).

Quest: I got nothin’. Average rooms-and-corridors, nemesis of average difficulty, perhaps.

Quest artifact: The Pen of Shakespeare (Lawful by default). Artifact magic marker that writes the desired spellbook or scroll without fail, charges permitting. Starts with 99 charges, is recharged to 99. Grants double wand damage when carried, can be invoked for charging as from the PYEC.

Other details: None.

I’m pretty cold on this class. I understand the idea: take the great starting gear from the Wizard and make “put great items to good use” its own class. But putting great items to good use is one of the core features of every class. By the mid-game you have a huge collection of a variety of things regardless of role, which I think would cause the Tinker to lose most of its identity rather quickly. By DL20, everyone is a Tinker.

However, I do see a hole in the game caused by the Wizard’s overhaul that we can fill with a new role. Without a class that starts with lots of random magical gear, how will top players seed their crazy conduct runs? To that end, I make this role suggestion:

Challenge Seed Class: Adventurer

Alignment: Any

Race: Any

Starting inventory:
a +0 dagger
a +0 leather armor
a 1:2 wand of wishing

Initial Basic skills: Dagger.

Skill caps:
Everything can be #enhanced to Skilled. Three skills can be further #enhanced to Expert.

Intrinsics:
Upon reaching L2, L7, L13, L21 and L30 the Adventurer receives a prompt asking what intrinsic he should gain. This works just like wishing for an item. The list of legal options is limited to the following: any resistance except stoning and magic, searching, stealth, speed, warning, see invisible, teleport control and polymorph control. An Adventurer that loses an intrinsic upon being level-drained my select a new one when he levels up again.

Stats: Average across the board.

God gifts: First sacrifice gift will be selected from a short list of ascension kit items not already in the player’s open inventory. A cursory list of potential gifts is: boots of speed, gauntlets of power, helm of brilliance, amulet of reflection, cloak of magic resistance and wand of death. Upon receiving one gift the game will start cycling through random artifact weapons and/or spellbooks, as normal.

Quest: Randomly selected from the list of possible quests, similar to how Priests randomly select their deities. The game will not generate a quest if that quest’s artifact already exists. Once the quest generates, that quest artifact cannot be wished for. If all quest artifacts exist before the quest generates, no quest generates and the game becomes unwinnable.

Other details: The Adventurer doesn’t break wishless conduct until making his third wish. (Wishing for artifacts breaks both conducts immediately, though.)

The goal of this class is to offer a lot of very early game versatility for players who want to create specific challenges without doing lots of savescumming. For example, a player attempting a foodless run might use his wishes on a ring of slow digestion and an artifact weapon. I do not believe Adventurer would be a good class for starting players; even wishing for the three best items in the game in the first room is not enough to carry an unskilled player very far — a lesson every seasoned NetHack player knows all too well!

Anyway, that’s enough from me. Let’s get back to dtsund’s class changes.

Berserker/Tank: Caveperson. Caveman/Cavewoman, in vanilla NetHack, is yet another of those “basically Valkyrie but worse” melee classes. This would be fine if it were intended as a challenge role, but Tourist and (in some variants) Convict both fit that better. Here, I propose to change it into something dramatically tankier and stronger, but with severe restrictions other characters don’t have to deal with.

Intrinsics: Sickness resistance (level 1), Speed (level 7), Warning (level 15). Sickness resistance is new. +2 innate AC per level up to level 8, and +1 thereafter. +1 damage per attack every three levels.

Quest: Given their new restrictions, they could certainly do with a good attack, even despite their added intrinsics. It should be possible to find at least one trident during the quest, as that’s one of the best options a Caveperson has for twoweaponing; perhaps horned devils are common Quest foes.

Quest artifact: Unchanged.

Other details: Cavepersons are incapable of communicating in modern terms. This means they are incapable of reading scrolls or spellbooks (the Book of the Dead’s ancient runes are contemporary for him, though), cannot engrave Elbereth, cannot chat except during the Quest, and cannot name creatures to genocide or items to wish for. When given a wish, a Caveperson may only indicate what he or she wants by gesturing to an existing item in his or her inventory, duplicating it. They can, however, pray. Owing to their primitive nature, they have a pet-like ability to sense curses; this may help them find naturally enchanted armor.

I really love all these changes, and I think the Caveperson is dtsund’s best overhaul. Here we have a class predominantly defined by its drawbacks. We already know that is fun and interesting, because Monk. The intrinsic protection and damage neatly offset the loss of enchant scrolls and E-squares. The restrictions on wishing offer a unique but not insurmountable challenge. This is a Caveperson that would actually be worth playing.

Petmonger: Healer. Healers are designed to rely on pets, and for the most part this is already true of their early-game play. It could stand to be encouraged a little bit more later on, but this is pretty much already accomplished by the spell school reorganization allowing them to cast Create Familiar.

God gifts: First god gift guaranteed to be a blessed figurine of a decent but not overpowered pet, like a gargoyle.

Other details: Healers might be given the ability to use cursed figurines as uncursed, uncursed as blessed, and blessed with 100% reliability, flavored as healing a cursed being.

These changes both go in the right direction, although I think it would be better to scrap the Healer entirely and invent a new pet class from the ground up. Just off the top of my head, a class for which taming is an ability rather than a spell (similar to how Priests can turn undead) would be an interesting and challenging way to cater to that playstyle. Instead of taking the Healer as the base, take Warcraft’s Hunter or Final Fantasy’s Summoner.

Cavalry: Knight. Knights are already designed effectively as mounted fighters; as with the Healer, all they really need are a few minor tweaks to help encourage this throughout the game.

God gifts: First god gift guaranteed to be the Lance of Longinus, an artifact lance. Identical to an ordinary lance, save that it cannot be broken.

Quest: Unchanged.

Quest artifact: The Magic Mirror of Merlin, instead of telepathy, magic resistance, and double spell damage, confers reflection and controlled flight to the bearer and any steed he or she may be riding.

Having just played a Knight myself, I love these changes as well. I might even go so far as to replace the Magic Mirror with a powerful artifact Lance that can be #invoked to convey resistances and lifesaving to whatever mount you’re on. My beef with mounted combat is that it always felt like a nice perk when I had it, but it just wasn’t quite powerful enough for me to actively go out of my way and make it work.

Back-loaded: Tourist. Tourists already have a weak early-game paired with a somewhat versatile late-game; no significant changes need to be made here, as the class is already working as intended and has a unique identity.

I agree wholeheartedly here. Tourist might be the most unique class in the game, in fact.

So that’s a lot of notes. I think these are all wonderful ideas from the standpoint of making the existing NetHack classes more enjoyable.

I really, really don’t think that’s what the game needs, though. These are the kinds of changes you arrive at if you start with taking for granted you’re going to have a Samurai, and then trying to figure out how to make the Samurai distinct from the Knight and the Valkyrie and the Barbarian. What I think the game needs, here in 2014, is for someone to really question whether or not we need a Samurai. Take a chisel to everything and build something modern, but still fundamentally NetHack. To use a D&D analogy, dtsund is trying to revise 3E, when what we really need is for someone to design 4E.

1 comment to dtsund’s NetHack Overhaul

  • dtsund

    Alright, rebuttal and explanation time! I don’t disagree with as much of this as you likely think I do.

    “My first thought upon reading dtsund’s notes were, oh cool, I wonder which classes he’s going to cut? He ended up not cutting any, and while I disagree I suppose it makes sense; his stated goal is to overhaul the classes by making the available ones more distinct, and not by doing any trimming.”

    My initial instinct was to do exactly what you said; in the IRC log (from #nethack4) at the end of the document, I proposed as much (actually, I proposed throwing out *all* of the classes and starting from a blank slate), but was shot down because “removing classes causes riots; making them more interesting doesn’t”. So I adjusted my goals for the document.

    “My personal feeling is that spellcasting is redundant in a game that already has wands and rings and scrolls and who knows what else, but if you’re dead-set on keeping the mechanic, I suppose this is a good way to go about it. A lot of classes in NetHack already start with spells the designers felt they should be naturally good with, and dtsund continues this trend as he overhauls the classes, but to me it feels artificial. A lot of the starting spells could just be simmered down into class abilities with cooldowns.”

    I do disagree with this part, though; as someone who tends to favor spellcasting classes myself, I think there is a place for innate spellcasting as distinct from magical items. You *could* have class abilities, but spellcasting serves as a universal library of class abilities that might be good to have on more than one class.

    For what it’s worth, though, class abilities with cooldowns have already been implemented; you can find them in SLASH’EM, that greatest (in the sense of size, not necessarily quality) of NetHack variants.

    The points about Ranger are well taken; perhaps it might help to weaken ranged combat in general in ways that the Ranger’s bonuses let him ignore, like increasing the destruction rate of arrows or disenchanting them by a point when fired.

    “It is my understanding that a lot of the play this class sees is from players doing challenge runs, because the starting trinkets allow for specially-crafted challenges and conducts. Removing the starting gear changes that aspect of the game a lot, but dtsund fixes that problem later.”

    That’s only partly true; the Wizard also trades fragility for the ability to completely dominate the game with his spells. They’re one of the more popular classes even in non-challenge runs.

    “I haven’t played a Rogue, and while I think dtsund’s idea for a quest overhaul is really cool I am skeptical that it is a compelling reason to play the class. The quest is a very small part of the game, after all.”

    I’ll cheerfully acknowledge that Rogue is one of the weaker overhauls in that list; I had difficulty really coming up with an identity for the class distinct from both Priest (utility magic) and Archeologist (avoidance), and I’m not sure I succeeded (I think Samurai and Barbarian were a bit more successful; have you played a game reliant on attack spells yet?).

    “I’m pretty cold on [Tinker]. I understand the idea: take the great starting gear from the Wizard and make “put great items to good use” its own class. But putting great items to good use is one of the core features of every class.”

    I think there’s room in the game for a class that is both better at using items (especially at using them early, when most classes are neck-deep in the ID game) than anyone else and more reliant on them because his other abilities are substandard. Perhaps this class should’ve been Rogue, though.

    For what it’s worth, both your idea to seed challenge runs and my random loadout will likely wind up obsolete in NetHack 4; ais523 really wants to base starting inventory on a point-buy system.

    “These changes both go in the right direction, although I think it would be better to scrap the Healer entirely and invent a new pet class from the ground up.”

    I would’ve done just this, if I were designing a document for my own use. Innate taming on a cooldown would be fine (perhaps it could be snuck in as an invoke effect sacrifice/crowning/Quest artifact); on a from-scratch class, I’d probably also put a magic whistle in the inventory.

    “To use a D&D analogy, dtsund is trying to revise 3E, when what we really need is for someone to design 4E.”

    This isn’t necessarily wrong, but getting to such a 4E would be dramatically more work, particularly given the shambolic state of the game’s spit-and-#prayers codebase. Implementing the Arc’s climbing shafts alone would be a heinous amount of work, and is feasible as a suggestion for NetHack 4 only because the game’s interlevel code is slated for overhaul there for unrelated reasons.

Leave a Reply to dtsund Cancel reply

You can use these HTML tags

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>