The Alexandrian

Big Red - CGP Grey

I’ve just had an interesting discussion regarding the intersection between random generators and railroads.

Hypothetical Situation #1: You’re running a standard hexcrawl campaign. You generate a sequence of six locations. Regardless of where the PCs decide to go, they will encounter those six locations in the order that you prepared them.

This is self-evidently a railroad.

Now, take this hypothetical situation and begin stripping off locations until you’re left with a single location. Regardless of which direction the PCs leave town, they will encounter that specific location.

Still self-evidently a railroad.

Hypothetical Situation #2: You’re running a standard hexcrawl campaign. You create a random encounter table of six locations. When the PCs leave town, you start rolling on the random encounter table. As they encounter locations you cross them off the random encounter table. Since they’re randomly generated, however, the sequence in which they’re encountered may vary.

Is that a railroad?

Does your answer change if I similarly strip off locations until the random encounter table consists of a single location?


Does this mean that random content generators are an example of railroading?

No. But this thought experiment does demonstrate the complexity of these issues and the danger of trying to create some sort of “railroading purity test” for various techniques without considering the motivation, context, and methodology of their use.

The core distinction here is whether or not the players are making a meaningful choice. In this hypothetical hexcrawl scenario, the choice of direction has been rendered meaningless (since you’ll have the same experience regardless of which direction you go). And if the choice is meaningless, why are you having the players make it? Why are you lying to them about the choice being meaningful?

Random anecdote time: Many, many moons ago my players needed to explore the sewers beneath a major fantasy metropolis. I didn’t want to map the sewers out and the only game structure I really understood for that sort of thing at the time was dungeoncrawling. So I came up with a system which randomly generated dungeoncrawling maps for the sewer.

This worked just fine and the players were having a great time… until they realized that the terrain was being randomly generated. Their interest in exploring the sewers instantly evaporated: They knew that their choices were irrelevant. There was nothing that could actually be discovered. They were using a game structure of exploration, but they weren’t actually exploring anything.

This taught me a really important lesson as a GM: In order for an exploration scenario to work, there has to actually be something to explore. If all choices are equally likely to get you to your goal (because your discoveries are being randomly generated or because the GM has predetermined their sequence), then your choices become meaningless. And meaningless choices are boring and frustrating.


I’ve talked frequently in the past about the usefulness of procedural content generators: In a dungeon, random encounters can simulate the activity of a complex. In open campaigns, dungeons can be restocked with random generators. In hexcrawls, random generators can simulate the activity of the wilderness or they can be used to generate new locations on-the-fly.

These tools are incredibly useful. But how do you use them in a way that doesn’t negate player choice? How do you use them without the railroad?

The solution is actually quite obvious: Make sure the players’ choices are still meaningful even with the presence of the random elements.

A really simple example of this is simply allowing the actions taken by the PCs to affect the random generators: In a dungeon, for example, certain activities will create noise and increase the likelihood or frequency of encounters. In a hexcrawl, choosing to go to the Old Woods will cause the GM to roll on a different random encounter table than the Volcanic Peaks. And so forth.

Another straight-forward variation, in the context of an exploration scenario, is to create an environment with enough meaningful detail that it renders choice meaningful while the random content provides an additional patina of variety. For example, for my OD&D open table hexcrawl I keyed specific content to 256 hexes. That pre-existing geography creates a ton of meaningful choice: The random encounters that are being generated on top of it simply provide additional spice. Another example is the dungeon complex where the keyed rooms provide the meaningful choices, while the random encounter table provides a variety of activity throughout the complex.

A more complex variation would be a procedural content generator that creates an environment as the PCs explore it. This only works, however, if the location at which a piece of content is encountered becomes meaningful. This rules out purely ephemeral encounters (you meet eight orcs and you kill them) because the location is meaningless. But if the PCs are heading west and discover that the Salt Flats of Doom are over there and that Castle Vampire is on the far side of that difficult-to-traverse terrain, that geographical placement becomes meaningful if/when the PCs start mounting expeditions to Castle Vampire. (You could also imagine a structure where placing Castle Vampire and the Valley of the Giants next to each other creates a unique alliance that wouldn’t occur if it turns out that the Valley of Giants is on the other side of town.) The ability to revisit and reincorporate content, as is so often the case, is the key factor here.

Let’s consider a non-exploration example: In Technoir, the plot map of the scenario is randomly generated by the GM through play. Both the GM and the players discover the truth of the conspiracy together. However, before the players make any decisions the GM creates the mission seed which forms the core of the conspiracy. The mission seed provides the pre-existing detail which makes choices meaningful as the PCs seek to unravel the conspiracy (and, usually, penetrate to the heart of the mission seed). Even though the outcomes of these decisions are random, the choices remain meaningful because (a) they determine how the random elements connect to the mission seed (even as they also continue to redefine the conspiracy as a whole) and (b) the mission seed itself is a hard truth which can actually be discovered.

There are a lot of other ways in which random generators could be rooted into meaningful choice (or surrounded by it) without losing their utility. (And we would doubtlessly discover even more if we started poking around other game structures.) What you’ll note, however, is that all of these techiques are very different from simply taking the encounter you want the PCs to have and putting it in front of them regardless of which choices they make.

Go to the Railroading Manifesto

Share on TumblrTweet about this on TwitterShare on StumbleUponShare on FacebookShare on RedditShare on Google+Digg this

8 Responses to “Railroading Manifesto – Addendum: Random Railroads”

  1. Todd says:

    “This worked just fine and the players were having a great time… until they realized that the terrain was being randomly generated. Their interest in exploring the sewers instantly evaporated: They knew that their choices were irrelevant. There was nothing that could actually be discovered. They were using a game structure of exploration, but they weren’t actually exploring anything.”

    I probably shouldn’t go down this rat hole, but I can’t help myself. There are several items of note here.

    First, the game was fun and engaging as long as the players thought they were making meaningful choices. The success of the game as entertainment had nothing to do with the actual moral intricacies of meaningful choice, which I implicitly knew, but makes my head hurt when I think about it too hard. I think I’ll avoid Plato’s cave for the moment.

    Second, I don’t believe their choices become meaningful just because you predetermined the results of those choices. The fact that chosing turns Left-Left-Right brought them to the goblin lair doesn’t become any more of less meaningful because you decided it the day before rather than after the last corner.

    What I’m feeling is that what really gives choice meaning is rationale. Meaningful choice needs to involve calculation. Unless I can come up with something that unbalances the choices, then I’m ultimately choosing based on some mechanic that’s unrelated to the outcome, which is the same as just flipping a coin.

    As a GM, to infuse meaning into choice, I think I need 2 elements. First, I need detail. Color. Dungeon dressing. Whatever you like to call it. When the players reach a decision point and notice that there’s a draft blowing to the left, or that there’s a faint glow to the right, or that the terrain to the west seems more difficult than the terrain to the north, then we’re making choices. If it’s a maze of twisty passages, all alike, well, we all know what that’s like.

    Second, I need consistency in my world construction. Those colorful details that inform the players need to rationally relate to eventual outcomes. That doesn’t mean that every damp tunnel leads to a water monster, or that signs need to be blatantly obvious, but it does mean that there needs to be a rationale that both the GM and the players can (in retrospect) accept that links the decision inputs to the result. It also means that, if players make “informed choices” according to their experiences in your world, that sometimes they’re rewarded for that with favorable outcomes.

    Without these 2 elements, the most carefully constructed dungeon may as well have been rolled up randomly during play.

  2. Michael Prescott says:

    So, rather than thinking of railroading as a continuum (“are we railroading?”), I think it’s helpful to look at it as a relationship between player and GM contributions.

    By ‘contribution’ I mean a deliberate attempt to include something in the game. When the players go east instead of west, knowing nothing about either, all they’re contributing is, “We’re ready for the next thing now.”

    I see railroading as a GM trying to subtly (or overtly) undermine, neutralize or eliminate a player contribution in order to substitute one of their own. Perhaps the players suspect there are ogres to the west, and go east instead to avoid them (contribution: no ogres), but then the GM brings in the ogres anyways (contribution: nah, let’s have some ogres).

    To me, this language seems clearer than talking about meaningful choices. It’s certainly meaningful that the players chose to go west if Smaug is there, but only in hindsight; their contribution at that stage was still, “Next thing please.”

    Random tables are great way of letting someone not physically present (the table writer) make contributions.

    The issue I have with using them on a just-in-time basis (“This room contains.. uh.. ogres!”) is that it prevents players from getting much in the way of foreshadowing, so their contributions are limited. When they find themselves in a hallway with puddles of acid-smelling water, the stone walls smoothed clean of lichens, the absence of rat droppings that were everywhere in the upper hall, then they can make a more meaningful contribution.

    I suspects it’s fairly complicated, psychologically, but if I realize that the GM is generating the dungeon randomly, as we move, then a part of me realizes that my contribution is actually to create more dungeon! This is the opposite of my character’s objective (get to the end of the dungeon, eliminate this threat, etc.), so there’s a kind of dissonance that pops me out of the character’s head.

    But (and I think this is worse) is that it reveals a hard upper limit on the impact of my contributions as a player – I’m pretty much stuck with “Next thing, please.” I’m unlikely to be able to infer something (perhaps I might detect a pattern in the dungeon’s design) that lets me skip to the end and avoid a lot of trouble, because no such pattern exists.

    (I’m ignoring all the choices that can be made with the random content that does emerge – like, this lead-handled mop might be just the thing to mess with the jellies or whatever.)

  3. Martin Kallies says:

    I am very much in agreement with Todd. A choice is meaningful when the player have the ability to not just chose their current action, but also chose between different expected outcomes. Right or left is never a meaningful choice. Fast or safe is a meaningful choice.
    To drag up the old example for countless hypothetical RPG situations, The Lord of the Rings, the choice of the Hobbits to go through the Old Forest was a meaningful choice. Taking the road would be fast and mostly safe, but has the risk of being seen by the riders. Going to the forest avoided that danger, but meant taking a much slower route through an area with lots of less great threats. Do they risk running into one encounter they can not win, or instead face lots of smaller encounters that could still cost them dearly? It’s two bad options, but the choice is meaningful.
    When I set up adventures, I always try to put plenty of clues that give the players some degree of being able to expect what lies ahead. Do you want to go down the corridor full of corpses or not? Do you want to rush after the enemy sentry that just spotted you and ran into the forest? There will be consequences either way and the players have to chose which likely outcome they prefer and what dangers to face.

  4. Gamosopher says:

    I would add another way to make sure you are giving meaningful choices : the ability to get information to base decisions on. The random generation of hexes (or sewer passages) does not offer meaningful choices if there is no way to know what will be in a hex/room before entering it. If the PCs asks questions about the surroundings, then the GM roll the random hex to give them the information, nothing is railroaded : the PCs can decide to go there or not on the basis of the information. They can decide to gather informations about another direction, and decide which one will be their destination. They can use those informations to prepare themselves for the voyage.

  5. bygrinstow says:

    One way to generate ‘meaningful’ random dungeons (or sewers, or whatevers) on the fly (if that’s what you must do): generate sections two or three steps ahead of the party, e.g. what’s out straight ahead, what’s to the left, what’s to the right. When they go right, you repeat the process.

    Yes, it’ll take a little more time and you’ll be generating more than you’ll use most likely — but it lets you throw in that “there’s a glow to the right, a breeze from the left” detail that let’s them make choices that mean something to them.

  6. Justin Alexander says:

    @Todd: “Without these 2 elements, the most carefully constructed dungeon may as well have been rolled up randomly during play.”

    I think you’ve nailed it exactly. Random generation is a really easy way to strip these elements out of the dungeon (although some forms of random generation can compensate for that), but there are also other ways to strip those elements out of your design. And you end up with the same problems in either case.

    The only proviso I would add is one that I’ve made before: Players making a choice without having relevant information is only a problem if they don’t have the ability to gain that information. The choice to not get that information is a meaningful choice. (Or the failure to do so is a meaningful consequence.)

    @Gamosopher: “I would add another way to make sure you are giving meaningful choices : the ability to get information to base decisions on.”


    This is also why I find design that revisits material to be really effective: Whether it’s monsters reinforcing areas that the PCs previously cleared or simply restocking an entire dungeon, the geographical knowledge gained on the previous expedition creates a rich landscape for meaningful choice when you come back to it.

    NPCs become more interesting the more you interact with them for similar reasons.

  7. Dan Dare says:

    Interesting discussion. So a procedural generator like a dungeon generator could, as an example, require generation a couple of rooms ahead in every direction from the player position. Those rooms should generate what I call “projections”, which are encounters or clues or foreshadowings that provide information about the new room. These should appear in existing locations that the players haven’t moved through this session, with at least one being in their path.

    A question is then how do you make this fast? One thing would be that only some rooms are worthy of this treatment, and others are just infrastructure that gets encountered on the way. Another useful thing is to have a legendary room that is supposedly somewhere. As players move through they encounter clues and foreshadowings that begin to constrain where that legendary room can be found. At some point they will have nailed it down and the room cannot appear anywhere else.

  8. Dan Dare says:

    Just had a thought about speed. Go back to the 6 prepared rooms. Now have each actually be a 2 room set. One is the “heads up” room and the other is the encounter room. As players progress to random rooms have each one have an exit or two that leads to one of the heads up rooms. The players are now solidifying meaningful choice situations that have been prepped ahead of time, delivered randomly and not forcing an encounter. The encounter rooms can even have clues about other encounter rooms, by describing their heads up room.

Leave a Reply



Recent Posts

Recent Comments