Story Driven Roguelike (Based on the Hero's Journey)

[Darlim]

So I recently had the idea to create a "roguelike" (don't know if it should technically be called that) that works in a similar way to the classic choose your own adventure books. Essentially, I'd have an algorithm which would generate a random "adventure seed", for example "The king of Storyland has recently been saved and the kingdom thrown into turmoil." The game would then branch that out into a number of chapters based on a predefined length. Each chapter would have a number of possible outcomes, which would grow as the story progresses (the first chapter could have three outcomes, so maybe the player decides to raise an army to reclaim the throne, seek dilpomacy, etc. The second would then have 3 outcomes for each outcome in the first chapter, so 9 outcomes, the next would have 27, and so on) until the final chapter, which would have a constant number of outcomes (for example, if the final chapter had three outcomes it could result in peace being brought back to the kindom, failure in the heros quest, or death of the hero).   Each chapter would be broken up into a number of "encounters", which would be lines of text describing a specific event (for example, the player enters the cave of Crombul and encounters the beast Crombul) The encounter would then have a set of choices, which would be what effect the direction the story goes (do you fight Crombul? Do you run from Crombul? Do you attempt diplomacy with Crombul?) In my vision of this game, after completion of the quest, an epic poem in the style of the Odyssey or Beowulf would be given to the player detailing the journey of the hero, so that they may share it with their friends

I think the algorithm would be based on Campbell's "The Heros Journey".  I know random generation of worlds is quite a simple task (check out http://donjon.bin.sh/), and concepts from his "Random Adventure Generator" could be easily extrapolated. It would take quite some time, but what do you all think of the concept? Has it been done before? What do you think of a computer's ability to pull off this task? Would the system be too complex to be fun? I know if it's poorly designed, the game would be exceptionally boring with vanilla plots, but i believe that if one could come up with a vast array of categories for stories, chapters, and encounters that would act as a skeleton for the quest, as well as a strong name generation algorithm, it can be done. As it stands, i plan to work on my own basic version of this as a sort of Viking saga type thing.

[reaver]

Procedural story is really, really hard to make it interesting. Overarching plot and cohesion especially imo. Check this out, if you don't know already: http://elderscrolls.wikia.com/wiki/Radiant_Story

[Bear]

It sounds like you want elements from Interactive Fiction -- in addition to, or instead of, roguelike games.

If you can do a decent IF engine that procedurally generates scripted story and encounters, that would be awesome.  But Interactive Fiction tends to be game as story, where roguelikes tend to be game as contest, so mixing elements is easy to do badly and hard to do well. 

Can you make a coherent hero's journey where the hero dies in the middle of the first act?  In a classic roguelike game, it's a contest, and the player character can die at any time.  In a classic story format, it's a story and the hero can't die until the story reaches a satisfying conclusion.  So there's a fundamental tension. 

Maybe you'd be happier doing a straight Interactive-Fiction game?

[mushroom patch]

If what you aspire to is randomly generated games of Dragon Warrior, I think this can be done without too much trouble. I've toyed with Campbell's ideas in exactly this connection, but never put anything real together. I think you have to really strip his formulas down to something concrete and simple to get anywhere, at which point I'm not sure you can retain any of the charm, depth, or variability of hero myths. On the other hand, even something fairly silly could be ground breaking in this area.

If you get anything like this to work, I think people here, myself included, would be interested to see it.

My sense is that you're right to worry that the added fun would not be worth the added work to develop it.

[Darlim]

I don't see this so much as a game I'm producing which people will play upon release, more an experimental game design, answering the question "can a computer write a story?" Of course the system would be made open source, and I would hope through collaboration, people could create a system which writes compelling, diverse, and emotional stories- i don't think the difficulty with it is so much the fundamental algorithms involved, but more the amount of content required. It would needs lists upon lists of words to put into flavor text, and synonyms for those words, and prefixes and suffixes for names to create a diverse population, categories for stories and encounters and chapters and such to fit into, which is where the open source nature of this project would allow it to shine.

[mushroom patch]

Doesn't sound like a game. You might try looking into efforts to write books algorithmically. This has been done and I think a lot of people are interested in it. It's not something that can be done well, from what I can tell, but that's another matter.

[Bear]

I don't see this so much as a game I'm producing which people will play upon release, more an experimental game design, answering the question "can a computer write a story?"

Actually, that's a fascinating question in itself, and a great project for the procjam coming up.  It is something that could be awesome in games, though probably not all that awesome for scripting the game's central quest itself, because, as I pointed out, the player can go off in some other direction from the story it scripts.  It would absolutely shine though, if you use it to make up legends and backstory, different for each game, that are tied to or introduce procedurally generated elements and artifacts.

[Rickton]

It's not necessarily a roguelike, but Michael Cook (the guy who organized procjam and does lots of procedural generation work, notably ANGELINA, the game-making AI) was working on a game with a pretty interesting-sounding procedural story engine.

Unfortunately, it seems like development is stalled (or at least not a high priority of his), but there are definitely some neat ideas in there that might help you out.

[AgingMinotaur]

Intriguing, although it's not the first time someone crops up declaring s/he is making something with a procedural storyline. Basing it on Hero's Journey sounds like a good idea. Also check out stuff like Vladimir Propp, Georges Polti's Thirty-Six Dramatic Situations, and this crazy list.

As always,
Minotauros

[koiwai]

Nice links, especially in the last message by AgingMinotaur. There must be a way to use them for procedurally generated stories. As a strating point, one can take some story they know well, and try to break it apart, describing it in terms of those typical narrative elements. It may help to see how to connect the pieces: 1) how to describe characters, objects, places, 2) how to connect the events - some sort of graph, a linear timeline of events? 3) how to take into account the personal development of each character, and how it may affect the story?

Even though it may be irrelevant to the topic, it seems to me that roguelikes usually take a bit different approach to procedural stories, at least if we look at Dwarf Fortress or URR, their authors start by creating a rich context, in which the story should (hopefully) emerge. Of course, it's not an entirely contradictory approach, but it seems more suitable for the roguelike genre, which does require this procedural world anyway. Then the story is built on top. I don't know exactly, how world generation / simulation works in Dwarf Fortress, but it seems that the history events and the map generation events interweave, so both the world and the story develop simultaneously.