Persistent Worlds

A game, as we defined it earlier, is a contest with rules and a winner. However, a good many online "games" are not really games at all. Asheron's Call , EverQuest , Anarchy Online , and so on are actually persistent worlds: permanent environments that players can play in, retaining the state of their avatar from one session to another. Persistent worlds have a number of special problems and design requirements, which we discuss here.

The Origins of Persistent-World Gaming

Persistent worlds have been around for much longer than the graphical MMORPGs that are so popular right now. Since 1978, a small but dedicated community of developers has been building, playing, and studying text-based persistent worlds called MUDs (multi- user dungeons or domains, depending on who you talk to) that could be played by groups of people over the Internet. In these worlds, everything was accomplished by typing commands, and a rich culture of online role playing evolved. MUDs have since fragmented into subcategories : MUSHes (multi-user shared hallucinations, dedicated primarily to personal interaction rather than combat), and MOOs (MUDs, object-oriented, in which people can use a scripting language to design their own objects for use in the environment), and others even more esoteric.

We won't go into MUD design in any detail here; they're not a commercial market, and a vast amount of literature on them already is available on the Internet. However, many of the design problems of today's MMORPGs, particularly those relating to social interactions among players, were already solved ”or at least studied ”long ago in the MUD community.

How Persistent Worlds Differ from Games

Computer games appeal to players in a variety of ways. Part of the appeal is the environment in which the player finds herself: a fantasy world where magic really works, for example, or behind enemy lines in World War II. Another part is the role the gamers will play in the game: detective or pilot or knight - errant . Yet another is the gameplay itself, the nature of the challenges the player faces and the actions she may take to overcome them. And, of course, there is the goal of the game, its victory condition: to halt the enemy invasion, for example. The victory is usually the conclusion of a story that is told partly through narrative supplied by you, the designer, and partly through the player's own actions.

Story

Because persistent worlds have so many players, and because they are intended to continue indefinitely, the traditional narrative arc of a single-player game doesn't apply. You can't take the player by the hand and lead her through an experience of your designing. A persistent world is not a story, but a playground. It's up to you to build the environment and to give the players things that they may do, but how they spend their time is for them to decide.

You can still incorporate two elements from storytelling in a persistent world. One is the setting itself and the overall conditions of life in the world. It can be a dangerous place or a safe one, a rich place or a poor one, a tyrannical place or a democratic one. You can slowly change these conditions or introduce threats from the outside that challenge players to respond to them.

Elmqvist's Law: In an online game, players find it rewarding to save the world. They find it more rewarding to save the world together, with lots of other people.

The other element is a quest or errand that the player undertakes as an individual or collectively with others. These can be small-scale (eliminate the pack of wild dogs that has been marauding through the sheep flocks) or large-scale (everyone in the town gets together to rebuild the defenses in anticipation of an invasion). Most persistent worlds have large numbers of these quests available for players.

Storytelling versus simulation: If you write a static story (or, indeed, include any static element) in your game, everyone in the world will know how it ends in a matter of days. Mathematically, it is not possible for a design team to create stories fast enough to supply everyone playing. This is the traditional approach to this sort of game nonetheless. You can try a sim-style game that doesn't supply stories but that instead supplies freedom to make them. This is a lot harder and, arguably, has never been done successfully.

The best emergent stories occur in purely role-playing environments with almost no game-like elements, such as in MUSHes. These aren't profit-making enterprises , however, and they work because they have a small population of unusually dedicated players. As a designer, there's no way to force this to happen; it depends too much on the imagination and talent of the participants .

The Player's Role

In a single-player, plot-driven game, the player's role is defined by the actions he is allowed to take and is constrained by the requirements of the story. In a persistent world, the player doesn't have a single storyline to follow, so he needs a larger variety of things to do. Most persistent worlds are science fiction or fantasy role-playing environments, and they tend to define the choices available. Some of the early ones were designed like traditional single-player RPGs, in which the only role to play is that of mercenary adventurer, but thankfully that has changed somewhat in recent years .

As the designer, it is up to you to supply an assortment of possible roles the player may play and to make them meaningful in your world. You should also give the player the freedom to change her role (though not always easily or immediately) as she sees fit. Because the world continues indefinitely without coming to a narrative conclusion, you can't expect the player to want to play the same way forever. Just as people change careers and hobbies over time, players need to be able to change roles.

Gameplay

Finally, there's the question of the gameplay. Without a victory condition, you have to decide what challenges the player does face and what kinds of things she can achieve. Most persistent worlds are designed like role-playing games, and the player's objective is to advance her character. This is (usually) accomplished by fighting AI-controlled "monsters." However, there are many other things she might attain as well: wealth, political power, fame (or notoriety), and so on.

If your game is narrow, it will fail: Your game design must be expansive. Even the coolest game mechanic becomes tiresome after a time. You have to supply alternate ways of playing or alternate ways of experiencing the world. Otherwise, the players will go to another world where they can have new experiences. This means new additions or, better yet, completely different subgames embedded in the actual game.

In a single-player game, the player is, to some extent, trying to read the designer's mind, to figure out what you want him to do and then do it. His play is often reactive, a response to challenges thrown at him. In a persistent world, the player is deciding for himself what he wants to do. He seeks out challenges if he feels like it, but he can spend all his time socializing if he prefers. His gameplay ”and, indeed, the entire nature of the experience ”is expressive and active rather than reactive. Role playing, especially when it's taken seriously by the players, has qualities in common with live theater and improvisation. This characteristic of persistent-world play has profound effects on their design, as you will see later in this section.

Schubert's Law of Player Expectations: A new player's expectations of a virtual world are driven by his expectations of single-player games. In particular, he expects a narrow, predictable plot line with well-defined quests, carefully sculpted for himself as the hero. He also expects no interference or disruption from other players. These are difficult and sometimes impossible expectations for a virtual world to actually meet.

The Four Types of Players

In 1996, a MUD developer named Richard Bartle wrote a seminal article called "Hearts, Clubs, Diamonds, Spades: Players Who Suit MUDs" for the first issue of the Journal of MUD Research . He proposed that MUD players fall into four categories depending on whether they enjoy either acting on (manipulating, exploiting, or controlling) or interacting with (learning about and communicating with) either the world or the other players (see Figure 17.1). Those who enjoy acting on other players he dubbed Killers, or clubs; those who enjoy interacting with other players he called Socializers, or hearts. Those who enjoy acting on the world he described as Achievers, or diamonds; those who enjoy interacting with the world he referred to as Explorers, or spades.

Figure 17.1. Bartle's four types of players.

graphics/17fig01.gif

Bartle went on to claim that a healthy MUD community required a certain proportion of each of these types of players and that adjusting the game design to attract or discourage any given type of player would tend to influence the numbers of others as well. In effect, a persistent world is a sort of ecology in which the players' styles of play influence their population. Bartle's data was drawn from personal observation rather than rigorous statistical analysis, so his conclusions are certainly arguable. However, his grouping of player types has been found to be useful not only in the design of MUDs, but in graphical persistent worlds as well.

Creating an Avatar

As we said earlier, playing in persistent worlds is more than merely a form of gameplay; it's also a form of expression. The first thing a player does in joining a persistent world is to create an avatar, or character who represents her in the game, so this is one of the most expressive things she can do. The avatar is a mask she's going to wear throughout the game.

The avatar is defined by both tangible and intangible attributes (Koster's terminology). The tangible attributes are those that the game software recognizes and makes use of in gameplay ”physical attributes such as strength and speed, and mental attributes such as training and experience. The intangibles include some that are part of the software, cosmetic things such as clothing color , and others that are really defined by the way the player plays: vocabulary, prejudices, and attitudes.

If you're making an online role-playing game that includes traditional avatar attributes such as speed, strength, and so on, consult Chapter 11, "Role-Playing Games," for more information.

The Avatar's Profile

In addition to their tangible attributes associated with gameplay, players like to have a profile that identifies and describes them to other players. This profile can include such things as these:

Unique name or handle. Unless your game allows totally anonymous play, people will need some way of identifying their avatars by name. That way, a player's name can appear in documents, on leader boards, in chatrooms and bulletin boards , and so on.
Physical appearance. People clearly have to be able to tell one character from another on the screen. The physical appearance of avatars should be as customizable as you can afford to make it. Even if it has no impact on gameplay, players identify with and respond to physical appearances .
History or experience. This is simply a record of the player's achievements for others to see. It could be experience levels, quests undertaken, kills in battle, or any other accomplishments the player might be proud of. You'll have to decide whether players might want to keep some of these things private and, if so, whether they should be allowed to.
Reputation. Players can get positive or negative (or even multi-dimensional) reputations based on their behavior in the game. Some games use the reputation mechanism as a way of tagging players who frequently take advantage of others somehow. The reputation attribute warns other players "This person is dangerous" or "This person is trustworthy." Beware, however: This system is subject to abuse through collusion if you don't place limits on it.
Player autobiography. It's fun for a players to make up a history for his character, a background that will introduce him to others in the world. It's another form of self-expression. If children have access to your world, you will need to have a real person approve them for suitability.

You might or might not want to include important gameplay attributes in the player's public profile; it depends on how this affects the gameplay. If a player wants to hide his attributes from the world, is that a legitimate part of gameplay or is it an unfair advantage? ( Monopoly players are not allowed to hide their property cards under the table ”everyone is allowed to see what they own.)

Modes of expression: You're trying to provide as many modes of expression as possible in your online world. Character classes (a mechanism classifying characters according to their attitude toward the world) are just modes of expression, after all.

World Models

Unless your world is really only a chatroom, it's going to give players something to do. The things that you give them to do, and the rewards they earn for doing it, constitute the world model . We don't mean this in the genre sense of "killing monsters" versus "battling enemy spacecraft," but rather in the sense of how the world wants the player to live and grow in it.

Raph Koster has identified five classic world models, although there are undoubtedly more. Some include elements from more than one of these:

Scavenger model. Players collect things and return them to places of safety. The game is primarily a large treasure hunt, and they don't risk losing anything they've collected. Such games usually need to be reset to their initial state from time to time because players are hoarding too much stuff.
Social model. The world is primarily an expressive space. The fun comes from role playing in character; most forms of achievements are social achievement (political power, adulation, notoriety, and so on). Players rely primarily on intangible attributes ”their degree of leadership, charisma, or duplicity, for example. The normal tangible attributes, such as strength, are used as a basis for role playing rather than computer-managed combat.
Dungeons & Dragons model. This is the best-known model. The player is primarily in conflict with the environment, fighting NPCs for advancement and doing some scavenging along the way. It relies heavily on the tangible attributes of the avatar for its gameplay and includes an element of positive feedback: defeating enemies advances the character, which requires the game to offer tougher enemies next time. Such worlds tend to include quests as a form of narrative and a way of offering challenges to the players.
Player-versus-Player model. In this sort of world, players advance by defeating one another at contests of some kind, often characterized as combat. Players advance through a combination of their natural skill and rewards from winning battles . For this to work successfully, they need to be reasonably evenly matched; you can't have the old-timers beating up the newcomers all the time.
Builder model. This somewhat rare sort of world enables players to construct things and actually modify the world in which they play. It's a highly expressive form of entertainment. People get kudos not for their fighting skills, but for their aesthetic and architectural ones, both intangible qualities.

Avatar Death

In any persistent world that includes combat, you have to decide if it's possible for the player's avatar to die and what will happen if he does. Most role-playing games treat character death in certain standard ways; these are addressed in Chapter 11. As with other games, avatar death must be accompanied by a disincentive of some kind, or players won't care if they die. The trick is to find a disincentive that is appropriately proportional to the likelihood of their dying ”to put it in simpler terms, it's a balance problem. If the avatar can easily be killed through no fault of the player (such as through ignorance or bad luck), then the cost of dying ”the disincentive ”should be low. If the player really has to be stupid to get his avatar killed , the cost can be high.

Permanent Death

Here's the most extreme case: The avatar is destroyed and loses all property that he owns (in which case the system has to decide what happens to it). The player is forced to start over from scratch with a new avatar. This makes sense in games of short duration, but not in persistent worlds. Players put too much time and effort into building up their avatars to do this to them.

Resurrection with Reduced Attributes

Bringing the avatar back to life with reduced tangible attributes ”strength or skills, for example ”is a common way of penalizing death. In effect, it sets the player back a bit in her quest to grow a powerful avatar. Players find it irritating , of course, and it discourages risky play, but it makes a certain amount of sense. If you've just been beaten to death by a gang of club-wielding trolls , you ought to feel pretty lousy for a while when you come back to life!

Resurrection with Some Property Missing

Here's another classic disincentive for dying: When a player dies, he loses his money, gear, clothes, and other items that he has on him at the time. How much of it he loses and what becomes of it can vary considerably from game to game. You can also allow players to have a vault in the game where they can safely keep items that they're not carrying around with them, and these items can be there when their avatar is resurrected. You might as well include this feature because if you don't, the players will create a second character that they never play with, known in MMORPG parlance as a "mule," to hold their primary avatar's things for them.

The Player-Killer Problem

No aspect of the design of persistent worlds has been debated more than this one simple question: Should players' avatars be allowed to kill one another? We won't offer a definitive answer here, but instead we'll try to summarize some of the issues so that you can make an informed decision for your own game.

Most persistent worlds are designed as role-playing games in which players advance in skill and power through combat. It's generally more interesting if this combat occurs against another player rather than against an NPC, as Koster observes:

It is always more rewarding to kill other players than to kill whatever the game sets up as a target. A given player of level x can slay n creatures of level y . Therefore, killing a player of level x yields n x y reward in purely in-game reward terms. Killing players will therefore always be more rewarding in game terms than monsters of comparable difficulty. However, there's also the fact that players will be more challenging and exciting to fight than monsters no matter what you do.

The Ultima Online Experience

Ultima Online was initially designed in such a way that players could kill one another without restraint (except in towns), and the designers hoped that they would establish their own justice mechanisms within the game. Unfortunately, the world quickly began to resemble Afghanistan after the Soviets left: unremitting random violence, feuds, continual victimization of the weak by the strong, and petty warlords or gangs of bandits controlling areas of turf. No satisfactory solution arose from the players, partly because the software did not offer any genuinely painful punishment mechanisms for them to take advantage of. (In real life, we either lock murderers away for a very long time or kill them permanently, neither of which any for-pay persistent world can afford to do.) A variety of different automated mechanisms for encouraging justice were tried, but most were subject to some kind of exploitation. In the end, the developers threw up their hands and divided the world into "shards" with different rules for each. Some allowed player-versus-player (PvP) combat, and others did not. Approximately 80% of the players chose to play in non-PvP shards.

Justice Mechanisms

Koster offers the following summary of approaches to regulating PvP combat:

No automated regulation. Anyone can attack anyone , and only administrators or social mechanisms (vigilante justice) are available for dealing with rogue players. Koster estimates that as many as 40% of the potential audience will avoid this type of game because they don't like PvP.
Flagging of criminals. The server automatically detects criminal behavior and flags the criminals, who become fair game for others to attack. The system can also reduce the attributes of criminals, in effect penalizing them for their behavior. This can be used for thievery and other crimes as well as murder.
Reputation systems. This is similar to flagging, except that players decide when to "report" someone for criminal behavior and can choose not to do so. In practice, they almost always do, however.
"PK switch." Players can indicate whether they are willing to fight other players; those who are not are invulnerable to attack as well. This can also be used to give temporary consent for duels and arena-based combat. Unfortunately, this mechanism creates suspension-of-disbelief problems when wide-area spells are being used: Three PK players get roasted by a fireball, while an adjacent non-PK player is untouched because the fireball was cast by another PK player.
Safe games; no PvP allowed. This is the least troublesome solution, but even it is not without its hazards. Players will still find ways of abusing one another ”for example, by luring an unsuspecting newcomer into an area where he will be attacked by a monster. Koster estimates that this approach will cost you up to 20% of your potential audience, those who like PvP.

You can also divide the world into safe and dangerous geographic zones, but in practice, people tend to either stay in the safe zones or play near the edges, hoping to lure a potential victim over the line without his realizing it.

Faction-Based PvP

A solution that has become more common in recent years is to allow players to belong to factions. These can be as small as gangs or as large as entire nations at war. The rules enable players to attack members of enemy factions but not members of their own faction ”in effect, it's team play. Different regions are under the control of different factions, so players generally know which areas are safe and which are not. For the most part, this arrangement solves the random violence problem that plagued Ultima Online at first.

Star Wars Galaxies further extends the faction concept to allow players to be overt or covert members of a faction. Overtly declared members are at constant risk of being attacked by members of a hostile faction, so they receive certain advantages to compensate. Covert members don't run the risk but are not allowed to do certain things. Players can also be neutral.

The Bottom Line on Player Killing

You cannot please everybody, so you are really better off deciding who you want to please and tailoring your environment to them. A good many player-killers are cowards, undesirable players who enjoy exploiting their superior strength to victimize weaker ones without ever putting themselves at risk. You can't please them without also providing them with victims who won't be pleased, so it's not worth trying. Ultimately, you need to bear two things in mind:

It's a fantasy-world. That means it's supposed to be enjoyable, escapist entertainment. People don't fantasize about being harassed, bullied, or abused. A fair contest among consenting players is one thing; perpetual harassment or an ambush by a gang is quite another.
People are paying to play. This makes your world distinctly different from the real world. The real world doesn't owe us anything; we have to survive as best we can. But as a game provider, you've taken the players' money, so you have obligations to them. Just exactly what those obligations are is open to debate, of course, but if players don't feel that you are meeting them, they will leave.

Players have higher expectations of the virtual world than of the real world . For example, players will expect all labor to result in profit; they will expect life to be fair; they will expect to be protected from aggression before the fact, not just having to seek redress after the fact; they will expect problems to be resolved quickly; and they will expect that their integrity will be assumed to be beyond reproach. In other words, they will expect too much, and you will not be able to supply it all. The trick is to manage the expectations.

The Nature of Time

In a single-player computer game, you have a great deal of control over the relationship between "game time" and "real time." Most games run at many times the speed of real time, and a simulated "day" in a computer game often takes an hour or less of real time to experience. You can also hand over control of the speed of time to the player when you want to; it's not uncommon for players in combat flight simulators to speed up time when flying to and from the combat zones, and then slow it down to real time when they get there. Finally, you can skip time entirely during periods when the player's avatar is supposed to be asleep, for example. You can blank the screen for a moment and then put up a text message that says "8 hours later " and continue with the game.

None of these options is available in multi-player games. You obviously can't have some players moving through time at different speeds than others, and you can't skip time unless you somehow force all players to skip it together. Although game time might be faster than real time, it must proceed at the same fixed pace for everyone.

As a result, you must be careful about designing time-consuming activities. EverQuest , for example, has a mechanism called "meditation" in which players simply have to wait around for a while to restore their magic powers. There's no way to speed up this process ”it literally does involve waiting. Nor can they log out of the game while meditating and log back in again later to find that it's done. They can't even switch to a different process on their computer! Verant, the developers, eventually built in a mini-game for players to play while they were waiting, but this is clearly a patch, not a real solution. If your game contains features that are so boring that you have to distract the players, you need to rethink the features.

Time Is Irreversible

One of the key design considerations of single-player games is that they can be stopped and restarted at the player's discretion. If the player can save the game, he can essentially "reverse time" by going back to a previous point in the game and replaying it from there. This robs single-player games of the emotional impact of events: Anything that happens, good or bad, can be reversed by reloading a saved version of the game. You can design the game so that some events are inevitable, but, of course, the more of them there are, the less interactive the game is.

In an online game, time is irreversible. Even if there were a convenient way to do it, you can't reasonably ask all your players to agree to reverse time to an earlier point (although the managers of some persistent worlds have had to "roll back" to a saved state when the game got into problems). In the ordinary course of events, when a thing happens in an online game, it's done and can't be undone. It might be repaired, but it cannot be made as if it had never happened .

Persistent World Economies

If the players in a persistent world can collect and trade things of value, then the world has an economy. We discussed economies in some detail in Chapter 14, "Construction and Management Simulations." Economies are much easier to design and tune in a single-player game than they are in a persistent world. You can control the actions of a single person fairly strictly ; in a persistent world, you have thousands of people interacting in ways that you might not have anticipated.

Ultima Online was originally designed to have a completely self-contained, closed economy with a fixed number of resources that flowed around and around. For example, you could mine iron ore, smelt it into iron, and forge the iron into weapons. Using the weapons would cause them to deteriorate, and when they wore out, they would return to the pool of raw iron ore available for mining. This last step wasn't strictly realistic, but it did close the loop.

What the designers didn't count on was that players would hoard objects without using them. Because they didn't go back into the pool, the iron ore was quickly mined out, and as resources dwindled, inflation ran rampant. The players had literally cornered the market and could charge extortionate fees for anything. Eventually, Ultima Online was forced to adopt an open economy in which new resources are dumped in at intervals by the servers.

It's particularly essential in any economy that players not find a way to create something for nothing; that is, to return a resource back to the system for more than they paid for it in the first place. Otherwise, they'll find a way to automate this process and generate an unlimited stream of gold. Koster observes:

Online game economies are hard: A faucet/drain economy is one where you spawn new stuff, let it pool in the "sink" that is the game, and then have a concomitant drain. Players will hate having this drain, but if you do not enforce ongoing expenditures, you will have Monty Haul syndrome, infinite accumulation of wealth, an overall rise in the "standard of living" and capabilities of the average player, and thus produce imbalance in the game design and poor game longevity.

Final Thoughts on Persistent Worlds

Here are a few more of Raph Koster's observations on persistent worlds, to wind up this section:

The secrets to a really long-lived, goal-oriented, persistent world of wide appeal:

Have multiple paths of advancement (individual features are nice, but making them ladders is better).

Make it easy to switch between paths of advancement ( ideally , without having to start over).

Make sure the milestones in the path of advancement are clear and visible and significant (having 600 meaningless milestones doesn't help).

Ideally, make your game not have a sense of running out of significant milestones (try to make your ladder not feel finite).

Ownership is key: You have to give players a sense of ownership in the game. This is what will make them stay ”it is a " barrier to departure ." Social bonds are not enough because good social bonds extend outside the game. Instead, it is context. If they can build their own buildings , build a character, own possessions, hold down a job, feel a sense of responsibility to something that cannot be removed from the game ”then you have ownership.

J.C. Lawrence on utopias: Don't strive for perfection , strive for expressive fertility. You can't create utopia, and if you did nobody would want to live there.