Advancement

Players advance in intangible ways, through the knowledge they gain and the friends they make. Characters also can develop intangibly to some degree, in that they exist partially in the minds of the players who encounter them. However, characters are primarily a tangible phenomenon: They are defined by code and data. The virtual world determines what they can and cannot do and how mobiles react to them. When a character “improves,” the improvement is measured in terms of tangible factors rather than intangible ones.

Characters don’t have to advance, of course; for non-game worlds, the aim is often for players to gain or hone skills, therefore the success or otherwise of a character is entirely dependent on what players learn and how they apply this knowledge. Still, though, it’s often the case that players will respond better if there is a reward in the form of some tangible improvement to their character, even if it’s of only metaphorical significance.

For those that do offer advancement, there are basically two approaches: single or multiple. In a single advancement system, one process completely dominates; in a multiple advancement system, characters can improve along several dimensions. At first glance, it looks like no contest: Why would you want single advancement when multiple advancement offers so much choice?

There are two reasons why. Firstly, single advancement is focused and unambiguous. Players know what they have to do, and can therefore set about doing it. In a multiple advancement system, players might have to make choices before being fully informed; they may therefore make mistakes. The second reason is that multiple advancement suggests multiple content, which is more expensive to produce and harder to balance.

As an example, consider a hypothetical virtual world set in Three Musketeers ¹ France. In a multiple advancement system, characters can acquire “military points,” “romance points,” “politics points,” and “fame points.” Thus, they can concentrate on acquiring points of a particular kind to advance their careers.

If more people want to pursue one career than there is content to support, new content must be added for it (and the content that already exists for other careers will be relatively underused). Players who start their characters off collecting military points only to find that the romance approach suits them better will feel that their time has been wasted. A single advancement system, where characters collect “social points” that are an abstraction of the other four, would allow characters to switch focus without suffering, and if content of one type was running low they could overflow into another until the first recovered.

Multiple advancement wins out when advancement involves different mechanisms. If military rank depended on campaign experience and political rank depended on electoral influence, the two could not be reconciled so easily.

Virtual worlds measure advancement in terms of advantageous changes to tangible data. In considering advancement, it therefore makes sense to begin by looking at what can be advanced.

Attributes

Attributes are the fundamental properties that define a character. Because they are tangible, they should properly only rate physical qualities, such as the character’s strength and manual dexterity. However, under the influence of table-top role-playing games they often include non-physical concepts, such as intelligence and wisdom. This just about makes sense to a hard role-player, who can work with “My character is stupid,” but it has its limits: “My character is a genius” would for most players entail pretending to be cleverer than they really are, which is not quite so simple. For most virtual worlds, though, these “mental attributes” are merely convenient labels: Intelligence is used primarily as a fiction to determine how much magic a character can use, and wisdom serves the same function for priestly types.

Attributes also are meant to be relatively fixed: If you have a poor constitution, tough luck, you just have to live with it. Aging effects could perhaps make a difference (older characters might gain wisdom, but lose strength), but on the whole attributes aren’t supposed to change much, thereby preserving basic differences between characters. There has been increasing slippage in this over the years, however, with the old notion that attributes should indicate intrinsic abilities increasingly falling out of favor. Instead, attributes can be improved on by action or training, which makes them more like skills (discussed shortly). There are several reasons for this, having to do with fairness, motivation, and choice.

The result is that after a while character attributes ¹ become fairly homogenized: In the real world, people can be any combination of good-looking/ugly and strong/weak, but in virtual worlds they usually have to start off more one than the other (for balance). They won’t end up being high in both. It may take considerably more effort to raise an originally low attribute to a high value than it does to raise an originally high attribute to it, but that doesn’t present a total barrier. Normally, one stat is raised relentlessly as far as possible first, with the lower ones being neglected until it’s their turn; this leads to a world packed with puny geniuses and muscle-bound simpletons.

The problem with attributes is that they do make sense in real terms, but this isn’t a sense that many players are willing to accept. Some people are simply never, ever going to be as strong or as fast or as tall or as fit or as good-looking as others no matter what they do, but part of the attraction of virtual worlds is that these physical limitations don’t apply. In design terms, fixed or ever-rising attributes are also undesirable: If you decide after some time that you want your character to become a ballet dancer but you built her like an ox so she could become a boxer, well, ballet dancing is out of the question.

For this reason, the trend is to treat attributes as little more than symbols during character creation, so that players have to think about their characters as they individualize them. This gives characters direction, but only in the minds of their creators; in practice, the relative differences between a character’s attributes are not usually all that important in the great scheme of things, especially if values are regularly boosted as a reward for achievement ².

The idea of attributes as important, player-defining properties is probably here to stay. However, whether they remain relatively fixed (as they do for most other objects) or they rise ³ over time (as do skills) remains to be seen.

Levels

In the development of virtual worlds, levels came first (as they also did in table-top gaming). Levels are the classic advancement system—easy to implement and to manage—and are based on the concept of goals and rewards. Characters undertake actions, for which they receive a payment in experience points; because the notion of experience is general enough to apply to pretty well any activity, this approach has wide applicability. When a character’s experience points reach a certain threshold, the player rises in (ability) level. This will typically involve things such as attribute boosts, skill boosts, resistance boosts, greater choices, and higher status (among NPCs and, often, PCs).

Looking at this analytically, boosting is implemented at times dictated by a step function. Why not use a more continuous approach? Instead of levels, raw experience points could be used. By using a finer grain for things such as attributes and NPC attitudes, a character who is one experience point below the threshold for level X wouldn’t have exactly the same stats as a character exactly on the threshold for level X-1. In pencil-and-paper games, where humans do (or used to do) the bookkeeping, this is reasonable—keeping separate track of these things is just too tedious. Computers, though, excel at tedious, so why the need for levels?

The answer is that levels provide goals. They give players something to shoot for, so that their actions have more purpose. They give players a sense that their achievements do have effects. Levels use a step function because players like to feel they have taken a step.

The time taken to reach a new goal ¹ must be appropriate, of course. If characters advance too quickly, levels effectively merge into a stream and offer no more bang than experience points. If levels are too far apart, players become disenchanted that their efforts rarely seem to get them anywhere. Note that although the experience points separating two fixed levels are invariant (at least within character classes), the time taken to acquire the necessary points will depend on two things—what the character does, and luck. Thus, characters do not level ² with monotonous regularity, but with some unpredictability. This leads to what is known as a variable ratio reinforcement schedule, which behavioral psychologists have shown to be the most effective there is for maintaining a high, constant rate of play ³—Las Vegas slot machines are designed this way to keep people pumping them full of money. The morality of using such an addictive system for virtual worlds is discussed in Chapter 8, “Coda: Ethical Considerations.”

Another reason for grouping together several rewards is that this enhances the overall experience; the bulk adds value. Damming a stream and then releasing it has more of an impact than would letting it run freely, whether the stream is of water or of presents.

The relationship between levels and experience points varies, depending on the virtual world. However, it’s almost always exponential rather than linear. MUD1 used a doubling scheme, whereby the points needed to rise from level X to level X +1 totaled the same as the points needed to get from level zero to level X ⁴. Most modern virtual worlds start off approximating a doubling regime, but gradually tend toward a more linear scale. Compare the examples ⁵ from Asheron’s Call shown in Table 5.1.

Table 5.1 Asheron’s Call Level System

A slowing curve enables more levels to be packed into the same range of points, which increases the rate of reinforcement for players of experienced characters. Even so, some virtual worlds allow characters to have so many points that they record scores using a floating-point representation. This kind of level inflation can get out of hand (in worlds without permanent character death, anyway), so there’s often a maximum level beyond which characters cannot proceed. Some virtual worlds allow remorting at this level, which means a character gets to keep its abilities but must start back at level zero as a different (sometimes more powerful) class ⁶. Other approaches are relative levels and ranking systems (“you are 386 th out of 4,034 characters”).

One of the problems with levels (for VWs which make use of them) is that content has to be available for characters of whatever experience. If one character needs a million points to go up a level, they’ll be doing different things to the other character who only needs a thousand ⁷. This means that characters which are of markedly different levels can rarely play together—the lower character will find high-end content too difficult and the higher character will find low-end content insignificant.

Another issue concerns where points come from. Some of the most intense experiences of players result in low overall gains—or even losses—of their characters’ points. If some monster beats you up, but you get healed then go back again and again until you beat the sucker, your character may be worse for it but you, as a player, may be better. Likewise, running around mindlessly collecting easy quest points is going to make your character improve, but it’ll do nothing but bore you as a person. Ideally, points should be awarded for things you find exciting, rather than as some kind of compensation for being fed up.

I’ve been describing levels here as a single-advancement abstraction of other progression measures. It’s possible to formalize this, with level changes being determined using a formula that accounts for a range of multiple-advancement schemes. Thus, a player may suspect from experience that they’re close to going up a level, but don’t know how much they need to improve their swimming skill or their dart-throwing skill to do so. This uncertainty makes a variable ratio reinforcement schedule even more powerful, and the morality of using it consequently even less clear.

Hybrid schemes are also being developed. Neocron ⁸ for example, has typed experience points: When your character is awarded experience points for doing something, the points go to a pool associated with one of five stats. When the pool fills, you go up a level in that stat only, and gain skill points that can be spent on skills related to that stat only. It’s a cross between a level system and a class system.

Designers are thus increasingly becoming aware that the benefits of levels can be used in non-traditional areas. In particular, applying a level system to individual skills can be very fruitful, assuming some provision is made for achiever-type characters to compare themselves against others (and even that may prove unnecessary). Skill systems are present even in many single-advancement, experience-based virtual worlds, and could eventually take over from them as the predominant measure of a character.

However, skill systems are not without their problems.

Skills

In virtual worlds, skills are tangible representations of the abilities that characters have (supposedly) learned.

In real life, a skill is a related series of actions that a person has internalized to the extent that they can apply it “without thinking;” skills are either innate or learned through practice. In virtual worlds, the notion of what a skill is remains the same, but skill acquisition is often only tenuously linked to “practice makes perfect.” In particular, some virtual worlds allow characters to improve skills when they go up a level, irrespective of whether the skills they choose to improve are ones they’ve actually used.

Why do virtual worlds need to make skills tangible? Actually, they don’t: Character skills are very useful as a means for players to individualize their characters, and they provide a good delivery mechanism for achievement, but they aren’t strictly necessary. The most important skills (like it or not) are those of the player, not of the character, and the virtual world can’t hope to rate these ¹.

Skills are interpreted by the virtual world in a number of ways, depending on the nature of the skill, of any tools required, and of the virtual world. The main four are

• Binary skills (feats ²). Either you can swim or you can’t.

• Discrete skills (skill levels). People can swim to an ability described by one of a fixed number (greater than two) of predefined levels.

• Continuous skills. People can swim with widely varying degrees of ability, from not at all to completely competent.

• Binary-continuous skills. Either you can’t swim, or you can swim with widely varying degrees of ability.

The idea behind a binary skill is simple command guarding. If you have the skill, you get to use the command, and if you don’t, then you don’t. Discrete skills take this a step further, with increasing levels opening up new commands (swimming underwater, swimming in stronger currents).

Continuous skills work slightly different in that everyone has a shot at performing any command. Sometimes, the effects are also continuous (the higher your swimming skill, the faster you can swim), but often it’s used as a form of binary skill based on the difficulty of the task at hand. For example, if swimming across a moat is rated at difficulty 30, then someone with a swimming skill of 30 or more will always succeed; someone with a swimming skill of 25 will fail some of the time, but not as often as would someone with a swimming skill of 10 ³. Binary-continuous skills work the same way, but with the proviso that if your skill rating in the area is zero then you can’t even attempt it (this is in addition to any other applicable preconditions—for example, armless characters don’t get to swim no matter how much they practice).

Skills can be active or passive. Passive skills are switched on the entire time, whereas active ones must be called into play explicitly. For example, if a character has a “deflect arrows” feat then that would be passive—the character automatically attempts to deflect any arrows fired at them. A feat such as “berserk fury,” on the other hand, might be active: Players have to decide when to sacrifice defense for offense, it won’t happen on its own. Most skills are passive, but only come into use as the result of an active decision: Your swimming skill is effectively dormant until you enter a body of water, but while there it’ll be taken into account passively. Because putting active and passive skills in the same list can be confusing for players, there is a growing tendency to refer to active skills as powers rather than skills.

Skill Organizations

The “skill levels” approach is an example of a linear skill progression. To acquire a specialized skill, you have to learn its parent (or prerequisite) skill first. Thus, you can’t learn, say, “diving,” unless you already know “swimming.”

A skill hierarchy takes this idea one step further and allows several specialized skills to be learned following the acquisition of a general skill. For example, having learned a “single-handed weapon” skill, a character may then learn “sword,” “axe,” or “club,” and from there further specialize into particular weapons (longsword, shortsword, rapier, cutlass, and so on).

A skill network is like a skill hierarchy except that it allows for skills to have two or more other skills as preconditions. You can’t learn the skill of “horse archery” unless you already have “horse riding” and “archery” skills; you can’t learn “espionage” unless you have at least three from “stealth,” “disguise,” “bluff,” and “hand-to-hand combat.”

Looking at skills this way, it’s easy to see how attributes came to be regarded as the same thing. “You can’t be a blacksmith unless you’re strong” is just like “You can’t be a pearl diver unless you can swim.” A concept such as “strength” becomes the concept of “being able to use strength;” although this is ultimately dependent on a character’s physical prowess, it nevertheless offers enough leeway to allow for increasing the “attribute” as a device for character improvement.

Doing this has certain implications for skill systems, though. Ideally, all skills should be implemented using the same basic mechanism, so as not to distract the players. If two or more skill systems run in parallel, they must be disambiguated somehow (this is why feats get a separate name). When attributes are handled just like skills, this means that skills and attributes should “look” the same. In other words, if attributes are on an integer scale then they should all be on the same scale and it should be shared by skills.

In practice, this actually works quite well. There are some awkward intrinsic properties of characters that don’t fit naturally into the scheme (the number of genders is fixed, for example) but they can usually be shoehorned in without too much bother. This isn’t to say that there aren’t better ways to do it, of course; however, it does mean there isn’t too much incentive at the moment to do it some other way. As usual, as long as designers understand why things are the way they are, they can make informed judgments as to whether to keep them that way.

It’s worth considering whether skills should be applicable to objects other than player characters. They can quite obviously be extended to NPCs and even some monsters, but what about objects in general? Could not a shield’s ability to deflect arrows be implemented using the same skill that a character might use to deflect them? Or more frequently, shouldn’t a fireball wand simply cast the spell? Skill transference like this is particularly useful for virtual worlds that use magic, where a wand could be given the same spell-casting skill as a player or a player given the same petrifying gaze skills as a basilisk. Could a sword improve, through use, to become better? Could it be trained (that is, enchanted) to learn new skills? If object qualities are implemented using the same system as player character qualities, at least these are options (albeit not necessarily ones every designer will want to pursue!).

Skill Sets

The biggest problem with skills lies in deciding which ones your virtual world should have. How many should there be in total? Is 100 skills too many? Too few? From whose point of view—newbies or oldbies?

We should begin by determining what is meant by a “skill.” Most virtual worlds would admit derivatives of crafting—making or repairing things in the virtual world—as qualifying as skills. Some would include weapon skills, and others would also add spells. The disparity is in part due to the different mechanisms by which acquisition works: If you get crafting skills by collecting “skill points” for going up levels but you get weapons skills by buying “training,” the two are distinct. It would be confusing to call them both “skills,” so commonly some other term is used instead (for example, “weapon specializations”).

Ideally, all forms of skills would use the same mechanism. Why should the system by which you learn to make a barrel be any different from the one by which you learn to swing an axe, to brew a potion, or to let rip a fireball? If they are fundamentally different even at abstract levels, that’s fair enough; I wouldn’t necessarily advocate implementing equipment as skills, for example. However, if they’re the same but are implemented differently, perhaps the reasons for that different implementation should be examined.

Spells, for example, are often linked to a character’s level. When you reach level 43, suddenly you gain the ability to cast sidewinder fireballs in addition to regular ones. The spells you could cast at level 42 you can still cast, but with slightly more efficacy. If this approach is good enough for spells, why not use it for crafts? When you reach level 43, suddenly you gain the ability to make mithril horseshoes in addition to regular ones. Alternatively, if you’re using a “skill points” system for gaining new craft and/or weapon skills, why not apply it to spells?

Using the same skill system for crafting, weapon use, spells, and everything else (acrobatics, horse-riding, speaking Elvish) doesn’t mean they can’t be radically different in effect, of course. However, it also doesn’t mean that the problems of deciding what skills to have go away.

Typically, however many different crafting skills, weapon skills, spells, and general-purpose skills a virtual world offers, a handful will be so essential that everyone has them, most will be secondary skills that people have because they had to choose something and another handful will be utterly useless. In a Wild West world, for example, only people who actively like building unusual characters are going to pass on the horse-riding skill.

If a skill is considered by all players to be essential, all characters should be given it anyway. Characters don’t usually have a “walk” skill, because it’s a skill everyone can be assumed to have. As a rule of thumb, if any one skill is adopted by more than 90% of characters ¹, they all ought to get it. Similarly, if a skill is unused it should also be given to everyone (or derived from attributes)—it costs nothing to do so, and it means the remaining skills should be of roughly equal value to players. Of course, it may be that having done all this you find that a new handful of must-have skills emerges. If this happens, either repeat the exercise or split the skills into disjoint sub-skills.

It’s not too difficult to hardwire a universal skill once it’s been identified (assuming you do so before the virtual world leaves beta test). It’s quite tricky to decide what should be skills in the first place, though. Should “horse training” be a skill? Is it different from “dog training,” or are they both examples of “animal training?” Should there be a skill for tree-climbing, or are so few people going to want to do this that it’ll be a waste of time implementing it? Should there be some weighting scheme so that unpopular skills cost less to learn than popular ones ²? Should some skills be more difficult than others anyway—isn’t learning a foreign language comparatively harder than learning to ride a bicycle?

There are, unfortunately, no clear-cut answers. My advice when deciding what skill set to have is to consider what message it sends to players. I’ve seen designs for virtual worlds that have a weapons skill tree five levels deep, yet which had a single “mining” skill that worked as well for iron as it did for diamonds. The message they were sending was, “This virtual world is all about combat.” Similarly, I’ve seen designs that covered the minutiae of everyday life, with skills covering everything from corn-planting to calligraphy to bee-keeping to savoir-faire ³. The message here was, “Nothing is all that important.”

If you want crafting to be a viable career, the skill tree for crafting should be deep enough for people to take it seriously. If you want people to travel a lot, include several skills that make traveling easier. If you think players will want to play bardic characters, include bardic skills. If you don’t want to attract sneaky-thievy characters, don’t have sneaky-thievy skills. Rate each skill for its attractiveness to various player types; if any types are over- or under-represented, adjust accordingly. If you fear spreadsheet abusers, make key skills mutually exclusive (but be aware that this will create a de facto character class system).

Although the number of skills of a particular kind available to characters is one way to send players a message, it’s not alone. In general, skills should be of roughly the same level of abstraction as one other. If in one area they are represented in greater detail than in others, this also can point to an increased sense of importance. Were the “swordplay” skill to be described in terms of how much it increases the damage done, critical hit percentages, weapon recovery time, parrying capability, and resistance to being disarmed, this would give it more significance than a “lumberjack” skill defined solely in terms of how quickly you can fell a tree. Crediting skills with a false sense of importance this way is a particular example of selective depth, a trap that often ensnares the designers of virtual worlds.

As with objects, skills don’t have to have a tangible purpose. If characters need an “interior decoration” skill to change the wallpaper in their houses, then if they want to change the way their house looks they can either acquire this skill themselves or go to someone who has already acquired it. Either way, the result is a matter of aesthetics: The virtual world doesn’t necessarily ascribe any meaning to how the inside of your house looks. Of course, also as with objects, it doesn’t necessarily not ascribe any meaning to it either. Unless you want to make some kind of “we have some skills just for the fun of it” point, if you can find a way to give tangible value to skills, you should do so. Having your house redecorated to a higher standard could win you the envy of NPCs, say, or reduce the chances of its being infested with termites.

Skills that can be used for more than one thing are generally good to have but difficult to contrive. If you can use an axe in combat, does that qualify you to use one as a lumberjack? Or is that a different kind of axe skill? If they’re different, how many people are going to bother with a lumberjack axe skill? If they’re the same, characters will become lumberjacks so they can increase their axe skills while making money. Multipurpose skills have more justification for existing than do single-purpose skills, but they also can have annoying consequences.

Caps

Levels and (in most implementations) skills operate on a linear scale. A character starts at 0 and improves incrementally. A level 22 necromancer isn’t as powerful as a level 55 necromancer, but would be roughly on a par with a level 22 monk or assassin or whatever and much better than a level 11 necromancer ¹. So what’s the highest value that a level or skill can have?

If you have no limit, characters will improve relentlessly, getting more and more powerful until newbies are little more than insects by comparison. Once-daunting content will appear trivial; players will demand more to satisfy their appetites.

For this reason, many virtual worlds have caps—absolute limits on how high a level or skill can go. For level caps, this stops characters from getting too far ahead of what the virtual world can offer as content; however, it can lead to frustration as there’s nothing to do except wait for other people to catch up in sufficient numbers that new content will be added. For skill caps, players will max out their characters in all the important skills so they still end up ultra-powerful.

The only solution that is truly guaranteed to stop this from happening is permanent death (PD), but this is often taboo—especially in commercial virtual worlds. Some designers have mooted the idea of having PD as the price for cap-busting (if you want to go off the scale you can do, but your character can then die permanently), a solution that has been implemented many times already in non-commercial worlds. The uproar from paying players who want the skills but don’t want their characters to die would likely be considerable, however. Players offered Achilles’ choice between a short life of glory or a long life of obscurity will want a long life of glory.

Without PD, either characters are going to become all-powerful or (achiever) players are going to become frustrated. That really is the bottom line. It’s possible to relieve these symptoms without resorting to PD, though. One way is to have an advancement curve that makes it increasingly difficult to progress to new levels, but not actually impossible. As long as the powers at level N are tangibly better than those at level N-1, players will want to have them. Eventually, though, an effective cap will emerge beyond which few characters will ever rise. Taking the caps off might mean someone can increase their archery skill to 134 instead of being stuck at 100, but they’re never going to reach 1,000. It’s a bit like world records for athletics: They’re always falling, but there has to be a limit. No one is going to run the 100 meters in 7 seconds unless the authorities change the rules currently disallowing the use of motor bikes. Steep advancement curves like this can work, but they really only ever delay the inevitable: Sooner or later, players decide that the treadmill is simply no longer fun.

An approach pioneered by Ultima Online was to make skills atrophy. Characters can learn several skills, but the more they learn then the less attention they give to other skills, which lose their values. This allows for characters to excel in only a few skills at once, while retaining some scope for changing what those skills are. Players can remain powerful, but can use different skills as ways to express their power. It sounds good, but unfortunately players don’t like it: If they learn a skill, they believe that they should always have that skill. They do have a point: It’s not like an English-speaker who learns French will forget the French if they then learn Spanish; indeed, a knowledge of French could help you learn Spanish. It’s true that if you become unpracticed at something, then it’ll take you a short while to get back up to speed, but this will be nowhere near as long as the first time you learned it. Being rusty at something is different than completely forgetting all about it.

The first-century Greek philosopher Plutarch said, “The mind is not a vessel to be filled but a fire to be kindled.” ² Atrophy systems embody the “vessel to be filled” approach: You can only learn so much; if you’re all learned out, you have to unlearn something to make room for new stuff. Although UO players weren’t exactly happy about skill atrophy as a principle, what really annoyed them was the fact that they couldn’t state which skill they wanted to atrophy next. Star Wars Galaxies uses basically the same idea, except that atrophy doesn’t occur automatically: If you want to recover points from one skill to spend on a different one, you have to do it explicitly. Skills you already know cost less to advance in than others, so the temptation is to specialize rather than take the jack-of-all-trades route.

“Vessel to be filled” approaches cap the totality of skills a character can have while leaving the distribution of individual skills relatively free. A more “fire to be kindled” way of doing it is to cap earlier in the process—that is, limit the amount of time that players can (or are willing to) spend in a virtual world (but not so much that they don’t feel it’s worth their investment). Ultima Online also pioneered this method ³, although not until 2000. The idea behind UO’s power hour is that advancement happens faster during a character’s first hour of play each day, but drops off thereafter. Interestingly, the advancement rate to which it falls is the same as it was before power hour was introduced, but players are reluctant to rack up points in it because they’re “paid” less for their efforts than in power hour. Consequently, whereas the average time players spent in UO before power hour was introduced was three to four hours a day, afterward it dropped to around one hour a day. The result was that although characters advanced faster per game hour, they advanced slower per real hour and therefore took much longer in real terms to rise than they had done previously.

This approach worked for UO, but it can only be taken so far: If a character that was played four times as much as another character could only ever be 110% as good as it, for example, that would certainly lead to the loss of many long-term players. It has other disadvantages, too, in particular it discourages players from adventuring in groups and it isn’t easy to make remotely realistic within the virtual world’s fiction.

Skill Improvement

How do characters improve their skills? To preserve the notion of what a skill is, there ought to be some relationship between what you do and how your skills improve; but what?

If skills are rated using a points system (as most are), the question becomes one of deciding when to increment a skill by a point. There are basically two ways to do it:

• Random chance. Each time a skill-enhancement opportunity arises, there is a small chance of causing a skill point increment. This chance may increase over time.

• Cumulative chance. Each time a skill-enhancement opportunity arises, a random number of pebbles is thrown into a bin. When the bin is full, the skill point is incremented.

The advantage of the random approach is that you can gain a second skill point shortly after having gained a first, if your luck is in. The disadvantage is that you may never gain a skill point while your luck is out, and you will complain loudly about this. The cumulative chance approach does have some randomness, so you can’t ever be sure when you’ll gain a skill point (unless you can see the numbers); however, you can be absolutely certain that it won’t be immediately after having gained one.

Where skill nets exist, it may be possible to arrange for some skills to support others. It’s not a precondition of veterinary science that you know human anatomy; however, if you do know human anatomy, then you could expect to pick up veterinary skills a little faster. This can be managed by increasing the random chance of success or by preloading the bin with pebbles so that fewer new ones are required to fill it ¹.

It’s in deciding what qualifies as a “skill enhancement opportunity” that things start to get interesting. There are several ways to do it, none of which are entirely satisfactory.

The traditional method is improve from success. Each task requiring skill use has a difficulty rating; when a character attempts the task, his skill rating is compared against the difficulty rating to see if he succeeds. If he does, it counts as a skill enhancement opportunity; otherwise, well, he should maybe try something less difficult next time.

A variant of this is improve from failure. The theory here is that people learn from their mistakes, not from their successes, therefore characters should get a skill enhancement opportunity for coming close to success but not actually succeeding. If they do succeed, well, they’re happy anyway because they performed the task; this way, they’re rewarded for nearly succeeding too. If they try something so hard that they don’t even come close, well, maybe they should have attempted something easier. This is my own favorite method, by the way, but it’s not for everyone. It can produce inappropriate messages, too: You don’t really want to know that your character’s shield-use skill has gone up if the blow it only just failed to stop was fatal.

Many virtual worlds use improve from leveling. Although some increase in a skill may occur through use, the idea here is that experience points are a sufficiently good reflection of player activity that skill acquisition can be linked directly to character level. When you go up a level, you (may) get skill points to spend on improving yourself (also known by its fiction name, training). This approach has three advantages: Players look forward to going up levels more; it’s much easier to balance for skills that are used with wildly different frequencies (fighters swing swords more often than tailors make robes); characters don’t have to waste their time repeatedly shooting arrows at static targets, trying to learn from being bored. The disadvantage is that the skills that players decide to invest their skill points on may have no bearing on what they did to get the skill points in the first place—that is, it’s unrealistic. Some games try to limit you to only spend points on something you did in the last level (that is, you had to make a piece of clothing at least once to advance tailoring). Players will find and jump through these hoops very quickly.

Star Wars Galaxies uses a very novel approach that impressed designers because of the way it manages to kill several birds with one stone. The idea is that for craft skill points, the measure of success is whether anyone wants to buy your goods or not. Therefore, the more people use the products of your skill, the more skill points you get. This gives crafters a reason to produce the items that PCs want rather than the items NPCs will pay for. It helps stake-holding, it helps community, and it integrates crafters more fully into the economy. Players are as excited over the potential use of their skill as over the actual use of it. Presumably, there could be an improve by failure equivalent, whereby characters gained skill points by making items nobody wanted, but this has none of the advantages of the SWG original. Unfortunately, it does have one very important advantage that the improve by success version doesn’t: It makes sense. In real life, people buy a crafter’s goods because the crafter is good at making them, not the other way round.

There is a final skill advancement mechanism that isn’t used much at the moment but may become more important in the future: improvement from purchase ². If the virtual world’s business model and player base are right for it, characters could gain skills by simply paying for them. This was first done implicitly in the early 1990s by Avalon, which allocated skill improvement points to characters depending on how long their players stayed logged in (at a time when it charged by the hour). Achaea is more upfront about it: You pay real money for credits, which you can then turn into skill improvement points.

When skills can rise inexorably along a linear scale, it’s worth asking what the values mean. Is a character with a strength of 100 five times stronger than someone with a strength of 20? Unlike the case with character levels, it could well be possible to measure this directly—for example, by considering the maximum weight that the virtual world lets a character carry. If something approximating a proportional relationship is established, this raises a number of issues:

• If the principle is established for one skill, should it not apply to all skills?

• Is it realistic to have skill disparities as large as this?

• Are players agonizing at character-creation over whether to give their character a strength of 15, 16, 17, or 18 simply wasting their time thinking about what’s basically just noise?

That mention of player design brings us conveniently to our next topic.