Mücke's Musings on MMORPG Making

"Die Spieler machen das Spiel."

Newest Postings
Table of Contents
I can be contacted as hobold at this domain name.


Newer Postings


Tracking 01: Pacing Anomaly in Hard Mode Raids

This entry is meant to keep track of a posting on Blizzard's forums, just for convenience.


#23. Connection: Storytelling through Architecture

The title of this entry might remind you of the quote "talking about music is like dancing about architecture", but actually there is a strong connection between comics and dungeons. I will use the concepts and terminology of Scott McCloud's eye opening book "Understanding Comics". You don't have to read that in order to see the connection. But McCloud's book is an entertaining and enlightening read (actually a comic about the medium comic!) which I can heartily recommend to anyone who likes comics and to anyone who doesn't.

A basic comic is a sequence of images (or panels, in the respective terminology), that each present one scene, or one moment, or one view. It is the reader's imagination that connects successive panels to form a continuous story, or continuous motion. McCloud calls this action of filling in the blanks closure.

Reading a comic is a bit more interactive than reading literature, in that letters and words are being decoded at a relatively constant speed for each reader (actual understanding of a potentially very complicated text notwithstanding). Comic panels, in contrast, can invite the reader's view to linger, to scan for important details. Furthermore, closure requires active participation by the readers. So it takes a bit of time depending on how long the conceptual jump is between panels. Simply speaking, comic readers (and comic writers, of course!) have some control over the pace at which the story unfolds.

(There are even a few experimental comics where readers can choose the order in which they are reading the panels, and the story still makes sense. But such extreme forms are mostly unexplored.)

The exact same process of closure can be employed in dungeon design. When a dungeon is constructed as a (mostly linear) series of connected rooms, each room can present one scene, or one moment, or one view. And successive rooms can depict stages of some development that is happening right as players are advancing deeper and deeper into the dungeon.

My favourite example of this simplest and most direct analogy is a dungeon from WoW that goes by the name Blackwing Lair. In the first room there are lots and lots of dragon eggs. The third room is swarming with dragon whelps (I skipped the second room because it is part of another, overarching story). The fourth room is an alchemy lab populated with crazy scientists, grown dragons, and weird dragon-humanoid crossbreeds. The next room is a large hall with several dead dragon bodies, and a cage at the far end. That cage holds a ghastly monster that has been endowed with an unnatural combination of draconic powers. Finally, in the last room, there is the throne of the evil mastermind who has devised this scheme of breeding an unstoppable army. (There is one more panel of the story being revealed during the final fight, but I won't spoil it here. :-)

The sequence of scenes presented in the rooms of Blackwing Lair is telling a story as clearly as a sequence of panels in a comic would. All without ever showing the intermediary stages, or even only describing them in text. Not a single dragon egg ever hatches. Not a single dragon whelp ever grows up (but dragons do die at the hand of the players). All the gaps are being filled by the players themselves. Closure.

(Here I wonder if it would make sense to let players trigger prefabricated animations/transitions between rooms. Perhaps imagination delivers more detailed and more colorful FX? Or perhaps closure is about getting the point, about discovering the connection by yourself?)

Blackwing Lair actually isn't the most striking example in WoW (that would be Karazhan, in my biased opinion). But for other such dungeons I would have to explain a lot more background from Understanding Comics, especially the different types of transitions between panels. And I would have to go into more detail on the freedom that players have in choosing their path through a dungeon. Blackwing Lair is fully linear and transitions straight from stage to stage of a development, so it is pretty much the most straight forward implementation of the analogy.

What I described here was only the bare skeleton of the rich and detailed comic that was painted as Blackwing Lair. A lot more detail was required to make it actually work. But that would have only obstructed the view on the essential idea.


#22. Basics: The N² Problem

This one is a bit more technical, but a very fundamental limit of virtual worlds even with the fastest computers and the fastest network connections available. Consider a player in a virtual world as seen by the server that is running the game. Every player sends character actions to the server and receives information about the reactions of the game. For the sake of simplicity, let's say that the server needs to communicate one kilobyte of data per second with one active player. (This is just an arbitrary example, not a realistic value. Additionally, I will be a bit sloppy with the math for the sake of clarity.)

Now consider what happens when there are two players within virtual earshot of each other. They can both see each other's actions and reactions in addition to their own. The game server has to communicate double the number of events to both players, thus two players now require a bandwidth of two kilobytes per second each. The total communications load on the server is therefore four kilobytes per second.

In the general case of N players within virtual earshot, each of the players can see N other active players. The total load on the server will be N*N kilobytes per second in this simplified example. So just ten players already require a hundred times more bandwidth than a single player, and a hundred players require ten thousand times more bandwidth! This is the fundamental reason why virtual worlds cannot handle arbitrarily large numbers of players bunched up in close proximity.

Back in the early days of commercial MMORPGs, this was sometimes abused by vocal players. They could crash the virtual world simply by organizing big meetings. However, I have to disappoint contemporary leaders of player revolutions. The real cause for the crashes back then was that network equipment was not as mature as it is today, and unreliable under high load. These days, virtual sit ins will still cause the participants to be disconnected, but are highly unlikely to affect the game server at large. Each player's individual internet connection is much more easily oversaturated than the fat pipe of the backbone where the game server is hosted. And each personal computer's network stack is much more likely to be buggy than that of the mission-critical "big iron" server.

Still, game designers should not encourage players to all bunch up in the same virtual spot. Doing so will not lead to especially epic experiences. Instead it will make that game content unavailable to all but those with the fastest internet connectivity. Originally, instancing was a mere workaround for this very problem.


#21. Advanced: Instancing and Phasing

I was a bit unsure whether to file this posting under basics or under advanced. But the concept of instancing is a relatively late addition to virtual worlds, and it carefully breaks some of the fundamental rules. So despite this trick being used in most of the contemporary MMORPGs today (that would be June 2011 as of this writing), it isn't really a basic mechanic. (In my humble opinion, designers need to exercise due care wherever they use instancing. It is not a mechanic that you can employ and it just works.)

As an illustrative example, let me explain the idea within the context of a dungeon tuned for a group of players. If it's a popular dungeon, lots of players will want to venture into it. But if the dungeon is tuned for a huge target group size, it will be hard to coordinate that many players. Worse yet, the dungeon will effectively be unavailable game content whenever too few players are logged in.

The standard solution is to tune the dungeon for a smaller, more manageable, group size, and give every group their own independent and exclusive copy of the same dungeon. In other words, different groups are seeing different instances of the same dungeon template. They can play the dungeon at different times and at a different pace. Effectively, this makes the dungeon much more available on demand.

The downside is that players no longer share a common world with a single consistent time line. Some groups will defeat the dungeon's evil denizens, but others won't. Yet this is quite tolerable, because it is still a group of players who are sharing their dungeon adventure. Dungeons being self-contained works out to an advantage here again.

Instancing has been used for parts of the open world, too, but the paradoxes are more severe then. If I am not mistaken, the first MMORPGs that used open world instancing did so as a workaround for technical limitations. It was the lesser of two evils.

A recent extension of instancing embraces those paradoxes, and deliberately presents players with different views on the virtual environment. The idea is to let players see their progress manifested in the game world. The designers prepare a number of distinct stages of some chronological development in a particular area of the environment. Each of those stages then lives in its own instance when the game is running. Players can be moved between these instances individually to see those changes unfold as they play each at their own pace. All players who are at the same point of progression still share a common view of the one true game world.

In WoW this is called phasing, and the stages of development are called phases. That same game also uses phasing in disguise to implement distinct difficulty levels for the same dungeon. Only very recently has phasing of dungeons been used to implement chronological development of a whole dungeon.


#20. Basics: Dungeons

I have been referring to dungeons a few times already, so I probably need to explain the concept for future reference (better late than never :-). Dungeons are not a defining element of virtual worlds, nor are they exclusive to that genre of game. But they are a fairly common design pattern, even when they are not explicitly emphasized by the game rules.

So dungeons are self-contained pockets of the game world. This can be a dragon's lair, a haunted ruin, a mage's tower, a monster infested cave, or the headquarters of your local neighbourhood evil overlord. The place can be swarming with monsters, or plastered with traps, or filled with puzzles and secrets. Dungeons that stick to a small number of strong themes tend to work better, because the whole purpose of a dungeon is to concentrate and intensify a particular type of game content. A consistent story, and a consistent style, and a consistent architecture, and appropriate inhabitants will all work to reinforce each other.

Within virtual worlds, dungeons are unusual in that they actually have an ending. The size of a dungeon can be tuned such that it covers a typical gaming session. If a virtual world were an anthology of novels, then dungeons would be chapters that can stand alone as short stories. A dramatic arc can be most easily implemented in more linear dungeons. The experience is that of invading ever deeper into enemy territory. The conclusion is typically a fight against an end boss, the local evil overlord.

Dungeons usually are game content of higher difficulty, and in virtual worlds they often require cooperation of a group of players. This again serves to intensify gameplay and make the dungeon a special experience. Due to limited size and linearity, it is often possible to design a controlled ramp of difficulty in a dungeon, which is a perfect match for both the dramatic arc, and game design based on challenge and learning.

There is another completely different set of reasons why dungeons are popular in games. Since they are self-contained, they can be designed in parallel by independent teams. You can have some in-house competition for dungeon designs, and you don't have to scrap a loser in favour of a winner. All designed dungeons can be implemented in the game, and the teams can learn from each other as they watch players streaming in.

For the same reason, dungeons can also be released (i.e. opened to players) independently from each other and even independently (to a degree) from the surrounding virtual world.


#19½. Guesswork Addendum: Is Tanking for Killers?

Now that I have been stirring the idea from the previous posting a bit more, I am beginning to think that tanking could well be for the constructive flavour of killers. The game imposes some responsibility on the tank to manage critical aspects of combat.

As I recall my own experiences of raiding in WoW (i.e. coordinated play in fairly large groups), only the raid leader and the tanks had the license to freely speak in voice chat during combat. Everybody else was supposed not to waste this critical communications resource. Occasionally a cheeky explorer would blurt out an important observation, but that kind of speaking license had to be constantly renewed with information that was of actual high quality.

The raid leader(s) (i.e. the person(s) who organize(s) and lead(s) the group) are, of course, the prototype of benevolent dictator in virtual worlds. They assume responsibility for a whole group of other players. And in the long run, selfish (i.e. less benevolent) people don't survive in such a position. (A marked difference between the real world and virtual worlds - and a consequence of the difference between virtual power and real power).


#19. Guesswork: Is Healing for Socializers?

Those of you who already know the combat functions of tanks, healers, and damage dealers can skip the following description. For everybody else: several MMORPGs (but not really a majority) are structuring multiplayer combat versus monsters by giving players one of three specific functions. In general, combat consists of inflicting damage on a foe. That damage then reduces a resource called "life", "hitpoints", or "stamina", which eventually reaches zero, meaning that character or monster is "dead", "unconscious", or "exhausted" (the exact terminology usually depends on what age rating the game makers aimed for). Thus, dealing damage to a monster effectively reduces that monster's time quota for active participation in combat. Those who are the last still standing are victorious.

That covers the function of damage dealers. Healers, in contrast, do not inflict damage on opponents, but refill the hitpoints of their injured comrades. That means, healers buy more time to fight. Healers are typically constrained by some other resource that they have to spend for a healing effect. This prevents combat from lasting forever. Intelligent monsters would kill healers first, because that would be the most effective thing to reduce enemy combat time. This is where the tanks come in. They have tools to grab the attention of monsters and keep them away from the healers. So the overall structure of combat in such games is that damage dealers are trying to remove opponents from combat as fast as possible, healers make their comrades last as long as possible, and tanks ensure that the others are not disturbed on the job.

I have been wondering if there are maybe correlations between the three combat roles and the four motivations. From damage dealers to achievers is the easiest connection to make. Dealing damage is about friendly competition with the other damage dealers, and about continually improving one's personal performance. Certainly sounds like achiever. That correlation would also explain why players of damage dealing characters are so irate when the game rules seem to favour the damage output of competing characters.

But then what motivation is driving healers and tanks? I suspect that both these combat roles appeal mostly to socializers. The healers quite literally invest in other players. If this was a birthday party, the healers would be the people in the background who keep everything running smoothly. And their reward is when guests simply enjoy themselves.

The tanks, in contrast, are the center of attention. Not just the attention of the monsters, but that of the healers and the damage dealers as well. The reward for tanks is the cheers from the crowd, the slap on the back, and the knowledge that they got everyone safely through. If this was a birthday party, the tank might be the host entertaining his guests.

Killers are not really part of this analysis, though. They probably are not even in the dungeon (dungeons are briefly described in the entry on fun through learning), but somewhere out on the battlefield, engaging other players.

Explorers on the other hand will want to explore dungeons, too. I guess explorers would not be very attached to one specific combat role. They'd probably try to fill the gaps, because their only chance to go exploring is with a group that has the required number of characters in each combat role. If anything, then explorers would be more attached to character classes that can conveniently fill several alternative combat roles.

If my assumption were true, it might explain why World of Warcraft saw a pronounced shortage of healers and tanks when the Cataclysm hit. During this latest expansion (as of this writing), both tanking and healing have been made notably more challenging. But challenge is for achievers, while tanking and healing might be for socializers.


#18. Guesswork: Towards Understanding Addiction

I tried to narrow down what addiction means in the context of computer games in general and virtual worlds specifically. The scientifically best established definition is from medicine. Regular consumption of certain substances leads to measurable and permanent changes in the brain of the consumer. The disadvantage of this definition is that computer games don't involve substance abuse. Neither have neurophysiological changes been documented for gamers.

A somewhat weaker definition of addiction comes from psychology. It is focused on compulsive behaviour, and the observation that patients have lost conscious control over their actions. Another unspoken but important part of this definition is the assumption that patients can overcome their addiction with force of will (and a lot of help, usually). The trouble with this definition is that psychology is an inexact science. Addiction in this sense can be a matter of judgement, not an objectively measurable condition.

The final and weakest concept of addiction that I am going to present here is based on behaviour alone, and is a sociological definition. It is based on the observation that the addicts no longer perform their functions and duties in society. They "set wrong priorities" or otherwise let their peers and fellows down. This definition is focused on the addicts having lost control over their time.

This weakest definition, finally, leads to a few insights with regards to designing games. When a game rewards players for adhering to an unusual schedule, especially one that conflicts with the established schedules of daily social life, then that game is more likely to be perceived as creating addicts. If, say, a hypothetical game were to hand out double rewards during the hour between 3 a.m. and 4 a.m., then some players would feel compelled to maximize their virtual gains by playing at right this time. That kind of unusual schedule would be noticed by the people around such night shifters.

If that hypothetical game would hand out the extra rewards in the hour from 8 p.m. to 9 p.m., chances are that players would feel equally compelled, and put aside other activities during that time. But other people would be less likely to notice, and less likely to stigmatize that play style as gaming addiction. That hour was gaming prime time anyway.

The worst kind of game-induced schedules are those that slide across the day, like the 20 hour crafting cooldowns that existed at some point in World of Warcraft. A player might (for whatever reason, probably not a good one :-) try to maximize the throughput of this artificially rate-limited character ability. He will be rewarded by the game for logging in at "arbitrary" times: 9 p.m. on the first day, 5 p.m. on the second day, then 1 p.m. on the third day, and so forth. Any sane person observing this hypothetical player would certainly think that the player has lost control over his time, or is setting strange priorities in life.

In my humble opinion, game designers could defuse the whole addiction debate a bit if they took more care to let players control their time. Game mechanics that reward playing at ungodly hours are not a cause for addiction, but they are certainly a welcome excuse for both wannabe addicts as well as anti-gaming activists.


#17. Advanced: A Hypothesis of Fun in Games

This is a critical reflection on "A Theory of Fun for Game Design" by Raph Koster. The book is good food for thought for anyone interested in game design, and an entertaining read. The author presents strong arguments for the idea that players experience fun in games as they beat challenges, and what enables them to do so is learning to master the game. If I may do the injustice of summarizing the book in one sentence: fun in games is a consequence of learning.

I actually think this explanation is a good one. From what little we know about intelligence (as displayed in humans as well as in a few other species on this planet), hunting for problems to solve is generally an experimental, not necessarily immediately purposeful, and often unsystematic activity. It's playful, done only when there is time and energy to spare, not in emergencies when tangible results are needed right away. But under such ideal circumstances learning is definitely fun; more so if the learners feel that they are making progress somehow (and winning is the most obvious such sign). Of course, if the learners are not seeing the point, it can be as boring as, say, school. :-)

This particular notion of fun correlates pretty well with the explorer and achiever motivations, but leaves socializers and killers unaccounted for. Koster's presentation in general pretty much leaves multiplayer issues unexplored. I regard that as a major gap in his hypothesis. Nevertheless, this school of game design seems to be applied to an important part of MMORPG game content: dungeons. These are little pockets of the game world where architecture, monsters, stories and treasures are orchestrated such that a group of players can share an intense adventure which follows a (mostly) self-contained dramatic arc. Asking for a group of players to cooperate serves to show the epic-ness of this particular adventure: a single hero is not powerful enough to take on the mighty foes in there.

An archetypical dungeon, viewed with Koster's hypothesis in mind, is a somewhat linear series of ever harder challenges. These challenges usually take on the form of especially powerful boss monsters that each require a special strategy to be defeated. Ideally, these combat encounters are sorted by difficulty, and the last one is a (minor) lore character guarding the dungeon's most valuable treasures or secrets. Lore wise, only the end boss is important enough that her or his death will have repercussions for the outside world's evil.

So dungeons seem to be the kind of environment that designers can control tightly, to establish a defined pace and a learning curve. Unfortunately, the designers cannot control all the players in the group. Whatever pace the designers intended to set, in practice it is set by the average or the slowest runner. This disrupts each player's individual connection between his learning and the group's success. Somebody who plays perfectly might not get a positive feedback (because the group fails to defeat a boss monster). Somebody who plays rather bad might not get a negative feedback (because the other players manage to pull through nevertheless).

This is why my reflection of Koster's hypothesis must be critical in the context of virtual worlds. Despite its broad applicability and sound foundation, it doesn't cover all implications of cooperative multiplayer games.


#16. Advanced: Pacing

So now we have a virtual world with a rich variety of places. We have captivating stories woven throughout. And we have various intertwined games to give players a motivation to go see it all. Following the standard pattern, we have arranged everything along some sort of progression. The natural thing to happen is for players to spread out across the virtual world, each at their own pace.

Pretty much anything that differentiates between players will act as a force to dilute and disperse their characters. Challenges will separate those who can master them from those who cannot. Time sinks will separate those who have time to spare from those who have not. Rewards will separate those who desire them from those who do not. Learning will separate the quick learners from those who are slower. And so on, for every little thing.

This is not necessarily bad. But it works against the massively multiplayer aspect, in that people play aside each other, but not with each other. I think that any virtual world which emphasizes group play will need a force to counteract the dispersion of players as explained above. It is not enough to offer opportunities or incentives for group play. The players need to be able to participate meaningfully in that particular game content. Their characters need to be in just the right place along the progression axis. Ideally, the players themselves are in some sense comparably "progressed".

This calls for active pacing mechanisms. Active in the sense that the game designers have given some thought to what a good pace should be. So they can implement catch up mechanisms for those who are behind, and either slowdown mechanisms or "you have done enough" signals for those who are ahead. I personally like the signaling better. Slowdowns can be an incentive for players to waste even more effort, hence can be addictive rather than fun. (My definition of addictive isn't all that positive. I will get to that. Someday.)

I want to emphasize that a matching pair of forces is needed, both "push forward" and "hold back", to justify the label "pacing". Slowdown mechanisms have been in MMORPGs forever, but they slowed everyone equally, and so did nothing to limit the dispersion of players. Worse yet, some of those slowdown mechanisms were based on randomness. By effectively setting a random pace, different for each player, the dispersion problem only got worse.

As a sidenote: pacing mechanisms can be regarded as a feature for socializers, because proper pacing brings players together. But of course no amount of game mechanics can make arbitrary people actually like each other.

Pacing in World of Warcraft evolved significantly over the years. In the base game and the first expansion (Burning Crusade), there were only slowdown mechanisms (some of which were based on randomness). Late in the era of the first expansion, a partial reward (as described in the entry on progress) was made much more available in more places of the game. And the catalog of actual rewards available for the collected points (Badges of Justice) was heavily extended to provide many more useful power-ups to characters.

The new currency was highly available and allowed latecomers to catch up fairly quickly. Players who suffered from extreme bad luck could fill in the gaps that the random number generator had left open. Not everything was perfect, though, in that first iteration. One undesired effect was that players could save up an almost unlimited amount of the new currency, and then immediately buy power-ups whenever the catalog was extended.

So during the second expansion (Wrath of the Lich King), the game designers introduced a new, different, kind of currency whenever they added new game content to progress through. This solved the issue with saving up points ahead of time, but removed much of the incentive to play slightly older game content. The handling of the various currencies was confusing, too, and they cluttered up characters' inventories.

Finally, in the third expansion (Cataclysm), the designers have settled on just two generic currencies. One of those is fairly abundant, while the other one is rather limited. Both currencies can only be saved up to a maximum, which is both an incentive to spend the points and a sign that players have done enough for now. The abundant currency buys basic equipment for the characters (appropriate for tackling the current point in progression). The limited currency buys high end rewards (to fill the random gaps). Basically, the abundant currency is the force that pushes forward, while the limited currency is the force that holds back (or rather, that turns random slowdown into paced slowdown).

By the way, this model of two resources, one abundant, one limited, is being used for example in Starcraft (both 1 and 2), too. Minerals are the abundant resource that enables both opponents to build a basic army relatively quickly (or reinforce their losses). Vespene gas (for powerful high tech units), though, is always lacking. Together, these two resources establish a defined pace for both combatants. The game mechanics are actively equalizing army strength, to maintain suspense, and to magnify the difference in actual player skill.


Older Entries:
0 - 15


Newest Postings
Table of Contents
I can be contacted as hobold at this domain name.