User:Phol ende wodan/Balance issues

This document outlines some of the balance issues with vanilla NetHack, and discusses possible fixes to them. I'm trying to address big issues first, so smaller ones can be considered in light of the fixes to the big issues.

Possible fixes for each balance issue aren't intended to be a pool from which only one could be chosen; any amount of them could be taken together, though some might contradict each other.

If an idea is described that I know came from a variant, proposal, or some other outside source, it'll be noted as such. If not, it's something I came up with myself. However, it's quite likely that many of the ideas are similar or identical to things others have thought of, and I simply didn't know about it at the time.

This document is a work in progress. Ideas and balance considerations will be added as I devote more attention to them.

Special thanks to Jonadab, Chris, NeroOneTrueKing, Amy, Ais523, FIQ, and Red kangaroo for their comments and contributions.

Why balancing is beneficial

Imbalance in a game can lead to boredom and annoyance. When a game isn't balanced very well, it's easier to identify the best available strategies. The player might feel compelled to pursue these strategies because they objectively give the highest chance of winning, which can become boring. Then, if the player tries to pursue an optimal strategy but can't, they can get annoyed. A similar topic has been discussed before; in terms of that essay, imbalances cause the headroom of part of the game to decrease.

In NetHack, the game provides a large amount of variety through random generation of levels, objects, and monsters. It's a very replayable game because the player must adapt to their current situation with the resources they have available, which are rarely the same. Yet this is much more true for some parts of the game than others: in the late game, for example, large components of most strategies converge (wear dragon scale mail, be able to death-ray the Wizard of Yendor at short notice, and save up scrolls of gold detection to locate the Planes portals, to name a few examples). There's no requirement to do these things, but they sure make the game easier to win.

The end goal of balancing a game should be to make it more strategically interesting and fair while preserving its variety as much as possible. Balancing might involve reducing variety, but hopefully only the bad kind.

Practical problems

For all possible fixes listed on this page, I am more or less ignoring the programming side of things, unless one of them would require such massive overhauls as to be impractical.

However, there are some issues with implementing balance fixes that lie in player approval of the fix: the majority of rebalances here take the form of nerfs, and nerfing things tends to build up distaste with the variant or version they are implemented in. The better balance does not stand on its own, and a variant consisting mostly of balance fix nerfs and nothing else is unlikely to be very popular.

Nutrition

NetHack is supposed to have hunger driving the player deeper into the dungeon, but currently there is so much food that this only really happens in the early game.

Balance issues

Egregious sources of nutrition in the dungeon include:

• Randomly generated food. Even in Gehennom, food makes up 16% of all random objects. This is most commonly food rations, which provide 800 nutrition.
• Soldiers, each of whom carries about 300-1200 nutrition in lightweight rations.
• Kops, who can be farmed for an infinite source of cream pies.
• Corpses. As the game goes on, monsters get larger and their corpses provide more nutrition. A non-vegetarian character could probably survive through Gehennom on randomly encountered dragons alone at 1500 nutrition a pop. On the other hand, corpses in the very early game seem pretty well balanced, with many things being inedible for an early character and nothing that provides a lot of nutrition.
• Prayer. You can currently rely on it to go without food through the entire non-Gehennom game. Not completely reliable as it won't keep working if you roll a particularly long prayer timeout, but overall you gain enough nutrition to last you until your next prayer.
• The ring of slow digestion. Currently it lets you carry around a couple pieces of food and avoid nutritional issues for the rest of the game. It basically takes all these egregious sources of nutrition and multiplies them by 20. Some players even start scum for one so that they can avoid the whole issue of nutrition.
• Polypiling. It's mainly a problem because it allows low-nutrition food to be transformed into the most common type of food, food rations, which are high-nutrition (and other high nutrition/weight foods as well).
• Trees, on levels where there are many accessible trees (e.g. the Ranger quest). Especially so since fruit from trees does not rot.
• Create monster sources. Allows on-demand generation of monsters, and therefore corpses. The spellbook lets this happen indefinitely as long as the player's power regeneration can keep up. The scroll, when cursed, also gets a lot of mileage, but the player probably won't get enough scrolls and markers off the monsters to keep going. Wands and the bag of tricks run out eventually and will usually not be worth recharging unless the player has the Platinum Yendorian Express Card. Therefore, the spellbook is the worst offender here, and probably one worth looking at.
• Stone to flesh. For a character who can cast it with any reliability, boulders will be encountered far more often than once every 2000 turns. With giant generation and monster tunneling providing additional boulders, the supply of boulders in the game is practically infinite.
• Horns of plenty, marginally. Tourists or anyone else who wishes up the Platinum Yendorian Express Card can use one for infinite food (the expected amount of nutrition from even an uncursed charge is much greater than the artifact timeout).

Not balance issues

Sources of nutrition that I don't consider to be too plentiful or powerful:

• Sokoban. There is a lot of food in it, but that's intentional and helps make it a strategic target for characters low on food.
• Delicatessens. Ones that the player is likely to see (in Minetown) are small. Ideally, a randomly generated delicatessen should be a boon to the player because it's not otherwise easy to find food, and a large one should mean that the player probably doesn't have to worry much about nutrition for a while. Though perhaps huge ones should be limited somehow, by turning into general stores over a certain size or something.
• Trees, when there are not that many of them (Medusa's Island 3, Bustling Town), since the food they produce doesn't have that much nutrition, and early characters have to watch out for killer bees.
• Amulets of life saving. They are a pretty valuable resource to spend on regaining some nutrition when being lifesaved.

Balance considerations

• One reason for food being generated as often as it does is that it facilitates vegan and vegetarian play. Reducing permafood in favor of eating corpses will make those conducts more difficult.
• Corpses are the primary means for intrinsics for most characters. Reducing the amount of corpses in favor of permafood would make getting intrinsics more difficult.
• Kops shouldn't have their cream pies removed, but something has to be done about their dropping 1-3 unused pies whenever they're killed.
• There should be enough food around for a player to have some leeway in backtracking a bit - returning to a stash, returning to Minetown, etc. But eventually, the player's available food should run out, and they should have to go deeper.
• Walking back and forth between different places is necessary in some games, so it should still be at least feasible to do so once or twice, though perhaps not for too far of a distance or too many times.
• The amount of monsters encountered when exploring a level for the first time (ignoring special rooms) versus the amount of monsters encountered when wandering around that cleared level. Sometimes this ratio approaches 1:1, which doesn't make sense in context with the general idea that you're clearing out the level.
• The whole too-much-food balance issue gives rise to a bunch of other ones, like spell spamming and regeneration abuse.

Idea #1: cut the chance of random objects being comestibles sharply

• All the way down maybe to 0% in some places.
• Does not affect guaranteed food in any way (delicatessens, Sokoban, the Castle).
• Vegetarians have a rough time of it. They have to survive on lichens and starting food, hoping it lasts long enough to flee to Sokoban and stock up.

Idea #2: adjust corpse nutrition so that they don't give as much.

• Monsters that grant more than a couple hundred nutrition are rare.
• This would have the effect of pushing the player more towards permafood.
• Avoids the concern above about resistances, since neither the number of corpses nor their chance of giving intrinsics is not reduced.

Idea #3: don't generate food as a death drop, since a death drop can only happen when there's going to be a corpse anyway.

Idea #4: do not generate random food below some place in the dungeon.

• With a hard threshold like this (say level 10), this would just encourage players to hang out on levels above that farming death-dropped permafood, unless death drops rarely or never generate food.

Idea #5: make the chance of generating random objects as food steadily decrease based on dungeon level and possibly dungeon branch.

• Has interesting implications if you do that with other object classes. Right now the probabilities are just fixed per-branch and in containers.

Idea #6: in addition to corpse rotting, ordinary types of food get rotten over time, and can rot away if not eaten.

• The amount of time to become rotten is specified per item type. Examples: egg = 80 turns, banana = 100 turns, apple = 500 turns, food ration = 1500 turns, cram ration = 2500 turns, lembas wafer = 10000 turns
• This should probably be randomized somewhat so that players can't name their food with the turn they picked it up and know when it's still good to eat.
• Perhaps the player should be shown warnings that food is going bad - taking an example from GruntHack corpses, "an off-color cram ration", "a smelly cram ration", "a stinky cram ration" - though the food becoming this should also be randomized per-item so the player doesn't actually know when it becomes rotten.
• The amount of time for a comestible to rot away completely could either be also fixed per item or just a multiple of the time to rot, like 1.5 or 2 times.
• Tins don't rot at all, though they still might have a preparation method of "rotten" when opened. This makes tin openers more useful and blessed tins of spinach become the best permafood in the game. Player-made tins will only give 50 nutrition and are heavy, so I'll say tentatively that tinning kits and player-made tins don't need to be nerfed yet.
• Lizard and lichen corpses still do not go bad, but lichen corpses may possibly need to be lower nutrition to prevent players from farming or hoarding them on low levels. This is already true for lizard corpses; they don't provide enough nutrition to last the player until they encounter more lizards.
• Food stored in an ice box does not age, just like corpses.
• Ice boxes now can generate containing non-corpse food as well as corpses.
• Blessing the food should possibly make it last longer without rotting.
• Rotten non-corpses should be treated either similar to how cursed food is now ("Blecch! Rotten food!" plus possible negative effects and nutritional value severely reduced)
• Or, rotten food might even be able to become tainted (though not all food items, and only when it's very close to rotting away).
• This hurts players who like to peek at levels early (to generate its monsters at a lower difficulty, for instance) or otherwise are on levels temporarily without exploring the whole thing, for whom all the food may be rotten when they return. This could be ameliorated by having the age of the food increment only when on the active level, which is in line with the concept that "time is frozen on other levels" but I don't know how easy it'd be to implement. It could also lead to degenerate strategies with players leaving food next to the downstairs, or on a branch level they won't have to visit again, so that it will be fresh when they come back up.
• Polypiling food does not change its age. If food polymorphs into something which should have rotted away by now, it immediately rots away. Possibly, tins can no longer be generated by polypiling.
• This hurts Sokoban as a food source, because much of the food found there will expire in the time it takes to complete the levels. To ameliorate this, some of the food items could be guaranteed tins.
• Something may need to be done with the ring of slow digestion. I can see a lot more people startscumming wizards for slow digestion because it's the only way to avoid nutrition issues in the early game. Then again, that already happens to some extent, and a wizard is the only role that can be startscummed for this.
• Food can no longer stack, because there's no clear and fair thing that should happen when two food items with different ages stack together. This could make it annoying for roles like the Knight who starts with up to 40 apples and carrots. Maybe food of the same type with the same age can stack, but I'm not sure I'd like that.
• Food can be rotproofed and made free of any rot via a confused enchant weapon scroll, using the same erodeproof bit that other items have. Since food doesn't stack, you can't rotproof a bunch of it at once. Since enchant weapon is fairly useful and this is clearly not the best use for it, I don't think this has much potential for abuse.
• Rotproof food can show up in certain places (e.g. comestibles in starting inventory like food rations, but not "fresh" food like apples and carrots). Maybe randomly generated food has a small chance of being rotproof as well, or a fraction of the food in Sokoban, or something. Not sure whether the rotproofing should be obviously visible to the player or requires a full identify.

Idea #7: steeply decrease the chance of soldiers carrying C and K rations. Clearing out Fort Ludios should net the player only a few of each.

Idea #8: find more things to incur small additional hunger penalties, and possibly just up the hunger rate.

• Jumping's hunger penalty is oddly variable, make it more predictable, perhaps based on distance.

Idea #9: monsters besides Famine who specifically destroy food or cause hunger.

• Probably R class.
• jonadab suggested locusts, a class of course, which destroy food in inventory and possibly even in containers, and multiply while doing so. The effect of such a monster, even if it has weak attacks, could leave a player hungry and with no ready source of food.

Idea #10: heavily nerf the sources of arbitrarily renewable food; in particular, sharply drop the nutrition conferred by all corpses.

• This originally came from Chris_ANG, based on the view that there's nothing wrong with the permafood supply as it stands, and it's good for the player to be able to have enough of it to meander - but eventually, it needs to run out and the player can't survive without exploring new areas for more food.
• Reducing corpse nutrition makes flavor sense in that the hero generally doesn't know how to (and isn't anyway) butcher and prepare a corpse for eating, and isn't getting the maximum nutrition out of it that they might. (A side idea is that Barbarians do have more knowledge in this area, and as a role benefit can get more nutrition from a corpse than other roles.) Another flavor justification is that a corpse is not designed to provide a good complement of nutrients like a food ration is. At the same time, by not touching corpse intrinsic behavior, it avoids all problems with messing up that balance.
• Perhaps the trickiest part of implementing this is how to shut off the create monster -> altar farm -> pray for food feedback loop.

Gehennom items

Many objects generated in Gehennom are useless by the time the adventurer comes across them.

Balance issues

• 20% are weapons (the player likely has already found a good weapon, and even an artifact generated here is likely not to be worth it).
• 20% are armor (the player likely already has most needed armor, and even if a rebalancing change made it so that's not the case, the multitude of ring mails, orcish chain mails, and leather armors isn't useful to anyone).
• 16% are food (see above; most players don't need extra food in Gehennom).
• Scrolls and potions are 1%, and spellbooks 0%, for reasons I'm not aware of. Maybe because it's supposed to be hell and they all burn up? (Of course, Gehennom isn't very hellish, but that's not a balance issue.)

Idea #1: cut the percentages for weapons and armor and food down to small values

• Then increase the remainder proportionally to bring it back to 100%.
• But reduce the overall item generation rate to compensate.

Idea #2: armor and weapons that generate in Gehennom have desirable qualities

• Good positive enchantment, erodeproof, etc
• Still not very good since the post-Castle player should have most of an ascension kit ready, which is a wish balance issue.

Idea #3: (Chris suggestion) generate few "magical" items outside Gehennom, and very many in it. Thus, the player has to explore Gehennom to gather the components of their ascension kit. Also perhaps make demon lords optional, but they each guard a specific ascension-kit item, which the player can fight them for.

Gold

Gold has little use after buying 9 points of protection (at post-quest experience levels). If you can afford another point, you can probably afford to buy everything useful from every shop in the game. Yet it continues to be scattered all over Gehennom.

Balance considerations

• Fort Ludios is a significant and accessible source of gold in the Dungeons of Doom. There's probably more gold in Gehennom overall, but it's a lot harder to collect it all.
• The uncertainty of whether Ludios even exists steeply increases the variability of the total gold in the dungeon.
• Collecting gold in Gehennom is rather pointless beyond maybe grabbing some from the first level below the Valley and buying clairvoyance from the priest. Asmodeus and Baalzebub will be happy with taking two of your three gold pieces.
• The gold in Gehennom can be a nice cushion if you angered the gods and lost your divine protection somehow.

Idea #1: put something late in the game that takes a lot of gold.

• Perhaps Asmodeus and Baalzebub can demand, say, ninety percent of the player's visible (or even hidden?) gold, but with a high minimum amount, $10000 or $20000 or even more. They could either allow the player to remove the gold from containers first (reasonable but possibly difficult to implement), or simply attack if the player can't pay up. On the other hand, Asmodeus and Baalzebub aren't really scary enough in vanilla to bother paying off...
• Asmodeus and Baalzebub can also just camp on the downstairs until the player pays up, which allows the player to go back for more gold and makes a bit more sense flavor-wise (they'd rather have the gold and not risk their life fighting the hero; this is echoed a bit in what they do now, where they'll accept a bit less than their initial demand). If the player comes up from below while they're still on the downstairs and haven't been paid, they become angry.
• The upper bound for the amount of gold for safe passage should be very high, say $100000, so that it's not easy to walk around being sure you have enough gold to pay them.
• Chris suggested adding a djinn king who grants wishes for exorbitant amounts of gold in a response to jonadab's dungeon overhaul proposal. However, that would fit into dnethack a lot easier than vanilla.

Idea #2: have the number of gold piles placed on a level taper off with dungeon level.

• Possibly only after you enter Gehennom.
• This is a less compelling idea than having a large money sink in Gehennom, because any gold left in Gehennom will still be useless.

Idea #3: guarantee Fort Ludios to appear, or at least do something like UnNetHack which practically guarantees it.

• This makes it a strategic decision for the player to clear out, and the player can strategize about it with the certainty that it is in fact there.
• A way to guarantee it on top of Un's change is that if there are actually no vaults below dungeon level 10, stick the portal in an ad aerarium closet on the level above Medusa.

To-hit formula

In the midgame, to-hit stops mattering and you don't really miss anything again. Tends to be a contributing factor to the fact that a lot of roles play the same way after the midgame. Also affects the balance of single weapon vs single weapon + shield vs double-handed weapon vs twoweaponing.

Balance issues

• Once you get a luckstone and can maximize your Luck, that is a flat +13 to hit.
• Monks can overcome the -20 to-hit penalty and the loss of their XL/3 + 2 bonus with relative ease, to the point where monks can wear body armor through the late game with no real problems.
• For monks, the XL/3 + 2 bonus is fairly useless outside the early game anyway, when XL is low.
• Mid to late game characters can generally hit with anything they pick up regardless of whether they are Unskilled.
• Chris's analysis: "The root of the problem seems to be that nethack was based on 1e/2e D&D's AC system, in which ACs are limited (barring exceptional outliers) to the range -10 - +10 and clustered around +2 - +9. This range worked well in a system where to-hit bonuses were limited to a maximum of +9 from level, +5 from weapon enchantment, +3 from strength, and +3 from skill (and in which the actual bonuses were typically much lower than these maximums). This range of ACs does not work as well in nethack's system, in which applicable bonuses can include +14 at minimum from level, +13 from luck, +7 from weapon enchantment, +3 from strength, +4 or more from dexterity, and +3 from skill. On the flip side of the discussion, nethack does not have the same desired balance point that D&D had. Specifically, because the nethack character is expected to single-handedly fight and kill many more monsters at one time than a D&D character was, nethack characters should be able to hit their opponents much more easily than a comparable D&D character could. Another note: in 1e/2e D&D, a character in non-magical plate mail and wielding a shield had AC 0 (helm, gloves, boots, and cloak were not counted as armor). In nethack, a character in similar armor has AC -4 (+7 from plate, +2 from shield, +2 from dwarvish iron helm, +1 from gloves, +2 from shoes, and +0 from dwarvish cloak). This difference becomes far more pronounced at high levels, since each piece of armor in nethack can be enchanted for up to +5 additional AC per body slot, whereas in D&D only the armor and shield were available for holding magical armor, and getting +5 in both was far from certain."

Not balance issues

• Weapon enchantment. Enchantment influencing both accuracy and damage is very visible and understandable.

Balance considerations

• Strength and dexterity bonuses, encumbrance penalties, monster size bonuses, monster-specific bonuses and item-specific bonuses are mostly fringe cases or have small +1 or +2 effects; they're not really important enough to worry about at the moment.

Idea #1: heavily nerf the effect of luck on to-hit.

• With such a nerf, it would now matter a little, but not that much, and not as much as XL.
• Perhaps use Luck/3, Luck/4 or Luck/5 instead of Luck/1.

Idea #2: nerf the effect of XL on to-hit.

Idea #3: Scale monster base AC generally with monster difficulty.

• Demogorgon's base AC is only -8, which makes him easier to hit than a disenchanter.
• It never goes below -10. There might be some reason for this, but I'm not aware of anything like that.
• It should generally be below -10 for Gehennom and endgame monsters.
• On the other hand, an endgame character with a good weapon and a high level shouldn't be struggling to hit things like they were at XL1. Except if they're a monk wearing body armor.

Idea #4: overhaul the entire AC, damage, and to-hit system. This flippant comment has evolved into an actual proposal to overhaul the entire AC, damage, and to-hit system.

Idea #5: buff the effects of strength, dexterity, and skill on to-hit in combination with nerfs to other things.

Idea #6: (UnNetHack implementation) Flat 25% chance of missing when fighting with an unskilled weapon. This has annoyed many players trying to wield cockatrice corpses.

Idea #7: (suggested by Chris) Weapon enchantment from the scroll times out.

• Weapons generate at a certain enchantment, and if they are enchanted more from "artificial" means, it eventually goes away and the weapon reverts to its original enchantment.
• Means that a non-artifact +5 weapon in a skill you can only enhance to Basic might actually be better than artifacts or your favored +0 weapon.
• Would probably be quite a shock to players and make many people angry. Not that that means it shouldn't be attempted.

Energy regeneration

Spellcasting roles have to conserve spells and play as bad melee fighters because energy regenerates too slowly in the very early game.

Balance issues

• Rate of regeneration is based only on experience level.
• Wizards get a specific bonus to regenerate energy faster than other roles, but it's still slow even for them - an XL14 Quest-ready wizard will still only recover 1d3 energy every 12 turns. This is an average of 1 every six turns, meaning the wizard will have to wait 30 turns on average to cast another level 1 spell.
• The only way to get fast energy regeneration is via the Eye of the Aethiopica, which means that wizards have to wait till the quest for it, and everyone else either can't get it at all or have to wait till they have a wish and can suffer the artifact blast.

Idea #1: Rather than long delays between multiple points of regen, have shorter delays between 1 point of regen.

• Natural energy regeneration is always 1 point.
• Int and Wis play a role as well as XL in the delay between recovering points.

Idea #2: Energy regeneration (as provided by the Eye of the Aethiopica) regenerates one per turn, not d2 or d3 or d4 per turn.

• Makes energy regeneration follow the same rules as HP regeneration.
• The player's normal energy regeneration happens in addition to this instead of being overridden by the energy regeneration source.

Idea #3: (dtsund Class Overhaul proposal) Make Wisdom the primary factor in energy regeneration.

• Experience level and intelligence provide either a smaller impact or no impact.
• Intelligence probably becomes the spellcasting stat for all roles.
• Most of the numerous methods of exercising Wisdom are removed. It is now quite hard to gain maximum Wisdom, around as hard as intelligence.

Ascension kit degeneracy

Most roles use very similar items in their ascension kit.

Balance issues

• Dragon scale mail is objectively better than all other body armor and choosing to wear it doesn't represent any tradeoffs or other hard choices the player has to decide on.
• GDSM and SDSM are the only really used colors, to the point that there's a wiki article comparing the advantages of these two specifically. Occasionally you see a YDSM, but not often.
• There are only three mundane items that convey reflection, plus the Longbow of Diana (and Dragonbane in 3.6.1).
• There are only two mundane items that convey magic resistance, plus several quest artifacts and Magicbane.
• Speed boots are superior to all other boots, assuming the player can't perpetually cast speed or make speed potions. There is no other source of permanent very fast speed.
• Cloaks are not as bad, but it's really just a choice between magic resistance and protection. Robes may also be competitive. However, the stealth, invisibility, and displacement provided by the other cloaks is more or less useless in an ascension kit, since invisibility can be gained intrinsically, displacement is useless if you are already invisible, and stealth may be gained as an intrinsic but is little use for the end game when there are few sleeping monsters.
• There's no single optimum helm for everyone, but your role and fighting style usually have a first choice: brilliance for spellcasters, telepathy for non-spellcasters who want extrinsic telepathy, and elven leather for those who don't.
• Same for gloves - thanks to the to-hit imbalance, melee fighters and non-spellcasters should almost always wear gauntlets of power because they don't need to worry about their drawbacks, and spellcasters should almost always wear gauntlets of dexterity because they are the only beneficial glove that doesn't block spellcasting.
• Free action is considered a must, and there is only one source of it. Though apart from this, there are a large number of strategic choices for the other ring slot - polymorph control, levitation, teleport control, increase damage, protection, conflict.
• Life saving and reflection are the obvious choices for the amulet slot, Plane of Water magical breathing shenanigans notwithstanding.

Balance considerations

• The game is designed so you can't have it all. There is no way to have acid resistance, water walking, displacement, permanent drowning attack protection, conflict, regeneration, free action, life saving, magic cancellation 3, magic resistance, and reflection all at the same time. But some of these are clearly superior to the others, and most of them can be obtained at once. Hence the problem.
• Anything short of body armor, in general, can be quickly switched out for something else. For instance, one can switch out speed boots for jumping boots and quaff a couple potions of speed when about to enter the Astral Plane, and get both effects. But for most slots, such maneuvers don't happen.

Idea #1: add more items with good but non-intersecting effects

• Amulet of flying is one. Not useful on Astral, but it's good up until then.
• Body armors with low AC but really good effects, or even effect bundles. For example, silver ring mail that provides 2 AC, reflection, and MC3.

Idea #2: change other effects or behaviors so that existing unused items are worthwhile parts of an ascension kit.

• For example, under a poison resistance nerf where 100% poison resistance is not possible to be gained intrinsically, the only source of 100% poison resistance might be the amulet versus poison. If poison is made into a larger threat (enemies spawn with and throw potions of sickness, for example), this amulet becomes an important consideration.
• Reducing or stabilizing the jumping hunger penalty and making it easier to initiate a jump makes jumping boots compete with speed boots, since they are actually faster.
• Water walking boots prevent drowning attacks and don't wet inventory on a failed drowning attack.
• Crystal plate mail provides reflection.
• Add otherwise unobtainable properties to non-GDSM and non-SDSM dragon scale mails. Fourk and dNetHack do this, with things like RDSM providing infravision, WDSM protecting inventory from water, YDSM providing petrification resistance, and so on. The problem with this approach is though it balances dragon scale mail, the bar for other armors to compete with it is raised even higher.
• A less powerful way to balance dragon armor is in conjunction with a resistance nerf. Red dragon scale mail might look like a better option for Gehennom if it and the ring of fire resistance are the only way to gain full protection against the fire traps and attacks there.

Idea #3: make some effects useful only in some places and other effects useful only in other places; effectively meaning you can't wear a single kit of gear through the whole late game.

Idea #4: (planned Fourk implementation) Dragon scales are cloak-slot armor that provide 0 AC and only their special effects. They can be enchanted onto the body armor worn underneath.

• Removes dragon scale mail as a top-tier armor.
• The player must make decisions about body armor weight, AC, and magic cancellation, because dragon scale mail does not exist to provide the best possible weight and AC, and they can choose which dragon property to imbue their chosen armor with (assuming they have the scales).

Idea #5: (Red kangaroo suggestion) Make the player choose between good AC (plate mails), full magic cancellation (mithril-coats adjusted for less weight and less AC), and good intrinsics (dragon scale mails adjusted for more weight and much less AC, but with non-GDSM/SDSM having additional properties).

• dnethack is somewhat like this - plate mail/crystal plate mail grants the most AC but is heavy, mithril-coats is extremely light, have full MC and allows your DEX to contribute fully to AC (at maximum DEX it provides the second best AC, only behind plates). DSM grants less AC than either options but almost every color is buffed.

Artifact weapons

There are only a few artifact weapons worth using.

Balance issues

• If you can get to Basic in saber, Grayswandir is clearly optimal.
• If you like long swords instead, the Brands and Excalibur are the best.
• For chaotics, even rangers, Stormbringer is the best.
• The game is so biased towards these for damage output that even wizards might decide to wish or sacrifice for one.
• Quest artifacts offer some alternatives, but there's no way to get them short of wishing or finishing the quest. Some quest artifact weapons are considered pretty good (the Staff of Aesculapius), some are okay and may depend on external factors (the Sceptre of Might, the Tsurugi of Muramasa), some are commonly ditched in favor of other things (the Longbow of Diana).
• The Banes are more or less all crap, because they suffer from bonuses that only affect a certain type of monster. Werebane is considered slightly better only because it's effective on all silver-haters. The 3.6.1 buffs to Dragonbane, Ogresmasher and Trollsbane do not make them competitive enough to wield as a permanent weapon. Grimtooth is generally not worth it either.
• The goal for almost all characters is to build up to an artifact melee weapon. Since worthwhile artifact melee weapons are overwhelmingly blades, this means that mundane melee weapons which are not long swords, broadswords, sabers, or daggers almost never are worth training.
• Players with long sword restricted will actively try to unrestrict it by getting a long sword gift or being crowned while wielding one. Unrestriction seems like it exists only as a side effect of the god selecting a random artifact to give the player, where it would be ridiculous if the player wasn't able to use the gift effectively.

Idea #1: Gods will never gift the player an artifact weapon in which their skill is restricted.

• Crowning doesn't unrestrict skill either. Unrestriction is eliminated completely.
• This applies only to the first gift; subsequent gifts might be types the player is restricted in, but only if all unrestricted artifacts the god can grant have been generated already.
• Artifact base types may need to be adjusted so that every role has at least one weapon they can be granted first.

Artifact base types

They are all weapons, except for some quest artifacts.

Balance issues

• The only way for a player to get a non-weapon artifact is through wishing or if their quest artifact happens to be a non-weapon.

Idea #1: Include non-quest artifacts that aren't weapons, obviously. Armor in particular is an underused object class.

Weapon damage

Non-artifact weapons have some problems with damage.

Balance issues

• War hammers should deal lots of crushing damage, yet they deal only d4.
• Most polearms are as good as or better than most starting weapons.
• Martial arts are a joke, and are completely non-competitive with even normal weapons, let alone artifacts.
• Most weapons never get used simply because they don't have an artifact to work towards.

Resistances

Resistances, once you acquire them, are perfect. (This includes certain damage-prevention properties like reflection and magic resistance.)

Balance issues

• With a resistance, you will never take damage from that type again as long as you possess the resistance.
• Poison resistance is mandatory because of the poison instadeath.
• Reflection is mandatory because of disintegration blasts.
• MR is mandatory because of touch of death and wands of death, as well as some other protections against very annoying things like the destroy armor spell and polymorph traps.
• The other resistances, like fire, cold, and sleep, are not mandatory but can cause very annoying effects, so they're essentially mandatory.

Balance considerations

• Currently lots of the things subject to resistances are binary and
• Properties like reflection and magic resistance don't lend themselves well to partial resistances, since they can't be gained intrinsically.

Idea #1: you gain a partial resistance each time from its source (for edible resistances)

• This cannot go over 90% resistance.
• To be nice to vegetarians, eating jewelry for resistances either works 100% of the time or gives a larger gain than a corpse does.

Idea #2: (SporkHack implementation) you get 5-10% resistance from eating a source each time

Idea #3: Every source eaten gives you half of the distance to 100%. First is 50%, second is 75%, etc.

Idea #4: Separate concepts of "resistance" and "immunity".

• Immunity can only be gained extrinsically or through polymorph.
• Immunity is the only perfect defense that blocks absolutely everything, like vanilla resistances do currently.
• Resistances gained by reaching a certain level as a role or race could be converted to successively increasing partial resistances (cool but harder to implement), or just simply immunities.
• Resistances only cut damage by a percentage or reduce the duration of an effect (partial sleep resistance means you will still fall asleep for a shorter period of time), and have a maximum value of 90% or so.
• The player can see resistance percentages through enlightenment, separate from immunities.

AC / damage reduction

AC determining both whether you get hit and your damage reduction is odd and different from how other games do it.

Balance issues

• There is no way to quantify that a monster should be hit a lot of the time but resist a lot of damage, or be hard to hit but fragile.
• Heavy armor ought to give high damage reduction but actually harm your evasiveness.

Idea #1: (Fourk implementation) If a monster barely hits you, damage is heavily reduced on top of anything your armor does. If it hits you more cleanly, your armor is the only thing that reduces damage. This makes evasiveness and the monster's to-hit matter.

Idea #2: Armor does not improve evasiveness, it only reduces damage. This has grown into yet another full proposal.

Idea #3: Overhaul armor class with a strictly positive index.

• For simplicity, call it Def after the NetHack4 stat.
• This value is zero when you are naked and increases when wearing +0 armor, and for points of armor enchantment.
• Whether an attack hits is now according to the following: roll of 1d20 >= max(1, (Attacker's level) + (Prebonuses) - rn2(Def)) + Postbonuses

Alignment record

It's subject to mass inflation.

Balance issues

• You get fairly large alignment gains from killing hostile and crossaligned monsters.
• This means that unless you're pacifist, you should always be near the alignment cap.
• After the very early game, penalties don't matter unless you do something really bad or get a large amount of penalties all at once (e.g. knight attacking a sleeping throne room, being a caitiff for not waking up the enemies)

Balance considerations

• It's supposed to be an indicator of how well you are adhering to codes of conduct.

Idea #1: (Fourk implementation) Gains from killing monsters are removed completely.

• Drawback is that you have to get fairly inventive to raise your record enough to be allowed into your quest, sacrifice unicorns, etc.

Idea #2: Remove the gains from killing monsters, and slowly gain alignment points for turns spent while your prayer timeout and god anger are zero.

• Kind of like "good is the absence of evil". You don't need to go out of your way and help monsters out of traps, kill enemy unicorns, and things like that for your god to be satisfied with your good conduct.
• Your god does not get any more impressed with your progress while prayer is still timing out or they are angry at you, however.
• Maybe the gain should be one point every rnz(500) turns.

Idea #3: Where you would previously gain positive alignment from killing a monster, you now get only +1, and only a small fraction of the time.

Wishes

The game dumps 5-7 guaranteed wishes on the player at the Castle, and other wands provide just as many.

Balance issues

• This is usually enough wishes to fill out an ascension kit with one or two to spare for magic markers.
• Compounds the Gehennom items being useless problem.
• If the player happens to find a random wand of wishing, they get another 5-7 wishes, which basically guarantees a win for an intermediate or experienced player.
• Early single wishes are almost always GDSM or SDSM, with maybe a quest artifact if the player has enough HP.

Idea #1: (Fourk implementation) Provide scrolls of wishing or magic lamps at various points in Gehennom. Of course, a scroll is good for only one wish. These scrolls are un-polypileable, take more than 99 marker charges to write, but they might be an extremely rare naturally generated find.

Idea #2: Make wands of wishing not rechargeable. Fewer guaranteed wishes in the game means that the player has to think harder about what to wish for.

Idea #3: If another wish source like scrolls is provided, remove wands of wishing entirely.

Idea #4: Spread the wishes out around the late game.

• Easiest way to set this up is to make wands of wishing guarantee exactly one wish (always turning to dust after granting one), and then scatter multiple ones around.
• This can already sort of be done with magic lamps, but that is less reliable.
• Good places to scatter wishes include demon lairs, the fake Wizard's Tower, the real Wizard's Tower (maybe with the Wizard himself), Vlad's Tower, and keep one or two at the Castle.

Conflict

Conflict affects nearly everything in line of sight, and can be incredibly powerful, especially on levels mobbed with monsters.

Idea #1: make the ring of conflict a chargeable item.

• Monsters are affected in a radius whose size depends on the charge.
• The radius for the Sceptre of Might or other conflict-generating items isn't defined by this. Should it be constant?

Idea #2: conflict works in a fixed size radius and line of sight.

Idea #3: conflict works as usual, but not in the End Game.

Idea #4: conflict works as usual, but doesn't affect @ or A at all.

• This specifically targets the Astral Plane, and also makes Ludios and the Castle more difficult since the soldiers won't respect conflict
• Can make it not affect priests specifically if conflict should work in those places.
• FIQ suggested that it not affect the set of monsters that doesn't respect Elbereth, which would include minotaurs and elves.

Spells

The level of a spell doesn't correspond very well to its utility. This, in particular, is incomplete and is going to end up being a huge section or possibly even its own page.

Balance issues

• Charm monster is very powerful for a level 3 spell.
• Dig is not powerful or useful enough to merit its level 5 designation.
• Invisibility is almost never used because by the time one can cast level 4 escape spells one will have generally come across a stalker corpse or a wand of make invisible.
• Protection wears off way too fast to be any use in combat.
• High-level spellcasters, especially Wizards, have to worry about far fewer things and spend far fewer resources than players who can't cast those spells. Magic (magic speed, magic levitation, etc) is effectively a basic requirement for the endgame, and spells can provide spellcasters with a lot of leeway.

Idea #1: spell level is based on its utility, not its "power".

Idea #2: there is a limit to how much a certain spell can be cast in a certain period of time, probably related to its level. This is backed up by D&D: complex and powerful spells tend to require components and preparation. Including components is too complex for NetHack, of course, but preparation (in the form of a timeout) is easy enough.

Max HP gains

It is fairly easy to increase one's HP to a ridiculous level.

Balance issues

• Potions of full healing have no diminishing returns.
• Combining nurse dancing with potions of full healing (nurse dancing first, then quaff).

Not balance issues

• Nurse dancing; it has fairly well-balanced diminishing returns.

Idea #1: (Fourk implementation) Full healing and plain healing don't increase maxhp. Extra healing does, however, with diminishing returns.

Idea #2: Use a formula like max(1, oldmaxhpincrease - (maxhp / 50)) to enforce the diminishing returns

Idea #3: Involve a role-specific stat in any diminishing returns so that roles which tend towards higher max HP will still end up with higher max HP than more fragile roles after nurse dancing or potion quaffing.

HP regeneration

At high levels and with maximized ability scores, HP regeneration is insanely fast.

Balance issues

• Below level 10, Constitution has no effect on regeneration.
• From level 10-20, at high (18+) Constitution, you are fairly likely to regenerate (level-9) hit points every three turns. Getting at least one point means that at level 10, regardless of Constitution, your natural regeneration is equal to one-third of a ring of regeneration.
• If Constitution is 12 or below and XL is greater than 9, you get only one point every three turns, averaging 1/3 point per turn, regardless of your level.
• However, when Constitution flips from 12 to 13, regeneration goes completely bananas. At level 10, you will average 1/3 point per turn. At level 11, 25/39 points per turn. At 12, 36/39 points per turn (nearly the same as a ring of regeneration). At 13, 46/39 points per turn. At 14, 55/39 points per turn. At 15, 63/39 points per turn. At level 22 and above, this eventually stabilizes at 91/39 points per turn, but keep in mind this is only for Con 13. A late-game fighter with XL 22 and Con 18 will average 208/54 points per turn.
• A side effect is that the ring of regeneration is only good for early game characters who aren't at level 10 or can't get their Con to 13 yet.

Balance considerations

• Regeneration this powerful might be balanced for the heat of constant battle. But since the player is usually able to flee from the battle for a few turns, it doesn't take long for them to regain as much HP as they need.
• Or if a battle takes off a substantial chunk of HP, regeneration continues after the monster is dead and the player continues to walk around the level, and they will generally be well healed by their next battle.
• As long as HP regen exists at all, sitting in a locked closet, or similar uneventful area, to regenerate to full health is probably going to be a valid strategy.

Unicorn horns

Most status afflictions are too easily cured with a unicorn horn.

Balance issues

Unicorn horns are a perfect, infinite source of fixing various afflictions.

• Attribute loss: they make potions and the spellbook of restore ability almost worthless.
• Confusion: the other cures are lizard corpses, prayer, and a blessed scroll of confuse monster; nobody ever actually uses these because the unicorn horn is superior and even if the player doesn't have one, waiting for the confusion to time out is less wasteful of resources.
• Blindness: there are more secondary cures, but few are ever used except possibly carrots.
• Sickness: the other cures cost either physical resources, prayer, or power. The spell of cure sickness is probably better than a unicorn horn for those who have it.
• Hallucination: the other cures are potions or prayer, but waiting it out might still be preferable to spending these other resources.
• Stunning: no other cures at all, besides prayer, but again it might be preferable to wait it out.

Not balance issues

• The spell of restore ability is also a perfect, infinite source of fixing attribute loss, but this is its entire purpose, so it works as intended.

Idea #1: Unicorn horns are not infinite (they have charges, maybe?). Spending them frivolously to regain attributes might be what amounts to a waste of valuable resources.

Idea #2: (GruntHack implementation) Unicorns themselves are very rare.

• The player has to either get lucky and encounter a crossaligned one early, or reverse genocide them. Neither is particularly likely to happen until the Castle.
• After the player obtains a horn, attributes can be fixed just as infinitely as in vanilla, meaning that this is still a problem in the late game.
• Unicorns should probably be rarer than they are already. Yes I know their frequency is already 1, but they still pop up an awful lot.

Idea #3: Unicorn horns simply don't fix [attribute loss|blindness|sickness|...], and players have to look to other sources for fixing them.

Idea #4: Attribute loss from abuse is a permanent loss.

• Restore ability does not work on abused attributes at all.
• It has to be gained back the hard way (exercise, potions of gain ability, or some other thing that increases that specific attribute).

Spellcasting barrier of entry

It's quite dangerous to read unknown spellbooks. Characters that could get at least some benefit from spells (i.e. some fighter roles like Rangers, Rogues and Samurai) generally have little incentive to pick up books, waste scrolls of identify on them, or lug them to a shop to price-ID. The one class that gets warnings, Wizard, is the class that has to pick up and read as many spellbooks as possible anyway.

Idea #1: When generating a spellbook on level N, the level of the spellbook must be at most N/2, rounded up.

• More dangerous (and unusable) books don't start appearing till later.
• Reading a book found on a lower level is far less likely to get you killed.
• Perhaps this could be N/3, though that seems a bit off.

Idea #2: give the spellbook read warning to other classes, though in a much less useful form.

• For instance, if you have less than a 30% chance of reading it safely.
• Should have a different message than "This spellbook is difficult to comprehend".

Quest experience level

Many players are ready to enter the quest at experience level 10, 11 or 12, but 14 is the hard lower limit.

Balance issues

• It's annoying to scour the dungeon for foocubi, wraiths, or potions of gain level.
• It's even more tedious to grind for them.
• Pacifists have an especially hard time of it.

Idea #1: Allow players into the quest at XL 10. If they're not prepared, that's too bad for them.

Idea #2: If the player is below the requisite level, the quest leader will summon a powerful monster with the player's permission. If the player can dispatch the monster, the quest leader will consider them ready.

Invisibility

Invisibility and see invisible's availability as intrinsics means that the problems of not having them eventually vanish completely.

Balance issues

• Permanent intrinsic invisibility is easy to come by (stalker corpse, wand, blessed potion), after which the player doesn't need any of the temporary sources again (spell, ring, cloak, non-blessed potion).
• Vlad's spellbook of invisibility is never useful because the player almost certainly has the intrinsic by then.
• Most stuff in the late game that you care about hitting you or finding your position is going to hit you anyway, so invisibility doesn't help much by then.
• Permanent intrinsic see invisible is also easy to find by the midgame (stalker corpse and blessed potion mostly, but also thrones). After that point, all the monsters zapping themselves with make invisible or chugging invisibility potions are wasting their resources, since invisible enemies are no longer a factor.

Idea #1: remove the ability to get these as intrinsics. Getting them as extrinsics (rings, cloak, perhaps a new item modeled after SporkHack's helm of sight which grants extrinsic see invisible) is the only way to get them permanently. A blessed potion or stalker corpse is still good for a few hundred turns. A wand of make invisible probably less than that due to their abundance, and the spell even less because it's fully reusable.

Early game unavoidable death

Several things are too potent against early game players and will probably kill them regardless of the player's skill.

• Poison instadeath
• Gnome with the powerful attack wand
• Chest traps
• A trapdoor on DL1 can theoretically make a player fall as far as the Castle

Idea #1: Allow the player to guarantee there is no trap on a door or chest.

• The #untrap command takes several turns to complete, probably based on Dex, Int, and XL.
• If there are no traps, it finishes with "You are sure there are no traps on the chest/door".
• If there is a trap, it is discovered with 100% chance, perhaps in fewer turns than that upper limit, though the player might still fail to disarm it.

Idea #2: Limit the number of floors the player can fall, to 4 or 5.

Idea #3: Nerf poison instadeath so it's no longer an instadeath. Variant solutions include permanent Str and Con damage, permanent maxhp damage

End Game difficulty

Many of the endgame challenges can be tanked, slain, or otherwise dealt with pretty easily.

Balance issues

• In particular, Moloch's Sanctum and the Astral Plane are anticlimactic - the priests and Riders are decent enemies, but the undead and demons in the Sanctum are not threatening in the least, the High Priest of Moloch is easily slain with a death ray (as are Famine and Pestilence, come to think of it), the Angels on Astral usually do 1d8 long sword damage per hit and their magic missiles are ineffective, player monsters are only scary if they have Vorpal Blade, and Team Ant is everywhere.
• Blessed-genociding ; makes the Plane of Water even more boring than it already is, with a scant few water elementals and snakes around.
• The Wizard of Yendor will go down in one shot from a death ray, except in the rare case he manages to get in an extra turn and steal your magic resistance quest artifact.
• The Elemental Planes' main challenge is to locate the portal and get there while not being bogged down by enemies, which is not much of a final Test. Actually, not getting bogged down is a pretty good summation of the whole ascension run strategy.

Idea #1: get rid of summon insects in the end game.

• It's a fine clerical spell in its own right, but it just does not work when there are large crowds of priests.
• I favor the priest AI simply not selecting this spell while in the Sanctum or on the Astral Plane.

Idea #2: buff the endgame "bosses".

• The High Priest of Moloch should have more powerful attacks than the standard mace + clerical spell. He should also wear a shield of reflection instead of his small shield, so the player's death rays don't work, and be shock resistant, so Moloch's divine lightning doesn't hurt him. Jonadab proposed he get a special attack that turns floor in the vicinity into lava and wear fireproof water walking boots.
• The Wizard of Yendor should have player-style magic resistance at least some of the time. In an earlier proposal, I suggested that every time he reincarnates, he should have a chance to generate with an amulet of reflection, a cloak of magic resistance, or a shield of reflection. If not given the amulet of reflection, he will have a second independent small chance to get an amulet of life saving.
• A number of things could be done with the Riders, but most important is that they can't be killed by death rays.
• Vlad should not be such a wimp. This goes without saying.

Idea #3: (jonadab's dungeon overhaul proposal) Monsters generated in the Planes do not respect genocide, and you cannot re-genocide them when you get there.

• The flavor behind this is that genocide affects only the mortal world, which the player has now passed out of.

Scroll of scare monster

It works on more or less all @ in 3.6. It was buffed as a compensation for making Elbereth far less useful than it was in 3.4.3.

Balance issues

• All Quest nemeses can be trivially killed by standing on the upstairs on a scroll of scare monster and punching them to death. Even Master Kaen.
• Basically any non-lategame boss threat can be dealt with this way, unless that threat happens to have ranged fire attacks.
• Once you have one and a supply of junk potions and holy water, it can last for as long as you have holy water. Bonus points if you picked it up blessed or out of a chest, so its internal flag isn't set.

Balance considerations

• Scare monster is the only scroll illiterates have any use for.

Idea #1: make the scaring effect resistable (based on monster level, like any resistable effect), and ineffective on all covetous monsters including quest nemeses.

Idea #2: it's no longer guaranteed that you can pick up, bless, drop, pick up, bless, drop ad infinitum.

Wands

Wands don't scale as the game progresses.

Balance issues

• Wands can be too powerful against the player in the early game. Intelligent monsters that pick up or start with a wand of fire, lightning, cold, death, sleep, and possibly even magic missile or striking, can kill the player with ease, sometimes even one-shotting them with no warning.
• Wands are ineffective against the player in the late game. Rays don't work at all because the player will have reflection, and the only beam wand monsters will use, force bolt, doesn't work because the player will have magic resistance (and enough HP not to care anyway).
• Many wands become useless to the player as the game progresses as well: speed monster is only useful if you happen to get hit by a cockatrice and lose your speed, magic missile does a piddling 2d6 damage, make invisible isn't needed after one zap. Other wands like fire, cold, and striking aren't really worth zapping as attacks any more and are only worth keeping around for their utility purposes.

Balance considerations

• Wands are a very user-agnostic form of magic. A system where a wand zap has the same effects regardless of who's using it makes for very good flavor. Ultimately this will probably have to be conceded though, because balancing them so that they scale is more important.

Idea #1: The Wands Balance Patch, of course.

Object weights

Lots of stuff doesn't really have a sensical weight. Potions are an exception to this, as their weight is working as intended (according to jonadab).

Skill slots

Characters get away with training up only a few skills, which underutilizes the skill slot system. Some roles tend to fall into combat styles that are numerically better but don't make that much sense flavor-wise.

Balance issues

• Several roles are able to enhance everything they need by XL 14, or at most 17.
• Roles like this include valkyries and samurai, where enhancing long sword skill and twoweapon skill is all the player needs.
• This is a contributing factor to the problem that characters don't have much incentive to level up more.
• Other roles, which in general means anyone that wants to cast spells, has to push that limit and perhaps go into the upper 20s.
• Lots of skill types go completely unused; this is partly an artifact weapon problem, but also because several skill types only apply to one or two weapons.
• Wizards are often built around throwing lots of daggers or darts instead of casting spells well into the midgame and late game, since daggers and darts can be enhanced to Expert and it costs no energy or resources to throw a dagger.
• Chris's analysis:
  • "As with to-hit/AC, nethack's skill system is strongly based on 1st and 2nd edition D&D, specifically the weapon proficiency system. However, nethack makes several changes, the net effect of which is to throw off the balance of that system. D&D characters received 1-4 weapon-proficiency-points (mages get 1, fighters get 4, others fall in between) which they must spend during character creation (and subject to rules forcing the player to spread the points out among multiple weapon types). After creation, characters receive additional points at a rate of 1-per-3-to-6-levels (mages get them slowest, fighters get them fastest), which they can spend according to rules that more-or-less resemble nethack's. The net effect, then, is that skill points are in much shorter supply in D&D than they are in nethack."
  • "In D&D most characters want both a ranged attack and a melee attack, so they need to spend proficiency slots accordingly. In nethack most characters are happy to pick one or the other."
  • "In D&D some enemies resist certain types of damage (ie, bludgeoning or slashing damage). It can therefore pay to be proficient in multiple types of weapons, so that you can use whatever is most effective against any given foe (though in most cases I don't think this mechanic is heavily used). In nethack damage types are listed, but totally unused."

Balance considerations

• The player is supposed to have to make strategic choices about which skills to improve, and if need be gain extra levels if they decide they need more slots.

Skills that have narrow usefulness or are under-used

• Boomerang: only boomerangs use it. This is fine because boomerangs are very different from any other weapon.
• Trident: only tridents use it.
• Knife: There are five weapons that use it, but the scalpel, knife, stiletto, and worm tooth are poor weapons. Crysknives see some use, but aren't reliable weapons even when fixed. Some variants merge knife and dagger skill on the basis that stabbing with or throwing a single-bladed weapon isn't much different from a double-bladed weapon.
• Axe: only axes and battle-axes use it, though Cleaver's existence makes it not so bad. Plain axes are poor except to train skill as a Barbarian. I'm not sure why a plain axe isn't a weapon-tool, since it seems to be used more often to chop down doors and trees than to fight.
• Short sword: there are four slightly numerically different variants, but they are all simply inferior to long swords, even for Rogues.
• Scimitar: only scimitars use it and no class can advance it to Expert. In real life, fighting with a scimitar is quite different from fighting with other types of swords, but in NetHack the gameplay is the same as other swords, but inferior.
• Saber: only silver sabers use it. However, because of Grayswandir, many roles find it worth taking a lot of effort to unrestrict and train its skill.
• Two-handed sword: considered to be dangerous around spellcasting monsters because of curse items; more importantly, twoweaponing +7 weapons almost always beats out the extra damage of the two-handed sword.
• Polearms: they are annoying to use as a primary weapon interface-wise, since pounding is the intended way to use one and are useless at melee range. Additionally, no character can train them past Skilled.
• Pick-axe: mainly suffers from a lack of artifacts, the fact that the dwarvish mattock is two-handed, and that archeologists almost always want Grayswandir instead.
• Morning star: only morning stars use it, and though it does have an artifact, that artifact is Trollsbane.
• Hammer: only war hammers use it, and war hammers have bad base damage. Characters with Mjollnir often do enhance it, though.
• Club: the club and the aklys are both very poor, and Cavemen try to ditch it for a staff or mace as soon as possible.
• Flail: it's not a great weapon, has no artifacts, and only the flail and grappling hook use it. The grappling hook is not much use as a weapon, and it's generally too obscure, rare, and hard to use to be worth training flail skill for.
• Sling: only slings use it, and its damage is poor. The Sling article outright says "Improving sling skill is generally a waste of skill slots" and goes on to explain how NetHack slings don't measure up to real-world slings.
• Whip: only bullwhips use it, and their damage is poor. Like the axe, it seems to be more useful as a tool (escaping pits, picking items off the ground, disarming enemies) than as a weapon.
• Crossbows: they are heavier than a normal bow, and bolts are considerably harder to come by than arrows (especially if you're a gnome; the most plentiful source of bolts is other gnomes, which you would have to kill somehow).

Idea #1: (a couple of variants' implementation) Add skills for armor and shield use.

• This gives players, particularly non-spellcasters, an extra thing to choose to invest skill slots in. Wizards are probably restricted or can get only to Basic in armor skill, and roles in between have some more options for how they can build out their character.
• Assuming a system that splits evasiveness and damage reduction into two different stats, armor skill reduces the effective weight of armor being worn and reduces or eliminates evasiveness penalties from that armor. Shield skill grants extra evasiveness at Basic and above (using shield to deflect attacks), and allows you to perform a shield bash by applying your shield in a direction at Skilled and above.
• Included in my combat system overhaul proposal, with much more detail.

Idea #2: adjust weapon skill categories.

• Tridents use polearm skill, as they are effectively a polearm, and trident skill is removed. (The Trident page even notes that it's "quite curious" that it doesn't use polearm skill.)
• Merge knife and dagger skill, for the reason above; they don't really fight differently.
• Merge scimitar and saber skill.
• Merge morning star and mace skill, perhaps even merge in hammer skill. All of these involve swinging a heavy object on the end of a stick at the enemy as hard as possible, and there are not enough distinctions as far as gameplay goes.

Idea #3: reduce Wizard's skill caps in dagger and dart to Skilled.

• With the 3.6 nerf to wizard multishot, this means that wizards will not be able to multishot anything.
• Magicbane remains a useful weapon, but spells are more powerful.

Idea #4: allow Basic polearm users to hit where Skilled users can now (knight's move away), and allow Skilled polearm users to hit where Expert users can now (two spaces diagonally).

Idea #5: make bullwhips and possibly axes weapon-tools.

Idea #6: add some special sling ammunition that weighs 1 and has decent damage and good multishot.

• Cavemen will start with this rather than a bunch of flint stones.
• Alternatively, make flint stones into this ammunition, reducing their weight to 1, but there has to be some way for slingers to make more of them, because flint stones are currently rare and not usually found in stacks. Perhaps there can be some method of breaking rocks to make flint stones (or other sling ammunition).

Idea #7: (NeroOneTrueKing suggestion) Make skills a tree, instead of an array, like advancement trees in other games. So getting a scimitar and training it to Basic isn't a waste of a skill slot if you plan to specialize in long sword: instead it trains a generic "sword" skill which can be trained only up to Basic, and once past Basic you can then leave the scimitar alone and train long sword up to Skilled and Expert. This suggestion was well received when proposed, but it would require some major architecture changes.

Midgame power spike

Over the course of the midgame, a good player's game tends to get a lot easier. Many of the challenges that are faced in the early game either disappear or are permanently overcome.

Balance issues

• The player manages to get or produce a holy water around this time. With it, they can dilute all their junk potions and mass-produce holy water. This lets them mass-bless all their scrolls of identify, enchant weapon and enchant armor, luckstone, and potions of gain ability. It also allows them to remove any pesky curses, either by blessing a scroll of remove curse or simply dipping things in holy water.
• Weapons get a lot more powerful because it's easier to enchant them.
• AC improves because it's easier to enchant armor.
• The chance of a blessed identify is high if the player can have held onto 3-5 scrolls.
• The player will either have found an altar and have a source of creating monsters or a unicorn and some gems from Mines' End; luck can now be maximized.
• Ability scores go up from blessed potions of gain ability. This can cause carrying capacity, to-hit, HP regen, energy regen (to a lesser extent), and spellcasting ability to skyrocket all at once.
• The player is likely, at least after finishing Sokoban, to have a ready reserve of food to use when there happens not to be any other food around.
• A bag of holding can probably be blessed, further increasing what the player can carry and reducing the need for stash management.
• The player is encouraged to save resources for this point so they can do all these things at once, instead of the progression happening gradually.

Not balance issues

• The Castle wand is another power spike, but see the Wishes section about that.

Idea #1: (NeroOneTrueKing suggestion) Cap the effective value of Luck to XL/3 whenever it's used in a die roll.

Potion of gain ability

Allows basically any endgame character to have racially maxed ability scores.

Balance issues

• Blessed gain ability is easy to come by; gain ability potions are frequent enough that the player should get several of them naturally, and they can additionally be alchemized up from full healing, which is also easy to mass-produce.
• Seven or eight blessed gain ability potions should bring just about any character's important stats up near or to their maximums.

Idea #1: remove gain ability as an alchemy target, and make them much rarer, comparable to potions of enlightenment or polymorph, or even rarer.

Idea #2: (some variants' implementation) Blessed gain ability doesn't increase each score by 1, instead it increases one score one or two points.

Junk scrolls and potions

As the game progresses, more scrolls and potions become useless.

Balance issues

• Many scrolls and potions are useless from the get-go (amnesia, punishment, booze, fruit juice).
• Some have only marginal uses (sickness, fire, destroy armor, stinking cloud).
• Some may get replaced by reusable sources, especially spells (restore ability, identify, remove curse).
• By the late game, only a couple types of potions and scrolls are needed (full healing, charging, gold detection).
• The abundance of more useless items ensures that there will always be a ready source of blank scrolls for writing and a ready supply of holy water.
• jonadab classified potions into three categories: ones that you alchemize to full healing or gain ability, ones that you blank for holy water, and ones that are rare and unimportant (and can usually be blanked too).

Idea #1: (GruntHack implementation) Blank scrolls can't be polypiled; if you blank all the junk scrolls you won't have enough marker charges to do anything useful with them. Potions are still problematic because a supply of holy water is useful even if it can't be polypiled.

Idea #2: Make potions more powerful and situationally useful - a potion of restore ability might not be worth diluting if a unicorn horn can't fix your ability scores, for instance.

Idea #3: Give everything at least some use, even if it's marginal, but preferably more than marginal if possible. Some examples are a scroll of fire buff I proposed that lets the user cast a blast of fire at range (included in 3.6.1), and the scroll of amnesia draining skill practice points so skill slots can be re-shuffled.

Mines' End loot

The available loot in Mines' End isn't very even across its variants.

Balance issues

• The Mimic of the Mines, in addition to having an uninteresting level, has basically no special loot. It guarantees two random tools, and that's about it.
• The Catacombs' guaranteed spellbooks make the game very good for a spellcaster and are fairly useless for a non-spellcaster.

Not balance issues

• It seems to be standard for all Mines' End variants to provide an uncursed luckstone and a good quantity of valuable gems and worthless glass. All level variants have this, and most if not all NetHack variants keep the luckstone in other level variants they add.

Idea #1: add some more interesting loot to the Mimic of the Mines.

Idea #2: (several variants' implementation) Decide that the Mimic of the Mines is not interesting enough to warrant a level, and replace it with some other level (e.g. Fourk replaces it with the Orc Temple).

Ammo breakage rates

The rate of ammo breakage (or "mulching") is detrimental to ranged combat, and leads to ranged characters favoring melee weapons or daggerstorming instead of using projectiles.

Balance issues

• Blessing is far too important. Unless the player's Luck is terrible, blessed projectiles have a substantially lower chance to break.
• Luck is far too important. The amount of time a player goes before they are able to maximize or at least permanently increase Luck is long, and this compels players who are intended to use breakable weapons early to use other, sometimes inferior, weapons until they can find a luckstone. On the other end, high Luck with blessed projectiles essentially ensures that nothing will ever break.
• Skill is not relevant at all.
• A non-blessed +0 projectile has a 2/3 chance of breaking, a +1 has a 1/2 chance of breaking, and a +2 or above has a 1/4 chance of breaking. Blessing it at low Luck reduces the rates to 1/2, 3/8, and 3/16 respectively. These rates are simply too high.

Balance considerations

• The Tourist early game should probably not play very differently. The starting darts are expected to run out eventually, wherein the Tourist should have found either more darts or something better to use by then. However, the Tourist should probably not have to worry about using the darts, to the extent of having to attack things bare-handed because the darts are too precious.

Idea #1: (Fourk implementation) Skill level and skill cap in the relevant projectile's skill factors into breakage rates. Enchantment is more important, and Luck and blessing do not matter at all. Specifically, restriction in a skill makes projectiles very likely to break unless saved by positive enchantment, and decent skill makes projectiles unlikely to break unless negatively enchanted. Additionally, hard stones are less likely to break and cursed objects are more likely.

Untrapping

It is not usually a very useful action to take.

Balance issues

• There's never a guarantee that nothing is there - untrapping is more or less a game of confidence that there isn't a trap there, where you have to decide how many times is worth checking for traps.
• Even if you find a trap on a map space or door, it is usually better to ignore it and avoid it (or blast the door down from a distance), making untrapping useless in this situation.
• With decent Luck, the bad container trap effects are very unlikely, and with near-maximum Luck you will very rarely get anything besides hallucinogenic gas, so you can largely ignore them. The chest in the Castle containing the wand of wishing is not supposed to be trapped, and is considered a bug on versions where it may be trapped.
• Bad effects of stumbling into an undetected floor or door trap are not very bad: about the worst that can happen is an unexpected polymorph, which isn't common and can be shielded. Note that making trap effects worse is not a good solution - it merely makes players have to spend more time searching for the traps and assuming there are none. In GruntHack, for example, the danger of trapped doors burning your inventory in a fire blast simply means you eventually have to try #untrapping every door.
• There isn't usually any reason to obtain a beartrap or a land mine.
• There is also some difference between ease of use for numpad and vikeys users, since a vikeys user must press 5 keys (shift-3-u-enter-direction) to untrap, whereas a numpad user must press only 2 (u-direction). This is only a problem because frequently, the player should untrap repeatedly.

Death magic

This refers specifically to instadeath magic, not Death's special HP and maxHP draining attack when you have magic resistance. The presence of an instant, unavoidable instadeath attack in the game is a serious design problem, and influences several other design problems and balance issues.

Balance issues

• If the game offers no protection against instadeath magic, deaths due to it will all be instances of bad RNG and fairly classed as YAADs, in which the player did nothing wrong except be unlucky enough to get killed.
• If the game does offer protection against instadeath magic (as it currently does with magic resistance), then obtaining protection against it is paramount and must be done as early as possible, because random monsters can spawn with a wand of death, and the number of death sources in the endgame mean that it is a necessary component of an ascension kit.
• Changing magic resistance not to protect against death magic (as many want to do) merely forces optimal play to be wearing an amulet of life saving or polyself to undead.
• All this is because it is the most blatant all-or-nothing effect in the game; it forces the game to decide whether or not it is over, immediately, without any player intervention. This is almost the only such effect in the game (poison instadeath being the other, and maybe random enemies ambushing you with Vorpal Blade being a third); all other instadeaths are either delayed, preventable, or due to intention or stupidity on the player's part.

Idea #1: (FIQ) remove all sources of death magic from the game, including the monster spell touch of death and the player spell finger of death, and make wands of death not generate, except with Orcus who still gets his wand as normal.

• Requires that Orcus' death zaps won't obey reflection; otherwise reflection is a perfect defense.
• Requires Orcus to be nerfed significantly in speed or some other way, and not be able to move and attack on the same turn, so he doesn't kill the player.

Idea #2: (FIQ) Remove death magic completely and buff disintegration and add new sources of disintegration in the late game, as a replacement.

Idea #3: Remove instadeath magic, and make "death magic" hit the player's HP, maxHP, and possibly attributes fairly hard.

Idea #4: Make the only sources of death magic happen after a certain point in the game - such as only once you reach Gehennom.

• Does not help at all with the ascension kit issues, since a kit will still want to guard against late game sources.

Average threat level per monster

ais523 brought this up. The number of difficulty-appropriate monsters attacking you simultaneously should generally take the same number of turns to kill you throughout the game. However, this is not the case as it currently stands: a low-level character can get doomed by being surrounded even by weak monsters like jackals and newts, while an endgame character can wade through mobs of hostile priests and angels in the End Game for many turns without being at serious risk.