Player ghosts are still a bad game mechanic

[Hellmonk]

Due to recent events on the CBRO server, I feel compelled to make yet another post about player ghosts. Player ghosts are an extreme case of poor design and are possibly the worst feature in DCSS. There's so much wrong with ghosts that it's hard to know where to start, but I think these are the four biggest problems with them. None of these arguments are new, but I feel like they should be brought up again since no major changes to player ghosts appear to be forthcoming.

• Player ghosts are essentially impossible to balance. They have a massive range of possible attributes based on the character that died (including HP, AC, EV, spell selection, melee damage, melee weapon brand, movement speed, resistances) and zero checks to ensure that they are a reasonable spawn for their dungeon level. Hence you can get a D:6 ghost with glaciate or you can get a spriggan ghost with no spells and 1 melee damage. A lot of player attributes don't translate well to monsters, which exacerbates this problem. I am thinking specifically of AC, where 20 is endgame material for monsters and achievable in lair for players. Then there's stuff like monster dazzling spray, which works different from the player spell for no apparent reason other than to make conjurer and wizard ghosts abnormally dangerous. Other monsters can get pushed around or have stats tweaked to make them an appropriate threat; ghosts cannot because of the way they're generated.
• The vast majority of ghosts are not interesting. People complain about bullshit deaths from eg. D:3 CeCK ghosts, but to me that is infinitely preferable to the typical ghost. Dangerous early ghosts can present interesting scenarios. Instead, most online earlygame ghosts are speed 10 melee only monsters with too much HP and AC but that can't climb stairs, so you get to manually explore floors while stairdancing repeatedly. Others are weak enough that you can kill them, stairdancing once or twice if the fight goes poorly. Lategame ghosts are almost all trivial bags of experience. I have seen perhaps two semi-interesting ghosts in all of my online games, but I've gotten dozens of annoying ones. Regular monsters could replicate the couple scenarios where a ghost is actually interesting, except they could do it reliably and their balance could be tweaked as needed.
• Ghosts encourage checking morgues. In addition to examining every ghost, you should always check the player's morgue. You get to see the exact AC and EV, which are not shown in game, so this is still worth doing even now that damage is displayed.
• Ghosts encourage metagaming. The number and danger level of ghosts depends on which server you play on, or the history of your DCSS install if you play offline. There are lots of ways that you can abuse this. You can run a shaftbot, or manually suicide formicids on an alt, before entering a new floor to reduce your chance of dealing with ghosts. You can also check how many players are dying and what combos they’re dying on before you start a game, and avoid hours when bots are playing to avoid getting their annoying ghosts. While actually griefing with a bot set up to run CeCK will get you banned, running a MiBe bot probably won’t and is almost as annoying. It is bad in principle that this kind of behavior could give someone an advantage.

The typical defense for player ghosts is that they provide interactivity. Assuming that this is desirable, it would be better to have a dungeon feature, like a gravestone, than a monster. Moving over it could give you information on the player that died: "Here lies dickylongcocking the poker, merfolk of okawaru". This would decouple interactivity from game balance. An alternative solution is to let players opt out of generating or receiving ghosts. This solution has been proposed and rejected in the past because it has an impact on balance, but ghosts already have an impact on balance because you can play on a different server if you want to get more/fewer/different ghosts. Naturally, complete ghost removal is an option as well. Any of these solutions would be preferable to the status quo. The current system of player ghost generation is not tenable.

[duvessa]

I know opting out of receiving ghosts has been specifically rejected, but has opting out of leaving ghosts been specifically rejected? You can already reliably avoid leaving ghosts, no matter how bad you are, by simply playing undead. It does not seem like an enormous leap to allow players to opt out of it while playing any species they want. If the argument against it is that it would lead to servers having too few ghosts, then at the very least you should make quitting leave a ghost, since many players (myself included) will quit when in a sufficiently deadly situation, in order to avoid leaving a ghost.

Also, since you didn't mention it:

Whenever the topic of player ghosts come up, there is the argument advanced that player ghosts create interactivity between players in this single-player game. This is somehow advanced as an argument as if that interactivity is a good thing, instead of one of the main reasons to pick a single-player game like Crawl in the first place. Players who use racist slurs and their character name are really only one snappy example of ways other players can intrude on my experience, but since my preferences for not interacting with idiots are not a mechanical issue with ghost generation the problem isn't really calling out for a mechanical solution.

(source)

Also, it's kind of annoying how many terrible monster spells have been added, with the excuse of "this terrible monster spell is just for player ghosts", and then later given to pan lords and even normal monsters (Shatter being the most visible example)

[chequers]

I wish undead left ghosts. It's a great example of flavour over mechanics.

[bel]

I can quote another post from the same thread:

The trouble with ghosts is that literally every single aspect of their design is bad.

Another way of coming to the same conclusion is to consider the following hypothetical. Imagine ghosts didn't exist, and someone opened a GDD thread to propose them. Now imagine the reaction.

[Shard1697]

Was draconian ghost breath fixed? It's been a while since I saw people talk about it and I don't remember.

[duvessa]

I wish undead left ghosts. It's a great example of flavour over mechanics.

Given that "flavour over mechanics" is the primary criticism of player ghosts in the first place...

[kuniqs]

Remove spells from player ghosts, cut their damage in half and give no XP for killing them.

[Tedronai]

Ghosts being completely removed from crawl would be the best thing to happen since the removal of item destruction!

If ghosts are absolutely to stay, I would prefer if they were generated only on the basis of the current depth. Meaning the ghost would keep the name of the player who died, but other than that it would just generate a level appropriate creature (not taking anything from the actual morgue into account).

[mattlistener]

Has the case really been made here that ghosts are structurally *impossible* to balance? To address the top-most cases above you could put a floor on their melee damage and a ceiling on their allowed spell levels, based on ghost level and/or dungeon level. You could scale their AC more appropriately. Those would be examples of improving the game balance of ghosts, yes?

Other arguments against them might prevail, but this one?

[mattlistener]

Assuming a player has memorized the entire Crawl Monster's Compendium, the only enemies that they still need to Xv and study upon encountering them are player ghosts, Pandemonium Lords, and... anything else?

This seems like a form of variablity in game experience that would be a shame to cut rather than improve.

[Shard1697]

Are you saying it's good for a player to want to x->v every ghost and check the relevant morgue, because it's different from what they do normally?

[VeryAngryFelid]

Assuming a player has memorized the entire Crawl Monster's Compendium, the only enemies that they still need to Xv and study upon encountering them are player ghosts, Pandemonium Lords, and... anything else?

This seems like a form of variablity in game experience that would be a shame to cut rather than improve.

If you need variability, it's better to implement it via normal monsters like black mamba or hydra. It is almost impossible that random monsters will be more interesting or balanced than carefully crafted ones. To take your approach to extreme, would you like it if all monsters had tile of a rat? Abominations had random stats and I am glad they were removed, it was awful to use xv on them.

[RBrandon]

Interestingly (and perhaps obviously) ghosts are bad design because they are not designed at all. Or at least are not designed in a way to provide a meaningful challenge since players are "designing" them.

[galbolle]

Ghosts should be more like folklore ghosts: they should be immaterial and unkillable. They would haunt the player (independently of their former character), inflicting hexes on the player until they have been appeased. They are appeased by revenge, id est, the player killing the monster which killed their former self. That monster should be distinguished, and the ghost could haunt the player more intensely if they are going away from their killer.

[ion_frigate]

Ghosts should be more like folklore ghosts: they should be immaterial and unkillable. They would haunt the player (independently of their former character), inflicting hexes on the player until they have been appeased. They are appeased by revenge, id est, the player killing the monster which killed their former self. That monster should be distinguished, and the ghost could haunt the player more intensely if they are going away from their killer.

This is basically like a minor version of hell effects, that goes away upon killing a specific monster. It's a neat idea, but I can see several potential issues:

• If the original player got killed by something really nasty, it's unbalanced. Examples include players on the orbrun, or players who decided to sit around and trip the OOD police. However, this is still better than ghosts, since we have ways to measure the difficulty of a monster (HD, experience, etc) and weed out the ones that don't fit in the current level.
• The target monster should be a durable summon, with unobtainable equipment. I know it's minor compared to NetHack, but you don't want any shades of bones-stuffing. Monster weapons (particularly ranged ones) could theoretically be desirable, and there should be no way to introduce a good item to a subsequent character from a previous one. Satisfying the ghost's quest could give a level-dependent amount of XP.
• The monster's equipment should be rerolled, and the player given a telepathic link with the monster, so that there's absolutely no reason to check the morgue of the dead character.

[Siegurt]

x-v should bring up the morgue file of the character that created the ghost.

[n1000]

If ghosts are to be preserved because of some desire for interactivity or continuity it has been suggested to turn ghosts into generic monsters (similar to Tedronai's point).

Rename a couple of existing uniques so they take on a recently dead character's name + identifying title so instead of fighting "Rupert" you'd fight "n1000 the warrior" or whatever. If you want to get really fancy, you could find a character that roughly fits the monster's specification (e.g., Rupert would be given the name of a character that died between XL:n-m and had a weapon skill higher than any magic skill, Boris--versa vice, etc.)

This way players would not be forced to play different versions of crawl because of their server choice. The same monsters would be generated, but you get to see someone's name.

PS: I don't think I've seen any players quite degenerate enough to do this, but it's beneficial to play on an empty server/off-peak hours and leave a bunch of XP potions for yourself in the form of pathetic ghosts. Speedrunners effectively do this by dying a lot early on with weak characters in quick succession.

[mibe420]

Even though speed 10 melee ghosts are not dangerous on their own, having a 40 damage nearly invulnerable monster wandering on the floor is not a good thing. While you can stairdance it and ignore it, there is always some chance it hits a tele trap or wanders while you are luring some other monster. So, if possible you should prevent the ghost from spawning right? You can easily do this by running 'dummy' games to spawn all the ghosts before you play the real game you want to win. Because of the way that ghosts work in DCSS (if a ghost spawns on a floor, and the player dies on another floor, the ghost is gone) you can bust all dungeon ghosts in a matter of minutes by shafting formicids. I'm not sure exactly how ghost spawns work, but IME running about 6 or 7 formicids every time before you descend to a new level resulted in meeting 0 ghosts.

You could take this a step further and play a dummy game all the way through lair, orc and portals. You would have to do this on a server or at a time where there are not many players dying. You can check with sequell which servers have the most noobs dying and at what times. While usually a ghost in lair isnt dangerous its still better to not have a komodo dragon with might and 25 AC (ie any minotaur ghost) or a monster who can do 40+ damage from range on the floor.

If you dont do this, and die to a tough ghost, its your fault.

Ive also thought about leaving naked deep elves wielding launchers strewn about dungeon as a source of free exp. If noone else is playing on the server at the time you could do this pretty reliably.

So, thats the biggest reason I dont like ghosts. There are ways to basically remove ghosts, but they are all tedious, metagaming, spoilery things that have nothing to do with the game itself. You can also just play offline, and delete all the ghosts yourself.

Theres also ways you could 'metagame' to get an advantage during a tournament. There is no explicit rule against unless you consider the vague "Please do not do anything that would give you an unfair competitive advantage over other players or clans. This includes stuff like scumming crash-on-demand bugs or running bots on your own account for speedrun points – just remember that the objective here is to have fun."

Is purposefully dying as a CeCK multiple times in the early D right when the top team is trying to continue thier streak an "unfair competitive advantage" or is it just using a game mechanic? How could you even determine the intent of the player losing CeCK's? If they are playing them manuallly (IE not with a bot) how could you even determine if they are trying to go for a real time speedrun or purposefully making the game harder for another player?

You also get 'N clan points per ghost kill after dying, where N is the dying character's XL minus 5. (No points are awarded if the dying character is XL 5 or lower.)'

What about a 'strategy' where teams have one player leave lots of deadly ghosts and the other 5 players purposefully die to it over and over? It could be viable at the end of the tournament, when streaks are broken, speedruns are done, and combogod points are worthless.

Or a clan full of bots (which is allowed in the rules) with the intent of scoring gkill points only. If they got even 1 point every 30 minutes, per bot, thats almost 5k clan points over the whole tournament.

Obviously in practice some or all of these things would get you banned, and maybe arent even practical. But they are possible with the way ghosts work.

It is also fairly obvious ghosts aren't going anywhere, but there are ways to change them to alleviate some of the issues I raised. I think the most elegant compromise is an rcfile option that when true means you dont leave a ghost. You can already do this yourself with ctrlq. Maybe you could just implement some random unique system, sort of like early game pan lords, and just name them after players. Or just make ghosts into derived undead so its no different than if the game had spawned a zombie, except its called dickstomper69 the kobold zombie.

[damerell]

I know opting out of receiving ghosts has been specifically rejected

I really suggest it should be reconsidered, with the same option also opting out of leaving ghosts. Yes, I know, it leads to players playing two different games. That's already happened, both with shaftrobin and given that (especially on low-population servers like CDO) what you meet depends on what people play.

I like ghosts. Right now with it not having moved to 0.20, CDO's population is tiny. When I meet a ghost there's a 50/50 chance the person who left it is someone I know IRL - quite possibly someone sitting in the room. It's fun. To know their exact stats I'd have to check the morgue? I'm a bad player, I don't know the exact stats of anything. I'd be sorry to see ghosts go or to be reduced to, aha, a ghost of their former selves by some of the proposals here. I'd like to see ghosts left on a level remain in play for the next unlucky delver.

An opt-out of ghosts is undesirable but it might be the least worst option. The alternative is either to force everyone to opt out by removing them, or to expend time and effort on something that in all probability will either have the ghost-haters hating the variability or the ghost-likers hating the blandness. An opt-out would be easy to implement [1] and leaves people on both sides of the issue reasonably satisfied.

[1] I appreciate I am writing this in the spirit of the Programmers' Credo: "We do what we do not because it is easy but because we thought it would be easy when we started."

[njvack]

Was draconian ghost breath fixed? It's been a while since I saw people talk about it and I don't remember.

Yes.

I tend to agree with KL's "every aspect of ghosts is bad" comment. Like, the concept of meeting the ghosts of players past is maybe kinda interesting, but every aspect of their implementation is problematic.

To save the concept, there could be a player ghost unique, with an algorithm to procgen some kind of "player ghost monster" but design it to be unpredictable, generally threatening, and yet not annoying. Maybe.

[VeryAngryFelid]

One way to keep inter-player interaction is to slightly buff monster who killed PC.
For example, you can meet Ogre the Sandman25 killer which is similar to standard ogre except it can have more HP or more EV or Sticky Flame/Throw Flame because RNG picks one of 3 three highest skills which Sandman25 had when killed by that Ogre: weapon/fighting lead to increase in HP, Dodging to increase in EV and Fire Magic leads to picking 1-2 fire magic spells.
If the ghost kills another PC, it can increase another stat (but no stat can be increased more than once or we will run into balance issue).
Of course uniques should not be improved this way.

[4Hooves2Appendages]

In my view, the idea of Player Ghosts is interesting. Although strong arguments can be made that a single-player game does not require this sort of interaction, or even that it detracts from the game.

The bigger problem, however, is with the implementation. The debate is quite polarised at the moment. On one side are defenders of the status quo, who like the idea, but end up defending the implimentation. On the other hand are a group of players who dislike the implementation so strongly that they advocate for removal of the entire feature.

What tends to get glossed over is the real possibility of change without removal. Like any other feature of the game it is possible to make general improvements to the form and function of every aspect of Player Ghosts. Some changes, like the removal of Draconian breath attacks, are testament to that possibility. That particular change shows a way forward.

Player Ghosts have several problematic aspects. These could be analysed separately and critically appraised within the context of the game. Then possible changes can be discussed and implemented. To focus the entire discussion on making a list of problems, and drawing the conclusion that only removal works, is, on the available evidence, not likely to be successful. It leads to unstructured discussion, full of quick fix suggestions.

My suggestion is to pick out parts of the Player Ghost implementation and discuss them in dedicated topics, trying to keep the debate constructive and focussed on those specific points.

The topics could be along these lines, in no particular order of priority:

• Offensive Ghost Names
  
• Ghost Defences
  Often too high in comparison to monsters at a similar stage of the game.
  
• Ghost Spells
  Some selections are extremely dangerous, out of proportion to other dangers at a similar stage of the game.
  
• Ghosts of Lategame characters in the early game
  Orb run ghosts seem especially unfair.
  
• Griefing
  Running into deliberately made dangerous ghosts doesn't feel intended.
  
• Meta-gaming Ghost generation
  Shaft-bots and deleting morgues, etc. again feels like an unintended consequence.
  
• Morgue Checking
  Gaining a big benefit from looking up outside information is clunky.

Some of the items on the list necessarily interact with each other, and the consequences of changes need to be considered carefully of course.

[damerell]

The bigger problem, however, is with the implementation. The debate is quite polarised at the moment. On one side are defenders of the status quo, who like the idea, but end up defending the implimentation.

I can't speak for any other defender of the status quo, but no - I am pretty content with the current implementation. Sure, you could meet an orbrun ghost and that would be pretty harsh, but that's obviously also extremely unlikely, and because I'm a bad player I have to get a bit lucky to win anyway so I'm not losing any sleep over dying to abominable luck.

However, I've offered up an implementation change, and here's another; until you attack them, ghosts are slow (slower than any non-Chei player) and don't attack, they just lumber after you complaining. It's up to the player if they want to awaken a sleeping giant.

[ion_frigate]

Allow me to suggest idea #47 to Make Player Ghosts Better (TM).

Just make them rare normal monsters that (potentially) get a name. Instead of directly creating bones, the game instead creates a (small) database of recent deaths, which includes the player name, XL, and highest skill. It then generates the ghost like any other monster, and tries to pick a matching name from the list. If none exists, it just uses a random name.

For example, on D:4 level creation, the game decides to generate a low-level mage ghost. It then searches the database for any characters that recently died with their highest skill being a magic skill, and who were relatively low level. If it finds a match, it uses their name. If not, it just picks a random one. The general format would be:

<Playername> the <powerlevel> <type> ghost

Alternately, if you want to get nice and flavorful, you could use words besides ghost to indicate the type (like spectre=mage, wraith=melee fighter, etc). That way, you're just introducing a few new generic monsters that can be correctly balanced and given sane defenses, you avoid horrible monster versions of player spells, griefing is impossible, meta-gaming is impossible, and morgue checking is pointless. The only problem you mentioned that isn't addressed is that of offensive ghost names - but that's a problem with public servers, not with ghosts. Those people need to be banned, since a racial slur is just as offensive in a high score list or IRC as it is on a ghost.

On the flipside, most of the flavor is preserved. While I'm sure some would disagree, I don't think that most people here would say some horrible loss of flavor has occurred because a generated ghost doesn't have *exactly* the same spell set as the player who died. The ghosts become more generic (since all ghosts of the same level and type are identical), but again, I don't think that's a huge problem. It's not like we care that all yaks are identical.

[VeryAngryFelid]

Idea #48 then. How ghosts were implemented in another single player roguelike - Demon. You find a grave and you have choice: investigate it or not. If you choose to investigate, you always find something useful but there is a high chance that the grave belonged to a ghost of another player who died on the same level. The ghost is harder than normal monsters and unlike DCSS you cannot stairdance it (Demon does not have upstairs) but still everyone is happy: some players can ignore the grave, some can gamble and hope it does not have a ghost, some are happy to fight the ghost for extra XP and fun.

[4Hooves2Appendages]

The bigger problem, however, is with the implementation. The debate is quite polarised at the moment. On one side are defenders of the status quo, who like the idea, but end up defending the implimentation.

I can't speak for any other defender of the status quo, but no - I am pretty content with the current implementation. Sure, you could meet an orbrun ghost and that would be pretty harsh, but that's obviously also extremely unlikely, and because I'm a bad player I have to get a bit lucky to win anyway so I'm not losing any sleep over dying to abominable luck.

You might be fine with auto-explore instant death Player Ghosts, but I suspect that it is a minority opinion. If I proposed that there is a 1% chance a player automatically loses the game every time they press 'o', I would get laughed at probably. If implemented, people would stop auto-exploring. If I proposed the chance is 0.0001%, it would still kill players if implemented. Not the exact number of course, but that's what you are arguing in favour of. (That is pretty much already the case, due to the way autoexplore is implemented and monsters like Rupert. Ghosts, fast FE's for example, increase that chance.)

There's a difference between "this feature doesn't bother me" and "this feature is good". In my view, the former statement isn't enough to argue for keeping something the same in the face of reasoned criticism.

However, I've offered up an implementation change, and here's another; until you attack them, ghosts are slow (slower than any non-Chei player) and don't attack, they just lumber after you complaining. It's up to the player if they want to awaken a sleeping giant.

If they don't attack, then they don't need to be slow, of course. That implementation would lead to tedium. You'd have to still examine every ghost you see, assess when you are likely able to kill it safely, then add a note, remember to look at the regularly, occasionally backtrack to pick up your XP reward. It fixes some of the issues with ghosts being too strong, but it introduces bad gameplay, so I'm not in favour.

[damerell]

Idea #48 then. How ghosts were implemented in another single player roguelike - Demon. You find a grave and you have choice: investigate it or not. If you choose to investigate, you always find something useful but there is a high chance that the grave belonged to a ghost of another player who died on the same level. The ghost is harder than normal monsters and unlike DCSS you cannot stairdance it (Demon does not have upstairs) but still everyone is happy: some players can ignore the grave, some can gamble and hope it does not have a ghost, some are happy to fight the ghost for extra XP and fun.

I like this, it's a bit like my slow passive ghosts.

You might be fine with auto-explore instant death Player Ghosts, but I suspect that it is a minority opinion. If I proposed that there is a 1% chance a player automatically loses the game every time they press 'o', I would get laughed at probably.

That's a truly hyperbolic exaggeration of the odds. As you say, sometimes the game just kills you - not very often, or people wouldn't streak. I've yet to see any evidence to suggest "auto-explore instant death Player Ghosts" form a significant proportion of those already very unlikely events, and the argument you make about ghosts could equally well be applied to anything that has a miniscule chance of killing you out of autoexplore - which is quite a few things, alas.

There's a difference between "this feature doesn't bother me" and "this feature is good". In my view, the former statement

I'm not making the former statement. I'm saying "I gain enjoyment from this feature". (And before you say that's not good enough to keep it, that's why my first proposal was to allow people to opt out of the feature altogether...)

If they don't attack, then they don't need to be slow, of course.

Yes, they do - otherwise you've got the thing stuck to you, it's easy to get trapped by it, you can't use AoE attacks...

[BobIsDead]

I'm going to side with the camp that opposes player ghosts. They just seem like a holdover from a time when the design philosophy of DCSS was less established, back when item destruction seemed like a good idea. I mean, if player ghosts didn't exist and someone suggested an idea for a unique that had some of the properties of player ghosts (can appear practically anywhere, can have a wide variety of skills and abilities, can vary wildly in terms of strength and weakness, etc.), it'd get kicked out of GDD in a heartbeat. Sure, some might argue we keep them around 'cos they've always been around, but part of Crawl's appeal is that it gets rid of old ideas when it realises they're unfun or encourage tedious play.

As for how to replace them, I think Hellmonk's suggestion about replacing them with gravestones is the right idea. It's a nice way of displaying failed dungeon delving attempts, which one can choose to pay attention to or completely ignore. Gravestones also open the possibility for fun with epitaphs ("Here lies Maud. Gone and best forgotten.", "Here lies Lava Orcs. Gone before their time.", etc.), which can add a bit of colour and variety to places that might otherwise be a little plain and generic.

[Factorialite]

Hide ghosts behind runed doors (could even be black doors that are "the entrance to a murder" or something flavorful )that don't have tele traps and include the monster that killed them, and only if the killing monster could be generated normally on that floor (not OOD or orbrun). You still get trolling people this way, but if someone is going to work hard to get a D:6 glaciator just to randomly troll one person down the road, that's something that they themselves have to live with.

[damerell]

Hide ghosts behind runed doors (could even be black doors that are "the entrance to a murder" or something flavorful )that don't have tele traps and include the monster that killed them, and only if the killing monster could be generated normally on that floor (not OOD or orbrun).

This is another version of the sleeping giant, and I think it's the best one so far. I think one could be slightly more generous about the monster difficulty (a bit OOD but not massively so). The door's flavourful description could give away the level of the deceased player in the same way that the current weakling...legendary ghost description does; this would also avoid the comedy case where someone losing to a panlord on the orb run happens to be finished off by an orc priest and the game goes "well, not OOD, fine, put it in".

I really like this. I don't really buy the objection to various proposals (which could be made to this one) that people will go back later and collect all the little bundles of XP and that's tiresome - if someone wants to do that they can go back to D:1 and wait until the OOD clock is getting serious, go to D:2, etc..

[Shtopit]

I like the vault idea. You could also have orbrun deaths spawn such ghosts. Just add a runed door and then "Here scours the restless spirit of an adventurer, killed while carrying the Orb of Zot towards the surface". Bonus points if the vault reproduces the place it was killed and is also inhabited by the monsters it last had in sight.

Once I found a hidden ghost on the river of the Ice Cave. It was fun, because I chose to dig him free. I think this is one of those situations in which giving the player a chance to roulette danger vs XP makes the game better.

[Rast]

How about this:

1. Remove "Getting loot from player ghosts (like bones in Nethack)" from the won't_do list.
2. If you die with at least one non-book non-weapon randart in inventory, you leave a ghost. If not, you don't.
3. When your ghost is killed it drops that randart (just one if you had multiples)
4. The player who gets your ghost can xv it to see what randart it has, so they know if it is worth killing.
5. When you leave a ghost, the game should look at its stats to decide if it is too tough for the level you died on, and if so move it to a deeper level.
6. Likewise, if the randart is too nice for the level you died on, move the ghost deeper.

[duvessa]

How about this:

If you die with at least one non-book non-weapon randart in inventory, you leave a ghost. If not, you don't.
When your ghost is killed it drops that randart (just one if you had multiples)
The player who gets your ghost can xv it to see what randart it has, so they know if it is worth killing.
When you leave a ghost, the game should look at its stats to decide if it is too tough for the level you died on, and if so move it to a deeper level.
Likewise, if the randart is too nice for the level you died on, move the ghost deeper.

literally on the wont_do list

[TeshiAlair]

I find ghosts really fun, yet fully agree they are TERRIBLE designs.

I think a genericized option would be nice: have a template for each skill, make a ghost based on the player's highest skill, scale it's stats/spellset based on DL.

[ZipZipskins]

An idea to keep interactivity is to name shopkeepers after dead players, maybe based on those players' skills.

Pros:
Keeps old player names reappearing like ghosts do
Gives the dungeon a haunted hotel-california vibe

Cons:
Welcome to pussysmasher's Used Wand Store! What would you like to do?
Idk coding or w/e

[Rast]

literally on the wont_do list

Good point. I edited my suggestion accordingly.