Sam Sorensen has three questions to determine the shape and context of an RPG he is reading:

  1. Is my goal to solve problems, or tell a good story?
  2. Can players (not characters) directly author the imaginary world?
  3. When the written rules disagree with the imaginary world, which takes precedence?

Without getting into his more detailed analysis, you can clearly see what is going on here. Depending on the answer to these questions the game you run will lean either towards the story game or the OSR play style. It is fashionable these days to claim that story games and the OSR are not so different. This mostly comes down to both play styles rejecting similar aspects of traditional D&D play. But the way they diverge is still quite distinct as these questions clearly demonstrate. With the notable exception of a game like Stonetop, it’s pretty rare to see designers try to thread the needle between these playstyles. Usually, a designer has some preferences about whether the focus is on gameplay or narrative, the type of agency players have, and whether the game is fiction first, or if the fiction is derived from the rules. One important tool in the designer’s toolkit is where and how randomness is injected into the game design. We see this in the shape of the randomness, where it occurs relative to player action, whether it resolves a question or poses one, and what exactly is being put at risk.

Distribution of randomness determines whether fiction or simulation is grounding the reality of play. The distribution of randomness says things about your game world. In so many flavors of D&D you roll 3d6 for an attribute. For players that want powerful PCs house rules are included to even further juice these numbers. In Mythic Bastionland the player rolls a d6 + d12 for an attribute. The difference is not simply between a 3-18 range, and a 2-18 range. The 3d6 roll is a symmetrical bell curve distribution, while the d6 + d12 is more of a trapezoid, flat across the middle and only sloping off at the tails, which makes it a comparatively more even distribution of outcomes. In practice that flatness means a Mythic Bastionland knight is far more likely to roll an extreme stat than a character rolled with 3d6s ever would be. So the Mythic Bastionland knights come out feeling more lopsided and interesting rather than clustered around an average. Distributions of dice rolls, in addition to helping form the shape of characters, also structure the kind of reality you are dealing with. In a Powered by the Apocalypse game you roll 2d6 and add a number for nearly everything. In other words, regardless of player choice, the chance of success, failure, or mixed success are broadly speaking equivalent. Players wield extraordinary powers to weave the details of the narrative relative to other forms of play, but barely have any power to alter the distribution of outcomes. Other games are more like Call of Cthulhu, and lean heavily on d100 rolls that establish a veneer of probabilistic simulation, even if this verisimilitude is more of a spectacle of reality modelling than the actual thing. D&D uses a variety of distribution choices, which adds complexity to the game, but provides more texture as well. Some adventures will have random encounter tables on a 2d6 which provides a distribution of encounter types, and these tables say something about the commonality or rarity of certain encounters, and others have you roll a single die, each type equally likely.

Randomness can happen prior to, or after a player action. You can think of pre-action rolls as game state randomization; random encounter rolls or rolling for the weather are examples of these. These are critical in OSR play where the GM has such a high degree of control over the game world. Offloading some of that responsibility to dice rolls helps facilitate the GM’s role as an impartial referee. The way these rolls are deployed also informs the type of game being played. A traditional random encounter roll will produce a random encounter with a 1-in-6 chance, or perhaps 2-in-6, or 1-in-12 depending on the flavor of the system. But critically either an encounter happens or nothing happens. Meanwhile the hazard die/overloaded encounter die generates some sort of event every turn. One type of roll is designed with a simulationist bent, while the other is designed with a gameplay focus. This type of rolling for game state is comparatively rare in story games. Sessions are designed to be heavily improvisational, and the GM has a much smaller hand in creating the world. Therefore this kind of randomization of game state serves less of a purpose.

Another word on the overloaded encounter die, it demonstrates how randomness is used to manage bookkeeping as well. Instead of having to track how long a torch or spell lasts, it is determined by the outcome of a die. You can even think of the original random encounter die as serving the same purpose. The random encounter roll assumes that there are NPCs roaming around the dungeon, but tracking them in real time would be tedious and not a good use of table time, so instead of tracking the particulars, this bookkeeping is abstracted away with a die roll. The overloaded encounter die takes this concept a step further and abstracts away other details that would otherwise require additional bookkeeping. Whether this is a good idea for your table or not will depend on how predictable you want things to be, and how much time you want to dedicate to bookkeeping at the table (in my experience most people would rather not). So randomization is also a tool for simplification of otherwise burdensome procedures.

Nearly any RPG that uses dice has randomization after a player action. Some game systems lean into this, but crucially, others minimize it. Despite it being such a prevalent feature of RPGs, I think most people agree these are the rolls that require the most thought and consideration in how they are deployed. I have memories from junior high school of D&D games where we would make strength checks to see if ordinary doors opened, and other nonsense like this. Random checks to that extreme turn your game into slapstick. Assuming that isn’t your goal a more tempered approach is called for. Typically, people will call for rolls either because the rules explicitly call for it, or else because the player has declared an action where the outcome is either highly uncertain and/or there are high stakes involved. Stated plainly, rolls are made for player actions of consequence. This is true for both story game and OSR play style. However, the OSR playstyle even sometimes takes this a step further and denies rolls for consequential choices, relying on player description of actions and GM adjudication instead. Instead of the traditional rolling to detect and/or disarm traps, the GM describes a room and the things that stand out about it, the player describes what they do to handle a situation, and the GM makes a determination if the threat or obstacle is mitigated. OSR games become in these moments a bit like a point-and-click adventure game, only instead of a potentially frustrating situation where you have to guess the designer’s intention for some obtuse puzzle, with potentially logical alternatives not working, in a tabletop game you have a GM that can handle unexpected player decisions. All of this is to say, sometimes you want a random chance to determine uncertain outcomes, but other times you might want to prohibit random rolls to facilitate enjoyable play that would otherwise be skipped over.

I should also probably briefly mention post-player action game state randomization. This isn’t something you’ll see mentioned in many blogposts, but I’ve seen it in enough tables in the wild, so it’s worth discussing. In this case, you have the players make some actions, and then the GM rolls some dice, and something else random happens. This is usually the result of magic spells, or in the funhouse school of dungeon design. Personally I hate this type of randomization, it turns the Information-Choice-Impact loop into Information-Choice-Impact-then some other Impact you didn’t choose. Maybe some people enjoy this sort of thing, it takes all kinds, but for people trying to play a problem solving type of roleplaying game, it’s extremely annoying.

Another aspect of randomness is whether the randomness is used to facilitate clarity, or prompt questions. Sometimes you make a roll to resolve something uncertain, but sometimes you make a roll to introduce an uncertainty. A to-hit roll resolves an uncertainty. Does my attack land or not? A roll in a story game does not resolve uncertainty, it poses a question. It determines whether an action is a failure, partial success, or success, but it doesn’t say how. It’s up to the table, usually the player who made the action if it’s a success, or the GM otherwise, to decide what happens as a result. Probably the most common question newcomers to story games have is how to handle mixed successes, since it is less straight forward than deciding if something happens or not. But failure too is subject to great artistic interpretation and license. In the interest of keeping things moving along, there is the idea of failing forward. So rolling for a lockpick check rarely results in simply not being able to pick the lock. What if it turns out you succeed, but success is bad? Similarly, with the to-hit roll, armor makes the to-hit roll harder, so presumably you might hit, but it’s just a glancing blow, doesn’t really harm the adversary, etc. This kind of interpretive ambiguity can either be destructive or creative to play, and papering over leaky abstractions in a system is a bit of a table artform. The desired quickness of play is relevant here. Rolls that are easy to interpret facilitate breezy, free-flowing play. Rolls that require consulting obscure tables, or involve a lengthy table discussion, will result in a more deliberate approach to play.

Randomness is frequently used to see if an action succeeds or not, as was previously discussed. However randomness can also be used to see if players avoid an outcome. Here the amount of GM agency is crucial. These rolls make no sense in a story game, where the GM only makes moves when a roll has determined that some harm or setback occurs. But this type of randomness goes all the way back to the saving throw, which is kind of an odd part of RPG history, so much so that Into the Odd and games informed by that design tend to ditch saving throws entirely. There are just (usually three) attributes, and they can double up as saving throws if need be. A reasonable enough simplification, but I also think there is something to the odd saving throws, if not in their specificity, then in the general purpose, which is that in a game system where the GM has maximal agency, the players have a chance to avoid harm, even when they screw up, as a matter of grace or luck. And so in the fringe corners of RPG design you are now seeing experiments here and there with ditching the ability scores, and only having the saving throws. Game design assumptions have now been properly flipped over from its head back onto its feet, and it is assumed that the players generally progress with their actions, and the only thing rolled for is to see what consequences they avoid.

Taken in sum, we can see randomness is a type of intentional mechanism that distributes the reality of the game world, responsibility among the players and GM, the management of actions, and reveals how the game world is revealed at the table.

As part of the Random Blogwagon this blog post is to be exactly 2000 words and posted on Jun 11th based on my rolls.