Putting the 'role' back in role-playing games since 2002.
Donate to Codex
Good Old Games
  • Welcome to rpgcodex.net, a site dedicated to discussing computer based role-playing games in a free and open fashion. We're less strict than other forums, but please refer to the rules.

    "This message is awaiting moderator approval": All new users must pass through our moderation queue before they will be able to post normally. Until your account has "passed" your posts will only be visible to yourself (and moderators) until they are approved. Give us a week to get around to approving / deleting / ignoring your mundane opinion on crap before hassling us about it. Once you have passed the moderation period (think of it as a test), you will be able to post normally, just like all the other retards.

Procedural Generation in RPGs

How do you feel about games that use procedurally generated content (dungeons, items, etc)?

  • I won't play games with procedural generation.

    Votes: 17 25.0%
  • I prefer handcrafted dungeons, but I'm willing to play games with generated content.

    Votes: 40 58.8%
  • I love procedural generation in RPGs.

    Votes: 11 16.2%

  • Total voters
    68

nlfortier

Esturia Games
Developer
Joined
Apr 28, 2020
Messages
128
While I enjoy playing a good hand-crafted story, I also love playing roguelikes and other games with procedurally generated content. I enjoy the unpredictability of the gameplay and it's exciting to never know what each playthrough will be like. What's your opinion on roguelikes and RPGs with procedurally generated content?
 

CryptRat

Arcane
Developer
Joined
Sep 10, 2014
Messages
3,561
There are two kind of things I like :
- What I like the most is spending 3 hours creating my big party, interesting dungeon layouts, puzzles, meaningful encounters, constantly new and unique. This requires mostly handcrafted content (generic is different from unique), and a save system. Occasional random events you must react to can make for some interesting resource management, typically you're going to use your ressource more aggressively in a hard encounter if you can't be sure to trigger again if you reload. But I prefer the core of the game, all the layout and puzzles and most encounters, to be fixed.

- What I like too, anyway, is skill checks, resource management, trap detection, item identification, permadeath. This works better with forced ironman mode, and permadeath and ironman mode work better with a single character, if the character dies I should restart from the start, no troop replacement. Then if I'm going to restart 100 times from the start then procedural generation makes sense, I see it as a consequence of a design rather than a goal.
Procedural generation does not mean you need procedurally created items for example, I generally prefer a good set of unique items which are randomly placed, and something similar can be said about opponents.

Therefore I like classic handcrafted RPGs and classic roguelikes but when someone proposes a mix of both, typically procedural generation but no permadeath, I'm generally very skeptical, for the mentioned reasons, I'm very skeptical about procedural generation as a goal (in RPGs, I'm not talking about other genres here) and only see it as a necessary evil of certain kinds of games, which are not my very favourite and are not classic RPGs.
 

KeighnMcDeath

RPG Codex Boomer
Joined
Nov 23, 2016
Messages
13,022
I love a solid handcrafted game but I also love procedural generation. Give me a handy-crafting but do me some random areas like say (1-2 infinite depth dungeons/area). Maybe extra dl content that is just random. A good game I will play again and again. A bad game I might give a once-over no matter if handcrafted or procedural.
 

DeepOcean

Arcane
Joined
Nov 8, 2012
Messages
7,395
Procedural generation is a powerful tool but it also needs another tool called brain, in theory, game designers should have a brain but, unfortunately that isnt always the case and procedural generation without a brain = homogeneous sludge of barely noticiable different content or 10000 tiny variations of the same thing, unfortunately, people blame the computer that is the "procedural generator" without questioning the wisdom of those that designed it.

Every generator has a scope, a game designed within the scope of a generator = good, a game designed with super ambitious design goals with an insuficient generator for the scope required = homogenous boring sludge for a content.

Problem for Daggerfal or Elite or other games notorious for their procedural generators is the over promising. "I gonna generate an universe!" are you going to simulate on your generator all the complicated interactions that can happen on this universe? "No, LOL, too much work, it was just a marketing point!", then the result is, you will have a massive scope with a very insuficient generator that cant keep with it. Same for Daggerfall and its copy pasted dungeons. Many designers think procedural content is just shortcut for less work, it will always fail if took from that view, procedural content isnt an easy shortcut to less work but key for a different approach, a systemic/simulationist approach to game design that is barely explored on traditional old rogue likes and new indie rogue lights.

There is a great potential for procedural content as is it can take cRPGs from the scripted theme park approach most have to a simulationist approach. In cases of cRPGs, it is a challenge because it isnt easy to simulate a good human designer with algorithms but maybe that wont be necessary as a systematic RPG would work on VERY different criteria from a theme park RPG and both have the potential of offering different experiences with different qualities and weaknesses. A terrible theme park cRPG like Pillows of Etenity or NWN 2 OC is as shit as a terribly designed systemic rpg leaning on procedural content, the problem is that people tend to compare the best case scenarios of one with the worst case scenarios of the other and with the lazy, hackish way AAA developer tend to approach procedural content as justa cheaper way to make games, it is no wonder people arrived to this conclusion.
 

Marat

Arcane
Wumao
Joined
Jan 6, 2017
Messages
2,603
Could someone please give an example of an RPG with procedural generation that wasn't absolute shit? Or better yet, one that was actually good? Daggerfall burned so hard into my brain I couldn't suffer procedurally generated RPGs in all their stultifying hideousness.
 

nlfortier

Esturia Games
Developer
Joined
Apr 28, 2020
Messages
128
Could someone please give an example of an RPG with procedural generation that wasn't absolute shit? Or better yet, one that was actually good? Daggerfall burned so hard into my brain I couldn't suffer procedurally generated RPGs in all their stultifying hideousness.

There are several good roguelikes. My favorite is Dungeon Crawl Stone Soup. As for more traditional RPGs, Diablo 1 and 2 both feature procedurally generated dungeons. They achieve good results by combining procedural generation with hand crafted content.
 

fork

Guest
I don't bother with games using pocedural generation for anything anymore.

Most of the time procedural generation of anything is used as an easy way out for the developer, to save them time and money on content creation. When all these things should be important to the dev... when I think about my dream game, the first thing that comes to mind are images, images of places, places which would have to be built by hand. Best example: Dark Souls. No kind of random proc gen would ever come up with a world like Lordran. There are dozens of other examples of worlds built by hand which no computer algorithm could ever come up with.

Another example: Dirt Rally vs Dirt Rally 2.

But it goes beyond level design. Even itemisation is shit when it relies on any kind of random process: Nioh vs Dark Souls.

Enemy placement? No, sorry, also requires human thought.

Maybe something like building the entire Milky Way justifies using proc gen? Well, Elite Dangerous' galaxy certainly is impressive, and I even travelled to the far edge of the Galaxy, Beagle Point, a couple of years ago (I'm sure they made it significantly easier since). But to be honest, I wish I could have my time back. Aweful game.

Then there's the argument about replayability. No, sorry. By the time your game gets boring, I'm done with it and no amount of randomly generated content can prolong its life and that's okay. Invest your time into an interesting, hand-crafted world, challenging difficulty and design to keep me interested, but don't waste my time with proc generated shit.

Proc gen is a powerful tool. But it has no place in gaming. Build your worlds by hand or they're not worth my time nor money.

And yes, roguelikes are shit.

Edit: And both Diablos would be significantly better with handcrafted dungeons. The randomly generated dungeons may be different every time you play them, but they're also bland shit every time. I'd rather play them only once or twice, but be blown away when I do because I recognise how awesome they are and realise there was a mapper or level designer or whatever they're called nowadays who really gave a shit.
 
Last edited by a moderator:

Jigby

Augur
Joined
May 9, 2009
Messages
336
The standard problem with handcrafted content is that you are a hostage of it. In a handcrafted content game experiencing handcrafted content to the fullest is paramount. If you interact with it, you do it only in predescribed ways, doing otherwise makes you afraid of breaking the game or breaking other handcrafted content. In Fallout 3, before you are allowed to set the nuke off which changes the gameworld irreversibly, you have to complete a quest chain. The purpose of the quest chain is not to give you exp, but to telegraph to the player that it's okay to interact with the content in this specific way. Bethesda is saying to you - it's okay shithead, we had accounted for this possibility, you can erase the city. In Ultima 7 you reload after you cast the Armageddon spell. Handcrafted content is rich and high quality.

You can draw parallels with chess and the queen problem. The queen is the high quality piece, the highest really. But the queen in some sense is the least free piece of them all. It is a hostage of its high quality. A lowly dirty pawn can bully the queen. You never develop the queen too early - it'd just get bullied by less important pieces. The queen is less free than a pawn.

In a procedural game a quest giver sends you to find a #123 necklace of king's nephew. You refuse the quest - I don't have time for this shit. The quest giver responds - don't have time? Fuck you bitch!
So you chop the guy's head off. And some guards enter the room so you chop their heads off as well. And you raise an army of skeletons and you raze the city to the ground.

In a handcrafted content game the quest will be of higher quality and you will suck the quest giver off, because experiencing handcrafted content is paramount.

So in short, if it's procedural, give me ways how to fuck with the content, preferably irreversibly. If it's a procedural RPG include a very rich chargen that has a lot of breadth. One criticism towards breadth - I think ever since d20 came out, diversity in builds really meant diversity in powerbuilds. You can choose between playing Kaze no Kama or some other 100 "interesting" powerbuilds. A rich chargen should allow you not to just create a demigod build, but also a gimped character. Gimped in nongeneric ways. If it's just about having 50% less BAB, that's no good. Caves of Qud has malformed mutations, Daggerfall has phobias. In Fallout 2 I can play a 10AP lucky sniper, or a 2AP slouching alcoholic with a pariah dog as a companion.
 

Jackpot

Learned
Joined
Dec 15, 2019
Messages
224
Nothing turns me off an RPG more than procedurally generated loot.
Being able to collect unique weapons and armor from ancient dungeons, powerful enemies, or stealing from elegant palaces is one of the best part of the RPG experience.
 

Desiderius

Found your egg, Robinett, you sneaky bastard
Patron
Joined
Jul 22, 2019
Messages
14,183
Insert Title Here Pathfinder: Wrath
People overrate the capacity of procedures to achieve comparable performance with (other) people.

Many such cases.

If you like it for it's own sake fine, but don't try to substitute one for the other or everyone loses.
 

bionicman

Liturgist
Joined
May 31, 2019
Messages
683
I don't mind procedural generation as part of a non-roguelike RPG as long as it's good, and I would define good procedural generation this way:

If a person plays this sort of game without knowing beforehand that it has procedurally generated elements, and never realizes during the playthrough that it has those elements, then this is good procedural generation.

Maybe I'm wrong, but whenever it's obvious something is "procedural" then it just feels lazy and random, i.e. it feels obvious it was constructed by a computer and not a person.
 

KeighnMcDeath

RPG Codex Boomer
Joined
Nov 23, 2016
Messages
13,022
Dungeon hack seems more like the random dungeon that didn't exist in EOBIII. Most of the creature assets are from there. Party would have made that game way too easy (and it was already pretty easy).
 
Joined
Jan 14, 2018
Messages
50,754
Codex Year of the Donut
with regards to level design, mixing procedural with handcrafted creates good results. Either through linking together hand-crafted mini-areas or generating parts of a level between handcrafted areas.
you should define what you mean by 'procedurally generated', do you include things like random encounters? unexpected random encounters are often some of the most fun parts of an rpg imo
 

laclongquan

Arcane
Joined
Jan 10, 2007
Messages
1,870,150
Location
Searching for my kidnapped sister
Procedurally generated content is a dream that still unfulfilled

Good application so far I know and acknowledge is Unreal World.
+ The world map get generated with every new game. While major geo-landmarks like coastal, big river doesnt change, caves, settlements etc change. This is important because if you have a cave area right next water source like lake or river, or best a white water river/rapid, you are golden.
+ The shops' content in each village settlement get changed in seasons or refreshed. This is important because if you are generated in the wild, the only way to get tool is visit settlement/village and if the shops dont have them, you have to travel half the world map, all the villages along the way, to buy.
+ The amount of prey (big animals) is pretty fixed but can be changed with unknown codes and procedure. You can hunt big animals in a, say 10x10 area maps, to exhaustion, but there's more flocking in from neighborhood, and possibily get respawn by your trap/ritual/quest/unknown reason.

Daggerfall I didnt play (and no desire to play) so I dont know if it can even reach this level.
 

Thal

Prophet
Joined
Apr 4, 2015
Messages
414
Nightime terror Missions in UFO: Enemy Unknown are some of the most suspenseful and best "designed" levels in isometric gaming. That shows that procedural generation can result in good and even better results than handcrafted approach. In fact, if UFO had handcrafted levels, it would probably be about as well-remembered as Chaos Gate: a solid game but not a timeless classic. However, by basing your game on procedural generation, you're still taking a big risk. Essentially you are trading your ability to create quality content in exchange for open-ended gameplay. That's a bad trade most of the time. UFO succeeds because the moment-to-moment gameplay is so good and, morever it only benefits from the open-ended nature. Also, the game is procedurally generated from hand-crafted pieces allowing reasonable complexity.

But can you apply this to rpgs? Probably not, though I'd still like to see a roguelike ARPG with well-developed combat and character development systems that tried. It would be something like a 1st person Nethack mixed with some handcrafted content here and there.
 

KeighnMcDeath

RPG Codex Boomer
Joined
Nov 23, 2016
Messages
13,022
Some consider X-com an rpg. I wonder if procedural generation in a mod for Crusader No Regret & No Remorse would be any good. I couldn't finish the second game as I just got sort of bored.
 

Polyphemus

Educated
Joined
Jul 5, 2020
Messages
34
Procedural generation doesn't really have a place in RPGs outside of roguelikes, and more games get away with being called roguelikes than there ought to be.

I especially dislike random generation of items in RPGs like the Divinity: Original Sins. It feels lazy.
 

V_K

Arcane
Joined
Nov 3, 2013
Messages
7,714
Location
at a Nowhere near you
Could someone please give an example of an RPG with procedural generation that wasn't absolute shit?
Unexplored is the only procgen game I know that generates levels that have an innate logic to them (sometimes even spanning multiple floors), like a really good human designer would do. But the result is so good that it shits on most handcrafted RPG levels. It plays much more like Ultima Underworld than your typical roguelike.
More roguelike devs should learn from it, especially given that Unexplored dev is very open about the approaches he used to achieve this quality.
 

JarlFrank

I like Thief THIS much
Patron
Joined
Jan 4, 2007
Messages
33,136
Location
KA.DINGIR.RA.KI
Steve gets a Kidney but I don't even get a tag.
Procedural generation cannot replace proper level and encounter design in any genre, let alone RPGs.

The idea that procedural generation allows for "infinite content" like some developers put it is a lie. Procedural generation does not create new content - it merely rearranges existing content in various ways, while being entirely constrained by the rules put upon it. A human level designer with at least an average amount of talent can easily create interesting scenarios that involve some outside-the-box thinking and custom scripting to make them unique. A proc gen algorithm cannot do that: it works entirely inside the box. It has no creativity, it just blindly follows the rules.

In a hand-made game, ideally every encounter and every location has some thought put into it. The designer consciously made the decision to put five goblin archers there, and three melee goblins, and one goblin shaman, and to put the archers behind cover, to create a challenging and fun encounter. In a procedurally generated game, encounters don't have any thought put into them - it's just a random selection of stuff deemed appropriate by the algorithm. You could end up with a piss-easy encounter of just two goblin melee fighters, or a way too hard one with three shamans throwing spells at you every turn.

But the procedural generation is never fully random. The developer designs a generic template which the algorithm then uses to create endless variations of that template. I do play roguelikes occasionally, but they're only coffee break games for me, I usually spend less than half an hour on one session then quit, while in proper hand-made games I can play for four hours straight. What I noticed about pretty much every roguelike I played is that the content gets very, very samey upon replays: the first dungeon level is ALWAYS populated by rats and kobolds. The third dungeon level has a chance to have the boss Broxflox the Necromancer, but he doesn't always spawn, just sometimes. Roguelikes have repeating patterns, they're not completely random, so after half a dozen games you've seen all the content the game has to offer. But none of the encounters are even half as good as in a hand-made game because they're just randomly jumbled together based on a generic template. And even if you end up with a really good dungeon design by chance, you will never encounter that particular level ever again - on your next playthrough, the rooms and encounters wil be arranged differently again, so that cool encounter you had in your current playthrough was a once in a lifetime affair. Meanwhile in a game with hand-made content, you can re-play your favorite encounter over and over again. The really hard Twisted Rune encounter in BG2. The encounter with Lareth the Beautiful in ToEE. No such thing in a proc gen game.

If a procedural algorithm were fully random, it would give you bullshit like a high level lich in the first dungeon level, while the final boss is a rat with two hitpoints, and you find the best weapon in the game on a mid-level orc you kill in a wilderness encounter. Also there'd be a complete mess of grassland, desert, and snow right next to each other, and a dungeon spawning right next to a village. Lots of stuff that makes no sense either from a worldbuilding or a gameplay design perspective. That's why the algorithm is fed with templates that makes sense: an undead-themed dungeon for level range 1-5. An orc-themed dungeon for level range 5-8. A pyramid with Egyptian-themed enemies for level range 8-12, located in the desert biome. And thanks to these rules, games with procedural generation soon become predictable. You KNOW there is a pyramid for high level characters in the desert biome... except sometimes it doesn't spawn so you're not even sure if you can look forward to that location. The starter dungeon with skeletons is always going to look similar: a bunch of hallways populated by weak skeletons and zombies. There are no real surprises. Everything is predictable, and therefore repetitive. And because proc gen algorithms can't think outside the box, you will never encounter anything non-predictable.

Procedural generation also works only for the simplest of level design templates. Usually 2D top-down dungeons with no verticality. Some of these newfangled survival games have proc gen for fully 3D worlds, but even then their architecture is way less complex than in hand-made games of the same genre. There are no algorithms that can create levels as good as the top-rated fan-made maps for Thief, Quake, Doom, Tomb Raider etc... all of which are games that heavily rely on good architecture and a fluid gameplay flow mixing some platforming/exploration with actiony combat encounters.

Fan-made levels are a good point here: they also provide almost infinite content as long as the fanbase is large and healthy enough, as proven by the Doom, Quake, Thief and Tomb Raider communities: all of these games are over 20 years old yet they still receive new fan-made maps every year, and the level designers become better at their craft every year, too. And even mediocre works by new authors are better than anything a procedural algorithm could provide. That's because hand-made levels created by a real person have a soul, a personality, a unique style. A great example are the Thief fan missions by Christine Schneider, who has made over 30 (!) levels since the early 2000s, with her latest released last year. Her early work is very rough and amateurish and not really worth playing, but I still enjoyed going through her entire catalogue because even in her crappy early work you can already see traces of what would later become her signature style, and with every release, her levels became better and better. It's fascinating to look at. You can witness a level designer's development from early work to mature work. Also by this point I have become so familiar with the Thief level design community, I can identify a popular author's style even if I don't know he made the mission.

Can a procedural algorithm give you level and encounter design with such a degree of personality? No, it can't. It will always just toss generic variations of a generic template at you. Lame and boring.
It doesn't even add as much replayability as the proponents of procedural generation always claim: due to the genericness and simplicity of proc gen levels, they become repetitive very quickly, and new playthroughs will feel the same even though the levels are (slightly) different from before.
Meanwhile I can replay some excellent hand-made games seven times without getting bored, just because of the high quality of the encounters. I derive more enjoyment from replaying an encounter I like - such as the Lareth encounter in ToEE's moathouse - with different party compositions than from replaying a roguelike where every dungeon level is of the same mediocre quality.

In fact, hand-made content deliberately designed to allow for different approaches is infinitely superior to anything a procedural algorithm could ever spit out.
 

Jigby

Augur
Joined
May 9, 2009
Messages
336
Stuff
.
.
.
Are you complaining about games not being fully random? Or are you justifying them for not being fully random? In one paragraph you seem to put them down e.g.
"But the procedural generation is never fully random. The developer designs a generic template"
a boring generic template that's not even fully random as opposed to an interesting handcrafted setpiece. But later on you acknowledge that it makes sense to not make things fully random
"Lots of stuff that makes no sense either from a worldbuilding or a gameplay design perspective. That's why the algorithm is fed with templates that makes sense".

This really goes from proc. gen. theory into randomness. Randomness on a large scale should obviously never be maximized, nor should it be the objective of proc. gen - you would just create a cryptosystem that way.
 

JarlFrank

I like Thief THIS much
Patron
Joined
Jan 4, 2007
Messages
33,136
Location
KA.DINGIR.RA.KI
Steve gets a Kidney but I don't even get a tag.
Stuff
.
.
.
Are you complaining about games not being fully random? Or are you justifying them for not being fully random? In one paragraph you seem to put them down e.g.
"But the procedural generation is never fully random. The developer designs a generic template"
a boring generic template that's not even fully random as opposed to an interesting handcrafted setpiece. But later on you acknowledge that it makes sense to not make things fully random
"Lots of stuff that makes no sense either from a worldbuilding or a gameplay design perspective. That's why the algorithm is fed with templates that makes sense".

This really goes from proc. gen. theory into randomness. Randomness on a large scale should obviously never be maximized, nor should it be the objective of proc. gen - you would just create a cryptosystem that way.

I'm explaining why "procedural generation" is bound to generic repeating templates that never push out any real surprises but stay fully within expected territory.

Full randomness would lead to wacky bullshit, but the template-based prodcedural generation just leads to generic, boring and repetitive crap.

This is why proc gen can NEVER replace proper hand-crafted locations and encounters. There's no algorithm in this world that can match the creativity of an actual human designer.
 

As an Amazon Associate, rpgcodex.net earns from qualifying purchases.
Back
Top Bottom