Narrative branching, done well, is fantastic—it gives the player agency and lets them make the story their own (as it were). But when you're creating the story graph, it's easy to get lost in it and lavish care on one path at the exclusion of the others. You can easily end up with one or two long, greatly-detailed paths, and (because dev time is finine, and you need to move on to writing other parts of the game) a pile of other paths that are shorter and less interesting. If the player takes one of the shorter ones, they end up missing out on all your coolest stuff. The tools I would design for the kinds of games I created specifically made it easy to create a main story trunk with side paths (that rejoined the trunk), and more difficult to branch/loop/etc.
Of course, that's not the only (or even the best) way to do narrative design—Disco Elysium is a masterwork because it did the branching, merching, loops, jumps, random checks, and so forth, so well!
Played so much AaaaaAAaaaAAAaaAAAAaAAAAA!!! back in the day, still never 5 starred everything. Holds up IMO. Their other games are cool too
My personal preference would be to have a single, on rails story, where the player don't truly have a choice. It's an interactive movie.
Or, pick a sandbox mechanic, and let the player do what they want directly, and compute the consequence (the most common type being the physics system).
There are over 14,000 games listed on https://ifdb.org.
Perhaps you should play some of them and adjust your perceptions.
I think both have a clear place in gaming, since different gamers obviously look for different things.
I can't bother to play those kind of games. A movie will be able to deliver its stories a lot better than a game.
But with choice and branching you get to appropriate the protagonist(s) and some story events can be a lot more impactful then. Lately I played Cyberpunk for which you have some choices in most missions and the endings hit different. If anyone involved in the DLC story is around: kudos to everyone involved in making the "face in the crowd" ending. You play some almost super heroic character and due to your choices (which involve betraying and killing a lot of people) you get to survive: alone and back to generic human power level.
I don't agree. Something like SOMA would just be a generic sci-fi B-movie but it's an awesome game, even though there's no real choice and is in essence just a walking simulator.
have you not seen the success of the COD Modern Warfare franchise? Their single player game is essentially an on-rails shooter, with pivotal story points completely scripted (you "press the buttons"). There's no choice, there's no branching (of the story).
But people like to shoot, like to run around, etc. It feels like they have control, and it feels like the heroics in the story is their contribution.
https://www.ifarchive.org/if-archive/games/competition2019/T...
just seing "my" story is enough
interactive movie type games are great but they're a different experience from choice style games which are also great, (and ofc the article shows that there's many styles within this too, all with a different experience)
I don't believe LLMs can recreate the same authored experience that has a point of view. I think they'll be okay at genre work soon enough though, for better or worse. But thats not a type of game I'm personally interested in.
Done poorly, you end up with something like Fallout 4. Theoretically there are choices but they don't matter that much and many of the branches get little to no QA love so have bugs and problems (eg when I did my playthrough of fallout 4 I unwittingly managed to avoid the main choice you're supposed to be forced into doing which is to choose which faction to align yourself with) which had numerous buggy consequences because the devs clearly had never expected anyone to play in this particular way (I basically did a breadth-first traversal of all the different quests but put off a couple of key decisions because I couldn't make up my mind. In the end this pushed me past the point where I was supposed to make the choice and I ended up friends with just about everyone. Except the institute. Screw those guys). Yet another reason that game was so disappointing.
[1] The whispering hillock. If you know, you know.
At the end, you have just two options: You can join the institute, or make war with them. I actually really didn't care about this faction and wanted to get out of that storyline, but that's what the game offers you.
Of course many games lead you to a specific denouement, which is typically one where you've defeated all the bad guys and win. But those games don't make any pretenses about making choices, whereas Fallout does.
I was really impressed with how many choices you had in Baldur's Gate 3. They definitely recorded an insane amount of dialogue just to cover all the possibilities. But I did find myself annoyed with the ending here, too, which only allows two choices at the end.
If there are two choices to make, it's basically the same as one, because the story just stops there. I'd rather games stop pretending. Don't pretend that there are so many gray areas. Just let the good guys win over the bad guys.
An alternative approach for such a game would be to have the decision tree be oriented around a kind of moral compass throughout, and your deeds decide what you can do. So if you keep killing or betraying people, you become more evil, maybe physically deteriorating into a kind of ghoul, while normal people start to fear you and refuse to barter or make allegiances. But as you get more evil, you gain access to eldritch abilities, make friends with monsters, and so on. There may be a point where you can make amends and return from evil, or vice versa, but at some point there's a point of no return where the final trajectory has been decided.
There are some games that have some something like this, where you have a "reputation" among the good guys that you can lose by doing misdeeds. Not sure if exactly what I describe has been done, however.
[1] https://dishonored.fandom.com/wiki/Emily_Kaldwin#Emily's_Dra...
yep, and i argue that most game devs wouldn't do it well enough to justify it. The limited time and resources available means they're almost certainly better off not branching, but make one good main story and polish it.
Take a look at the examples you listed - they're all award winning games, from developers with serious experience, grit and determination to make the best game. They spent ages, and lots of resources to do it.
Even big studios, with similar or more resources, fail at making such branching good. It's a folly to think that a smaller game developer with more limited resources could make it better. Disco Elysium is almost an exception that proves the rule (or another example is Pathologic).
Funny little game. I find it quite charming in its eccentricities.
Papers please had something like this.
It also tells you the % of people that managed to reach a certain ending or branch.
> My personal preference would be to have a single, on rails story, where the player don't truly have a choice. It's an interactive movie.
I stopped to play those games. Movies are better at being movies then games.
English nit, if you mean you no longer play those kinds of games, as I think you do baaed on context, you should write
"I stopped playing those games."
We are very very far off from an AI being able to come up with compelling stories that are logically coherent.
i think you mean there's barely any good ones. The examples you come up with are the exceptions that prove the rule. Look at a game like Dishonered, where the story have _some_ branches, but it's half-assed imho. There's plenty more games where having gone for a branching story made the game more expensive, less deep, and harder to sell as a result.
As different authors can start their own new stories, one thing I often have to deal with is that they want to design their story to have both long path lengths (multiple chapters before an ending), and also high choice count. Those of you who know something about geometric series know that this causes problems. I often have to tell them they can't have everything they want, which causes minor drama. :)
As a result, one of our stories basically shot its "choice budget" in the first few chapters, leading to many linear paths in the latter parts of the narratives, which is fun in its own way.
Another of our stories has just started playing with the "gauntlet pattern" as the article describes. For this one, we decided that all chapters must be in the "same universe", just following different characters' perspectives, and are planning for certain "anchor chapters" where all characters come together for a meeting. Probably the detective questioning them as a group (it's a murder mystery).
All of our stories are supposed to be literary, so usually in third person, sometimes first, never the second-person. So we don't tend to use choices and chapters as directions and rooms; it's all about how the plot moves. We also don't track state; they're designed to be able to be printed as books people can page through.
Overall a super-fun project for me and a handful of other writers, it's been a consistent way to spend a few hours of fun each week.
If you picked up the key earlier, turn to page XX
Otherwise, turn to page YY
It was entertaining to a) suddenly realize I had missed an important detail or b) allow me to “escape” if I can’t find the key or just don’t like that part of the book.
shields: 100
lasers: 100
troops: 100
hyperdrive: 50
days: 0
The plotline involved you racing to some planet to deliver news of an impending invasion or something and if you didn't get there within I think 30 or so days the news would be too late and you would lose.
However, being the nerdy kid that I was I mapped out every single possible route in the book and then simulated all events going perfectly on each route and there didn't appear to be a single way to actually win. The author had not done the math right and the absolute fastest you could finish was like 40 days even if you got crazy lucky with the dice.
More recently, Jason Shiga has used clever mechanics like this a lot in comic book form, notably in Meanwhile. He's just finished a three-part series aimed at younger readers, Adventuregame Comics. All Shiga's stuff is great, highly recommended.
Perfect example of the "Branch and Bottleneck".
It even includes similar graphics at the start of each act.
https://www.reddit.com/media?url=https%3A%2F%2Fi.redd.it%2Fm...
Looks like a "Time Cave"
Indexed here: https://gamebooks.org/Series/457/Show
And some exposition from Hackaday a few years ago: https://hackaday.com/2020/08/28/a-tale-of-tutor-texts/
It looks like I need to take a closer look, that last article says
> often the wrong answer pages take you on a detour path to correct your thinking before rejoining the main line of the book.
which is what I was hoping to find.
One interesting (and very complex) approach I've seen in VNs is multiple interleaved paths. Each path looks like a branch and bottleneck, but at certain points a decision taken on one path blocks or forces an outcome on another. You can linearize it into a single "branch and bottleneck" with extensive state tracking (this is how it's implemented, after all), but it's far easier to model it as multiple paths.
Detroit is interesting, in that it includes some choices made by passing or failing QTEs. They really did the "You will get emotionally stomped if you screw this up!" well in that game. I don't know its structure well, as I only played it once. (So experience only one path.)
Unless you count time caves like The Stanley Parable!
Came here looking for the same.. some kind of map from the game design angle more towards game theory.
Fun semi related tangent, I was curious to know authors background, and the About page quotes Borges “garden of forking paths” which jives nicely with tfa. Cataloging rather than inventing is an underrated activity in math sometimes, and we need to do both. Game garden taxonomy!
GTA2 with it's competing gangs seem to have a "state" tracker in the form of reputation scores with the game, while having an open world map. As your reputation/state changed opportunities would become available/unavailable.
I still think GTA 2's system is impressive.
I do not see that branch-and-bottleneck relies on use of state-tracking any more than gauntlet.
https://www.drivethrurpg.com/en/product/296847/t-t-solo-desi...
It is old, but I do not know of any other books really on this topic. I enjoyed reading it anyway. One of the documents in it describes a simple manual algorithm for how to number the sections in a reasonable way (just randomly assigning numbers is not very reasonable, as anyone will learn from reading a book where the author did that). I implemented that in a simple pandoc filter:
Basically LLMs have to be given assets and game components that they can easily compose.
They're jacks of all trades, master of none.
This has its uses, but they have limits, and for now at least, those limits are under the threshold for that.
I have actually tried using them to make a text adventure to help learn German. The result was at the lower end of the quality range I've witnessed from LLM output: a nice first draft, not shippable, missing a core element, missing a lot of content, too simple, the kind of thing where you'd give the output of the LLM as a code challenge to a job candidate to see how they improve it.
Something like:
Generate a choose your own adventure story about a young boy shipwrecked on an island populated by hostile pirates. A hidden cave holds treasure. There is a jungle on the island. Dangerous jungle creatures inhabit the island and the boy can not fight. Also hidden on the island is a boat. Each story section is around 200 words long and ends with a multiple choice question for the player to select which path they want to pursue next. The story is complete either when the boy dies or finds the boat, after no more than 20 story segments.
I have some doubts the AI will be able to handle all of that and keep it interesting and coherent. This sort of storytelling requires some attention to detail that LLMs usually struggle with.
The LLM walks the line between hallucinating too much and sometimes not. Either way you can pretty much guarantee that almost all stories made in games now are already mostly written by an LLM. It’s just the writing is edited and curated by a human.
There's a second technical problem that such stories are represented by a form of state-machine and that you would need to recompile it on the fly, making many checks very difficult (you would need to be able to check reachability on the fly, chunk transitions, etc). I think it would take years to get to the level of some of the great IF games with an LLM, and not just a cool PoC.
But as a consumer, I lean far to the side of "give me a handcrafted tunnel experience with the illusion of choice" in the divide between consequences yes or no. I don't think I'd actually want this "simulation behind an LLM facade". If I'm in the mood for reading (or for listening to voice actors reading to me), I'd rather have it be something more meaningful than just a game state. But to those on the other end of the spectrum, this might actually be the holy grail of game building.
