The Forge The Internet Home for Independent Role-Playing Games
About the Forge | Articles | Reviews | Forums

Applied Theory
by M.J. Young


There are among gamers those who like to theorize, to attempt to understand and explain our hobby, why we do what we do, and why it works when it does. For some gamers, this makes no sense. We play to have fun; we design games in whatever way seems to be the most fun. Some despise theory, and see no use in it. If you have no use for theory, then this material's only offering is that perhaps someone else might. I am among those for whom theories are fundamental, so I would be interested in theories if they had no practical value to anyone. However, since I think that theory is the foundation for action, I can't imagine any theory that would have no practical application. I am thus exploring the practical application of role playing game theory.

Specifically, I'm looking at the theory commonly known as GNS. This theory suggests that role play styles divide into Gamists who enjoy facing the challenges of play, Narrativists who enjoy great stories that involve themes or issues, and Simulationists who seek to know what another reality might be like. The theory, which owes much to many people over many years including the discussions on the newsgroup, first took this form with these names when formulated by Ron Edwards in his article System Does Matter, originally published on Gaming Outpost, but since lost and republished at The Forge. Mr. Edwards has expanded on this theory with several other articles, and debates and discussions of the details have been held on the forums of several gamer web sites. Periodically in those discussions, someone suggests that the theory isn't much use, because it doesn't tell you how to design a better game.

In response to this, it can be and often is answered that this is not really a theory about how to design games. It's a theory about what gamers are seeking when they play, and as such has its most effective application as a diagnostic tool for play groups that seem to be internally at odds. In this context, if we have players who are trying to get different things out of the game, having some terminology and definitions by which to discuss what each is seeking can be invaluable in resolving conflict. If all GNS theory did was resolve such conflicts, it would be valuable. However, one cannot read so much as the title of that first article, System Does Matter, without absorbing the idea that game design itself is part of the problem, and therefore could be part of the solution. Those who are asking how to do narrativist, or gamist, or simulationist design are asking valid questions; the answers generally given to these budding designers have been inadequate, as they in essence amount to telling people to design whatever they like and then test it through play to see how it works.

Answers need not be quite so nebulous in this area. Once the theory is understood, there are aspects to it which suggest practical approaches to designing consistent games that support particular sorts of play. This isn't about rules heavy versus rules light design, or about setting detail, or even about things like whether you play your character in the first person or the third person or have control over things beyond the character. It's about how to create games which support and facilitate one approach to play under the theory. Once you have the basic concept of a game idea, application of the theory can greatly aid many of the design details.

Game design has many areas; no one area will completely control how the game is played, nor is it necessary for design priorities to be considered in relation to all of these areas to be effective. The social context of the gaming group playing the game can have a significant impact on whether the game works at all, and whether it is played as designed. Once it is agreed that a particular group of players is interested in playing a particular kind of game, designing to that desire need not be so mysterious as some imply. Character generation, resolution mechanics, credibility distribution, advancement, and rewards are some of the aspects of design through which particular GNS preferences can be facilitated, and designers can devise approaches to each of these through such considerations long before test play begins.

Underpinning this article, which will be somewhere on the edge between theory and practice, is this basic principle: conduct will be preferred if it is rewarded, and avoided if it is penalized.

There's a lot to cover, so coverage will of necessity be sketchy; however, it is hoped that this will provide some foundation for practical applications of the theory to game design.

Character Generation

Nearly all role playing games include a section on how to create a character. Very few give more than a line or two to considering what you are creating when you do so. Failure to consider this aspect leads to design problems; in GNS terms, an understanding of what you are creating is far more important than how you are creating it. Put another way, if you know what you're creating, how to do that will more often than not fall into place. Too many games create characters without thought to what they are. Characters are not really people; they are functional components of a game world which are manipulated by players to achieve goals. They are, in a word, tools. It is at this point in play that you are attempting to guide the players into designing the right tools.

Put that way, it becomes obvious that GNS considerations are very important to the question of what you are designing. If you guide the players into designing hammers, they're going to wind up with tools that are very good for hitting things; if you want them instead to write stories, you need to have them design pens. You need the right tool for the job; if you don't have it, there will be a tendency to try to make the job fit the tool.

So what kinds of tools are needed for the major types of jobs?

Gamist tools are easy to recognize and easy to design. A gamist character has to be up to the challenges which lie ahead. What that means in detail depends on the nature of the game in play and the preferences of the designer. Some gamist characters can be extremely focused on the central challenges of the game. Combat is the most common example of this, and a character's effectiveness in a certain type of gamist design would be measured by his abilities to deal damage and survive damage, to stand up to the fight. In a very different sort of game, racing could be the challenge, and character design would be narrowly about how fast the character is without reference to much else. However, skill-driven games can also have a strong gamist design foundation, if the skills are geared to meet potential in-game challenges. Driving or piloting skill, medical skill, hacking, picking locks, and hiding are all candidates for gamist design, because they are there to provide the player with options, ways to beat challenges presented in play.

That's not to say that narrativist characters can't have either power or skills; they can. However, narrativist characters need to be connected to the world. They need to be built such that things matter to them, and they matter to things.

Just as there are multiple ways to design a character effective against the challenges ahead, so too there are multiple ways to integrate a character into the world. Creating relationships with other characters is a valuable factor; giving the character beliefs or principles which will be challenged by events is also useful. Character history and character goals might matter, provided these are of a sort from which issues arise. A long-standing feud might be merely fodder for another fight; done right, it might become an issue for exploration. To build a narrativist tool, you should have something that is already tied in to the ideas you hope to explore.

Simulationist tools are perhaps the most difficult to see or to design. There is a sense in which no words which describe a simulationist character don't apply equally well to another sort. He must be effective, able to change his world; but then, gamist characters must be effective in that sense. He must be human, seeming like a real person; this is true of narrativist characters, certainly. Perhaps the most important characteristic of a simulationist character is that he must be accurate, that is, he must clearly express something real and credible within the setting such that he has exactly the amount of impact on events and persons around him that he should have, no more and no less.

This does not mean and should not be confused to mean that a simulationist character is more detailed than any other. A simulationist character could have history, principles, character, goals, relationships, skills, and all the things that support other forms of play; he could as easily be three numbers on a statistics sheet defining his effectiveness. What matters is that he is given form as an integrated part of the world, where he fits as if he were born and raised within it. To understand him is to understand the essentials of the world in which he lives, and vice versa. He is what he is, and in some sense not what anyone outside his world wants him to be. He is in the world and of the world, and as a tool he reveals the world to us through himself.

Now that we've got some idea of what kind of tool, what sort of character, we're trying to create, how do we create him? Do we use point systems for gamist characters, lifepaths for narrativists, and dice for simulationists? Wrong on all counts. Those methodological considerations in themselves have nothing to do with what we are creating. You can create any sort of character with any of them.

Take lifepaths for an example. We could start a character in his teens and move him, by a combination of die rolls and choices, through military training, education, private sector work, and other areas through which he builds up skills that prepare him for the challenges which will come. We might instead start a character younger, take him through his early years, develop school friends, relationships, family connections, life partners, coupled with the sort of moments that form opinions and beliefs, and so derive someone ready to explore the themes of the game. We could have a much broader selection of options, creating characters who have far less focus and more breadth of background and experience, who thus feel more real, as the tools we will use to explore the world. The idea of using lifepaths didn't matter; it was the way we used them that made the difference. It isn't how you build the character, but what kind of character you build. You'll certainly have to adjust the character generation system to build the right sort of character, and you might find that you have more luck making one mechanic type work than another for what you wish to do, but the answer isn't so much in the type of mechanic as in the targeted result.

I make some suggestions on character generation systems in Game Ideas Unlimited: CharGen (which gives some general thoughts and focuses on freeform design) and Game Ideas Unlimited: Negative Points (ways to smooth out some of the problems in dice and points systems).

Resolution Mechanics

Mr. Edwards has said that system within a game is the equivalent of time. To understand this, you have to understand something about time: it is the medium for change. Without time, nothing changes. In the game, the system determines what happens, what changes; without it, nothing changes. Thus the system determines and controls change, and therefore is effectively time for the imagined world.

Yet this, too, can be very important in supporting or impeding GNS preferences. How outcomes are resolved matters very much.

Although it has been said many times, it is worth saying again that diceless systems don't in themselves support narrativist play. They may be used for narrativist play, but they may equally be used for gamist or simulationist play. So, too, such general matters as dice pools, bell curves, granularity, and the other aspects of system which garner so much discussion (particularly from system monkeys) are not in themselves relevant to GNS concerns. As with character generation, it is what you do that matters, and not these questions of how you do it.

What are you attempting to do? The function of system is to provide the medium for change; more specifically, resolution mechanics are there to empower players to make the kinds of changes they wish to make within the game world and to interact with the consequences. To the gamist, resolution mechanics are in a sense both the obstacles to overcome and the means by which to overcome them. To the narrativist, they are the means by which the theme impacts the character and the character addresses the theme. In simulationist play, these are both the limitations on change and the power to explore it.

For gamist mechanics, you want something resolute; there usually needs to be clear victory conditions, clear failure conditions. It also helps if the system is responsive to player choice, that is, if there are ways that the player, through his character, can impact the probability of success. This could arise from strategy, or from skill or equipment choice, or from any decision which should and does give the character an advantage. Few things are more frustrating to gamist play than for the character to do things that seem to the player to make sense as ways to improve the odds, only to have these amount to no effect.

Even unrealistic strategies are helpful as gamist tools. A game that gives combat bonuses for sound, conservative defensive strategy can be very gamist, but so can one which gives combat bonuses for brash and brazen boldness, charging, screaming, doing over-the-top stunts. What matters is not how the bonuses are earned, but that in fact it is possible to manipulate the chance of success through character choices.

Although combat is the example here, it should not be thought that it's only in combat that such things matter. If a character can improve his chance to pick a lock or hack a computer or repair a wound by taking particular actions, this gives support to gamist play. There is a challenge to meet. The resolution system will tell whether or not the player succeeded, with certainty, but the player has the ability to tweak his chance of success through his approach to the problem.

Although it may sound strange to say that a resolution mechanic need not be resolute, for narrativist play it is often better that it not be. A gamist wants to know whether he succeeded or failed; a narrativist wants to know whether his efforts had an impact. In a combat mechanic for the use of guns, it is quite sufficient for a gamist system to determine whether the shot hit the opponent and how severe the injury is; for a narrativist system, things are probably a lot fuzzier (from a certain perspective). The shot should have the power to frighten the opponent and cause him to flee, for example. From the gamist perspective, that would be a miss; from a narrativist perspective, that's a success. Thus it helps narrativism if the resolution mechanic provides more of a degree of success rather than a strict success/failure determination.

Simulationism wants to know what would actually happen, given the assumptions of the setting. That doesn't mean realistic, in the ordinary sense; it means believable within the bounds of the imagined world. A fighter putting his spear in the ground and then using it as a bracing point as he runs across the chests of his adversaries kicking them is not terribly realistic, but it does fit the imagined reality of a certain sort of world, and thus could be incorporated into simulationist play in that world. In fact, if it has been established that a particular fighter can do that, simulationist play would dictate that he do so in any situation in which that would be the obvious response, unless there is reason to think he would do something else at that moment.

Thus resolution mechanics which support simulationist play are those which make outcomes correct within the setting. Much as with narrativism, this is often served by some form of relative success and relative failure, a determination of how well the character did; but like gamism, this generally needs to be resolute. A simulationist doesn't just want to know that he missed; he wants to know how close he came to hitting.

It might help put the entire question of resolution mechanics in perspective by imagining that a character runs, perhaps fleeing from an attacker. The gamist wants to know whether he ran fast enough. The narrativist wants to know how running mattered. The simulationist wants to know how fast he ran. Although in a sense, all three are concerned about escaping the adversary, they view this in different ways.

Credibility Distribution

Before anything can be said about credibility distribution, some explanation of what this means is important.

In roleplaying theory, it is recognized that there is within the game a shared imagined reality in which actions occur. Players, including the referee, contribute to the content of this reality through statements made to each other. These statements amount to, "This is what I want to have happen in our shared imagined world."

Of course, player statements may be contradictory; after all, players have different aims. Bob's character and Bill's character might get into a fight, and Bob might say that his character hits Bill's in the nose, to which Bill answers that his character ducks that punch and knocks Bob's to the floor. Now we need to know what actually happens in our shared imaginative space, or we're no longer imagining the same reality. Game systems must apportion credibility to address these issues. Credibility is the degree to which any person at the table has the power to define what is happening in the shared space.

You might think that in traditional games, only the referee has credibility. That is incorrect. All players have a measured amount of credibility. The referee rarely is able to say what actions any player's character would take--only whether he succeeded in that action. Thus non-referee players have credibility, too, even in such games, as they get to state what their characters attempt. Credibility means someone gets to decide what rules apply to the situation, when resolution mechanics are used, what the dice mean, and ultimately what happens in the shared space; it also means stating what actions characters are attempting, what they are saying to each other, and how they are reacting. Credibility is always shared. The issue is how it is shared.

This is sometimes confused with something called narration rights, that is, who gets to describe the scene. There is some connection between the two, but it is not absolute. For example, a game could state that each player at the table is allowed to contribute one fact which must be included in the outcome of the event, and then the player who has the narration rights must state what happened in such a manner that all of these facts are included. He himself might not have determined anything that happened despite narrating all of it. In most instances, narration rights include credibility; yet even in games which pass narration rights around, it may be the case that the referee can veto something stated in the narration if it goes counter to something known to him but not revealed to the players.

Gamist play is best supported in most cases by narrowly and clearly delineated credibility. Because the point of play is to overcome the challenge, it is not usually effective for the player facing the challenge to decide that he was successful. Since it is also possible that the players may find themselves in competition, it would be equally problematic for that decision to be made by a potentially opposing player. It is important to gamist play that credibility be clearly distributed, and that the player who determines the outcome does not himself have a stake in the outcome. This is why traditional games placed this power with the referee. He was viewed as the neutral arbiter, and as long as the players trusted his neutrality he could determine what occurred in the game world without problem. It is not impossible to eliminate the role of the referee from gamist play, but to do so the design must clearly establish who has credibility under each circumstance, so that disputes do not occur over success and failure. Too much player credibility can actually thwart gamist play preferences, since a player who can merely decide his character has been successful has lost all sense that there was any challenge to the victory.

This does not mean that players cannot be given credibility beyond the control of their character actions. The credibility to add color and detail to a scene are not contrary to gamist concerns. What matters is that such credibility cannot provide ways to eliminate the challenge itself. As one of my sons observed, you can't give the gamist player the power to invent a plus four sword lying on the table within reach and expect the game to be functional at a gamist level. The challenge must be maintained.

Narrativism usually requires more credibility in the hands of the players. Players are not competing with each other nor trying to beat the game, so giving them credibility is not detrimental to play in the same way it tends to be for gamist play. Rather, players need to be empowered to address the theme. Director stance, that is, the ability for the character players to add elements to the setting and events on the fly, is not uncommon in narrativist play. It is not essential to it, but works better with it than it does with the other preferences. Severely restricting credibility tends to stifle narrativist play, as it takes from the players their ability to make the statements they wish to make.

It is much more difficult to address credibility distribution in simulationism. What matters here is the verisimilitude and consistency of the shared imagined reality; that is, all players must see the same thing and believe it. This does not preclude broadly shared credibility; it does require a solid agreement on the nature of the reality. If we're playing in a medieval fantasy world, exploring an abandoned castle, a player given credibility could announce that he saw objects on a table, and describe the objects he saw. As long as those objects do not upset the agreed nature of the reality, such credibility is not problematic. Thus it is evident that the objects could include bottles and lamps, perhaps swords and daggers, possibly jewelry, all things which would typically be found on such tables. Were the player to describe seeing laser guns or kinetic blasters there, this would clearly violate the agreed reality, and his credibility would cease at that moment. However, there are difficult cases here. The player might describe finding the famed lost jewel of Prince Balthazzar, or opening a bottle to release a djinni, or discovering a scroll with a map to a hidden treasure. These, too, are all plausible within the setting, but may be stretching the credibility of the player. For this reason, it is more common for simulationist games to prefer narrower credibility for the players and broader credibility for the referee. It is not a necessary arrangement, but it does tend to support simulationism better.

Again, credibility distribution does not determine the sort of play that will occur in itself; it tends to support different preferences when configured different ways, and thought should be given to the amount of credibility players should have to facilitate reaching their goals.


It must be asked whether it is necessary for characters to improve during play; the answer is that this is never necessary. It is not necessary for simulationist play, certainly not for narrativist play, and surprisingly not for gamist play. However, it is often desirable in each mode that characters have the power to improve and advance in some sense. The more significant question is, in what sense can the character advance?

Most of us are conditioned to think of character advancement or improvement in strictly gamist terms: a character advances by getting better at what he does. That is, his ability to face the challenges increases. That there could be character advancement that has nothing whatever to do with this is surprising to many players. Yet consideration of this mode of improvement should give us some clues regarding how to improve characters for simulationist and narrativist play.

In discussing character generation, it was recognized that the character was a tool which the player used to achieve goals. Improving a character means making it into a better tool. Thus if a starting character in a gamist game is a rubber mallet, improvement might take it through stages of being a tack hammer, claw hammer, ball peen hammer, sledge hammer, jack hammer, and ultimately pile driver. That is, the character gets more effective at meeting the challenge, because it is a tool designed to meet challenges.

If we consider the function of the narrativist character, we find that it exists to enable the player to address the theme, and as such it has to be tied in to the issues of play. Improving the character means connecting it more deeply or in new ways with the theme. It can mean deeper commitments, stronger relationships, more determined moral positions; it could also mean greater conflicts, increased doubts, more personal connections. In a game exploring issues of sexual identity, a character who has always decried homosexuality as a moral perversion could be advanced by the discovery that his best friend is homosexual, creating a tension between his friendship and his beliefs. It's not impossible for narrativist characters to get better at things they do, but it is far more supportive of narrativist play for them to advance by becoming more integrated into the issues. The tool that started as a pen has advanced to becoming a word processor: it is now able to address the issues at new levels and in more facets.

Since simulationist play is about exploring the imagined reality, character advancement is best if it enhances that ability to explore. Our magnifying glass gradually advances to an electron microscope; our field glasses to become the Hubble telescope. The particulars of how this works are greatly dependent on what the game is exploring. If the exploration is of a physical world, greater mobility within that world is the logical route to improvement. Given exploration of a complex society, increased contacts and exposure within the society provide the answer. Exploration of historic or fictional events requires greater access to the events. Combat effectiveness or skill improvement can be simulationist if these empower the player to explore more difficult or dangerous areas of the game world. The variety of possibilities makes it difficult to be specific, but the answer in any situation is found the same way: identify what the tool facilitates, and how to make it facilitate this more effectively. One thing that is consistent across simulationist play in this area is that character advancement, like everything else, must mesh with the in-game reality. A character exploring the setting by working as a local reporter can advance through being assigned to a larger beat, but only if it makes sense in the context of the world that this character would receive that assignment.

Again, it is not necessary in any style of play for characters to improve or advance. Gamist play can be about beating increasingly difficult opponents with the same resources with which you started. Narrativist play can interact with the world through a static character. Simulationist play can be limited to that which the character can access. All play styles can be enhanced by the ability to improve and advance characters within their own terms. More importantly, if a game design provides character advancement options, these will influence the way in which players approach the game.


I have written elsewhere of rewards systems, and the necessity that they be two-pronged. I first considered the issue on the forums at The Forge, and later contributed a brief statement on it to The clearest and most complete statement on the subject is in the aptly-named Game Ideas Unlimited: Rewards; but as that is for Gaming Outpost subscribers only I'll recap some of it here.

There are two aspects to rewards systems, both equally important. Many designers fail to realize this, and so design rewards systems that are internally conflicted--they encourage opposing play priorities.

There is a clear example of this found in examining the popular experience points systems of games in which you kill monsters and get treasure, which gives you points, which raises your character level or skills, which makes you better able to kill monsters and get treasure. This is a coherent gamist rewards system: everything in it is geared to encourage the process of killing monsters and getting treasure, that is, overcoming the challenges of the game. It is a system that does not need repair, because it works extremely well at doing what it is supposed to do.

However, there are many referees who don't like what it does. They think it encourages players to focus on killing monsters and getting treasure (which is correct, because that's exactly what it's supposed to do). They don't want that to be the focus of the game; they want to encourage role playing, or character development, or dialogue, or helping people, or any of uncounted other roleplay preferences. So they strip away at least some of the points gained for killing monsters and getting treasure, and instead give them for performing the desired conduct, whatever it is. Now a player character gains experience points by helping the poor, or pursuing his private hobbies; these points increase his level--which makes him better at killing monsters and getting treasure. The rewards are now given for one sort of play, but they still facilitate the other.

It's not necessary to have a rewards system in a game. Stripped of such artificial rewards, many players will discover that play is its own reward. After all, players play because they enjoy the game. They enjoy different aspects of the game, but whatever it is that they enjoy is inherent in the play itself. Rewards systems, in the main, are icing on the cake. Done right, they encourage the desired form of play. Done wrong, they can clash horribly with the entire game.

Thus when you design a rewards system, you need to look at both sides of it. What does this reward, that is, what does a player have to do to receive the reward? Winning, exploring the theme, and discovering the world are all goals and in a sense rewarding conduct; if you wish to encourage one of those, that is what you reward. You must then also ask what the reward facilitates. Does it make the character more powerful, give the player greater ability to address the theme, open up new areas of exploration?

Rewards systems, when they exist, are usually tied into character improvement. Thus if you've solved the one you've often solved at least part of the other. It need not be that way; you can provide rewards that advance player goals in one fashion and advancement that does so in another. For example, you could have a gamist game in which character advancement was built on improving skills by use, such that each time the player brought a particular skill into play in a significant way he received credit toward improving that skill. Independent of that, you could reward gamist play with success points, a small pool of points or dice on which the player could draw when he wished to improve his odds against a more daunting challenge or in a moment when success was more important. Rewards do not have to be tied to character improvement, even if character improvement is well designed for the goals of the game. Games with no character improvement at all may still have effective and functional rewards systems which facilitate the desired mode of play.


With sufficient consideration to what a game is trying to achieve, GNS theory can be very instructive in how best to achieve it. It does not dictate solutions to all of the questions that a designer must ask, but it does inform him of questions he needs to address which he might otherwise miss.

In examining character generation, resolution mechanics, credibility distribution, advancement, and rewards, it was shown that there were some ways in which GNS theory could point us to the best solutions for the type of game we sought to build. It is clear that at times designers are asking the wrong questions in these areas, because some of the things which we expect would matter are not relevant, but others that we often overlook are significant.

Although these five areas of game design are a significant portion of most games, they are not a complete consideration of all the areas which matter in all games. It is hoped that the consideration of these areas will not merely help the designer see that GNS can provide guidance on these game design issues, but also enable him to find the right questions and answers in areas not covered here.

I look forward to seeing the application of the theory to more games in the future.

M. Joseph Young is co-creator of the Multiverser role playing game and author or co-author of its various supplements. His Internet writings are indexed for convenience. He is available to discuss these ideas through the Forge forums and by e-mail.

The author wishes to thank Ron Edwards, Mike Holmes, Clinton Nixon, Ryan Young, Fang Langford, and Ralph Mazza for their editorial suggestions on the draft of this article. To recognize all those whose contributions were made through discussions on the forums of this site and others would require a separate article; please accept my thanks.

Similarly, there have been uncounted forum posts here and elsewhere that have contributed to the author's understanding of these issues. It has been wisely suggested that at least some of these be linked; alas, there are again more than can be acknowledged. Two stand out, however, as expanding on specific areas covered in the article, and in both of them the author here has made comments there which he hopes are of value. The concept of credibility appears to have been introduced by Vincent in Vincent's Standard Rant: Power, Credibility, and Assent; this author's comments on the top of the second page and near the bottom of the third page may be helpful in elucidating the use of Credibility in this context, and there is much on the thread that is useful. It appears that the earliest suggestion of the two pronged nature of reward systems was in this author's post, the second, in GNS and Player Rewards. The post illustrates by examples that games do not need reward mechanics for players to be rewarded, as play can and is often its own reward.

The Forge moderated by Ron Edwards and administrated by Vincent Baker.
All articles, reviews, and posts on this site are copyright their designated author.