The Balancing Act: The Challenges of Class Balance in MMOGs
Removed of their visual and aural splendor, video and computer games
are nothing more than a set of complex, interlocking systems and algorithms.
Just how many of these are you experiencing on a regular
basis? Several - it turns out. For example,
there are programs to determine how much damage you cause with
a single hit, how far you can jump, how tough your enemy should be, and
how the animation plays out when you land at whatever point the program
determines you should land. And, MMORPGs are probably the most complex
programming framework among the games that are out there because so
much of the decision making is left in the hands of the player. The
game must have a set of algorithms that respond to, and factor in your
At the heart of every online game is the concept of balance. Balance appeals to all of our notions of fairness and fun; it makes the boss mob finally beatable or the ignoble defeat a teaching moment. Combined with good information, it helps us decide what player or mob to attack successfully, and which would be pure suicide. And yet, it’s the most unstable of all game concepts, where changes to almost any aspect of the game - changes in a single spell, item, even crafting and the economy - can affect this balance in profound ways.
Craig Morrison, Funcom
It wasn’t so long ago that MMORPGs followed a fairly straightforward character advancement scheme based largely on their pencil-and-paper precursors. Put simply, new levels simply meant new items, new abilities, and more stats. This approach kept the math fairly simple, and many MMO game designers honed their craft by mastering the balancing intricacies involved in systems that seem fairly basic by today’s standards.
A storm of complexity was on the horizon, however, and the first few hesitating raindrops that fell were in the form of item proliferation. While more than a few online games recognized the inherent coolness of an ever-escalating progression of items and items sets that looked as powerful as they were, it was EverQuest that rendered these item sets with enough 3D graphical quality to show the incredible motivational power of an item whose actual existence was limited to bits and bytes. Armor sets with names like rubicite, adamantine, lambent, and ivy-etched were the status symbols of Norrath until they were replaced with other more epic and powerful (if not necessarily more meaningful or more stylish) items.
No MMOG has failed to capitalize on the player community’s consummate love affair with items since, but the familiar paradigm of armor, weapons, and jewelry might be called different things in different games. Global Agenda refers to these equipable items as “implants.” Champions Online calls them “upgrades,” with a special subcategory of “power replacement upgrades” that potentially alter powers in addition to providing stats. Champions Online has also formulated clickable devices, which powers designer Brian Urbanek says “were made, very intentionally, to be panic buttons.”
No matter their function, a carelessly spec'd item has the potential to imbalance the game (for an extreme example, look no further than the World of Warcraft’s “Martin Fury” incident). To prevent a character from becoming overpowered, designers must weigh the effect of the most powerful items available at a given level alongside the inherent abilities of that character.
But items are just one level of complexity in today’s MMOGs. Choice breeds complexity, and EverQuest’s alternate advancement system - a point-buy system of further class specialization available only to high level players - took mainstream ideas about character advancement to a new level. The “talent tree” that’s become something of a staple in the MMORPGs that followed World of Warcraft, is a direct descendant of EverQuest’s concept of "AAs."
Like Age of Conan, Global Agenda incorporates a fairly straightforward WoW-style talent tree where players of each archetype - assault, recon, robotics, and medic - can choose one of three advancement paths via the talent tree or mix their skills from all three paths. “When we introduced talents, that was another layer of potential exaggeration around how players could min and max and optimize their build.” Todd explained.
If MMORPGs like WoW, Age of Conan, and Global Agenda muddied the class balancing waters by adding a substantial amount of player choice to the advancement mix with talent trees, few MMOGs have allowed players such a huge degree of choice at any given level as Champions Online. In addition to the devices and upgrades described above. every few levels, players are able to choose any power from the total range of powers available in the game, meaning that Champions Online is a class-less MMOG. At least on the surface.
I asked content designer James Laird how the Cryptic team is able to even begin to mentally organize class and build options for balancing. “I think the saving grace of the Champions Online system is that creating an effective build in Champions Online is about exploiting synergies between powers. So the electricity powers add the negative ion debuff, and other electricity powers exploit that debuff.
“When you’re really looking at building a focused character, it’s not like you have a completely wide open smorgasbord, you have groups of powers that work well together. Of course you’re free to mix and match those powers; you’re free to take whatever powers you want, but in terms of having an upper limit on how powerful your character’s going to be for balance’s sake, we have an idea of the kinds of synergies that are built in.”
Brian Urbanek continued the thought by pointing out another added complexity of Champions Online: roles, which work much like “stances” in other games but also communicate your intended role to other players. “What roles are are simultaneous mechanics bonuses - if you go into defensive role, you just have more hitpoints and heals are just slightly more effective for you, but you generate energy just a little bit more slowly, your energy decays a little more quickly, and your threat is a little bit different, etc. What makes them interesting is that they are not just things that are there by default. They are things that you unlock by doing that thing. You unlock the defensive role by taking lots of damage over time, you unlock offensive role by doing damage. Eventually all players will unlock all 4 roles most likely, unless they play a dedicated healer from level one to level cap and never actually solo. We also have a mechanic where the roles auto-unlock for you once you reach a certain level.
“[Roles] are important to us because in a game where any player can take any power at any given time, it’s important to be able to not only self-identify as ‘I am tank, I am DPS, I am healer’ if you choose to, but also to be recognized by the community that you have that skill.”
So in addition to the powers, spells, and abilities that come with leveling, balancing a game must also take into account armor and weapons choices, clickable items and consumables, talent choices, and game-specific (or even class-specific) mechanics such as roles. How do designers even begin to boil all these choices (or appearances of choices) into decisions on how to bolster an underpowered class or nerf an overpowered class? Let’s take a look at some of their tools.
Tools of the Trade
While players might have legitimate feedback about what class might be overpowered or underpowered, our experience of the game is comparatively limited and tends to be highly subjective. Not surprisingly, statistics and numbers play a huge role in class balance decisions.
“My tool is Microsoft Excel, 2003 version, because 2007 blows.” Champions Online’s Brian Urbanek explained with his usual candor. “The metric I use is time, because ultimately time is the only non-arbitrary measurement. For how many seconds will these players survive? How many seconds does it take for players to defeat an enemy of a given class? How many seconds of that enemy’s class can a player sustain before they drop over? Everything in respect to time, actually.”
Hi-Rez Studios, makers of Global Agenda, has a few stat tracking aces up their collective sleeve. “The first tool, as far as outside of the game, is analytics and stat tracking. Really, that’s just a big investment in capturing statistics, storing those statistics in a data warehouse format, and then tools to report on that data and slice and dice it in a variety of ways. This is probably where we benefit from the fact that many of our engineers came from enterprise software and enterprise reporting backgrounds in addition to folks from MMO and game backgrounds, because a lot of that is about storing huge amounts of data in a way that you can do fast reporting on it and using OLAP tools to look at that data in a lot of different ways.”
Craig Morrison explained that numbers play a key role in Age of Conan class balancing, but was careful to note the human element too. “It is part science and part intuition and experience I think. We do, of course, have all the kind of ‘spreadsheet’ work in the back-end and development tools that calculate as many of the possible parameters as possible. On top of that, though, you then have the knowledge and skill of the designers involved.
“Working with a system, you have the general overview of how things interact and how players tend to behave in your game... Sometimes nothing beats spending time in the game itself and actually seeing how the players have been using the skills and abilities you have provided for them. Players are nothing if not inventive and they never cease to surprise designers with their ingenuity, so it is vital that the designers are also watching and learning themselves. ”
But what numbers do designers look at specifically to judge the effectiveness (or unacceptable überness) of a class? Our panelists answers were guarded on this point, and we can’t blame them for not being willing to give away their own recipe for the secret sauce. Todd Harris did, however, offer us this peek behind Global Agenda’s curtain. “The specific numbers we look at now are the ones you’d expect: kill / death ratio and total damage when we’re looking at classes, and when we’re looking at items we look at the amount of time it was used or in the player’s hand vs. the damage dealt. So one weapon might be doing a tremendous amount of damage but it also might be used a lot, so we can see how often it was equipped as well.”
“Because we also have the aspect of the FPS mechanics, we might also collect information not just on classes or devices but also the maps or specific areas. So for any particular area, especially on maps that are asymmetrical so there’s an attack / defend mechanism, what’s the win / loss ratio for that map and where are the hotspots in the map where players are dying often or making most of their kills. And that’s important to us as well.”
Brian noted that he’s partial to a fairly formulaic approach to numbers when talking class balance. “I start with a baseline value, the statistics that a player has. I then intend and dictate for players to improve their total performance through items by 30%. Then I derive what 30% actually means. For example, it is possible through statistics for a player to increase their total performance by 60%. Half of that 60% comes through items, because half of your total stats comes from items.”
So, with Microsoft Excel (2003) loaded on your computer, a working knowledge of statistics and analysis, and a hugely comprehensive data set, you could be well on your way towards discovering how the pros understand, analyze, and balance their games.
But as we all learned in middle school, unless we control the way in which we limit our variables, any level of experimenting can often result in skewed outcomes. How do developers control the variables in the wide-open testing environment of the MMORPG, where focusing too hard on a specific aspect of the testing environment might cut that aspect off from a wide variety of other variables that affect it in a fundamental way? The answer might surprise you, in the extreme.
Taking It to Extremes
Brian Urbanek explained his first steps toward class balancing see him pushing the functionality of the character to their natural limits. “My personal approach tends to be to find the extremes and make sure the extremes are balanced and fun. So to put together the most offensively DPS character that we could possibly build, then to put together the most healing-oriented character we could build, and then the most intensely tanking player and the most useful utility player we could build.
“We want to make sure that those four extremes are approximately equally valued, approximately equally fun options to play - the theory being that once the limits of the system, the outer edges of the system are in balance, everything else will fall inside that envelope and, at the minimum, won’t break anything, will be fun to play, and can be tweaked as we go along.”
Todd Harris’s Global Agenda team has a similar top-down approach. “At a high level we want each class to be interesting and fulfill a different role, so that does mean that the metrics may be different. Early on, we had a concept of contribution points, and early on we made this visible to players as well, but now it’s not. It basically was an attempt to equalize the contribution towards winning a match regardless of class. Healing would get points, assists would get points, defending an objective would get points. It’s not something that players see now, they just see the individual statistics.”
Reasonable Expectations and Accommodating Newbishness
Of course, not every or even most characters will get the best and most of everything available before structural inadequacies and imbalances rears their ugly heads. These might be planned inadequacies - such as when a level 12 is facing off against a level 19 in a tier one PvP instance or a level 27 is trying to solo a level 30 and up dungeon - or they might be of the unintentional variety.
Let’s solve the easy one first. Most of today’s MMOGs offer some sort of mechanic that allows newer players to pal around with their older, wiser pals. Call it sidekicking or mentoring, Brian calls it smart, and believes in getting players to participate regardless of level, especially in PvP. “Balancing for differences in levels is the kind of thing you do with meta-systems, not direct systems. What you do is you make sure that it doesn’t turn into a death spiral where the people who win get more rewards faster and just keep getting better. You make sure the rewards you give just for participating are enough to keep players coming back and eventually close the distance.”
However balanced a class might be, Craig noted that some perceived imbalances between classes are due to inherent and desired differences in skill level required. “In terms of accommodating a range of player skills in a class based systems you usually do that by the difficulty of the class mechanics themselves. Some classes are, often by design, inherently easier to master and play because they use simpler or more straightforward mechanics. This is not always a bad thing as it does allow for there to be some classes that more novice or newer players can pick up and contribute with. At a top level, of course, that also means that those who see that as an ‘imbalance’ don’t always appreciate the fact that a class can be ‘easier’ than another, but within the framework of an RPG system, where some classes are harder to master, it is also important to provide options that are a little more forgiving, at least on a situational basis.”
Todd explained that for Global Agenda, it was an iterative process of fun first, then came the math. “When we introduced the talent tree, we actually did a normalization exercise across all the weapons then so that from a math / formula based approach - things like how much energy they were all using per unit of damage or healing - we were starting from a firm math-based foundation. That was good and healthy for us, because that’s not actually how we started. We started more just designing things that were fun and had a really good feel. Then we did a lot of hand tuning on the balance again.”
The next question up for debate was this: since balance is so elusive: how do devs know when they’ve done all that they can possibly do with class balance? When is enough enough?
Craig noted that some imperfections in balance actually contribute to balance when viewed through the lens of the entire game. “ I think with any game that has an RPG system and, as mentioned above, caters for different levels of knowledge and skill, you will always have some kinds of inherent ‘imbalances’, or at least things that are more or less forgiving depending on the situation. What you strive for is that all the classes have a meaningful role in any given situation (be it PvP, Raiding or PvE) that provides the player with motivation and rewards their success. You want to achieve a situation where the players taking part feel that they are contributing to the success or failure of their chosen endeavor.”
The PvE experience, we can assume, is much easier to keep in balance because half of the experience is either scripted (in the case of raids and dungeons), or formulaically adapted to the players (in open PvE). That being the case, Craig and Todd were also quick to add that when Age of Conan and Global Agenda devs talk balance, they’re talking about balance across groups of players specifically in PvP.
From Craig: “For Age of Conan, for example, we don’t specifically balance for all one versus one combination of classes because our PvP content is built around group co-operation. So in a straight up duel there might be classes that are more likely to win much more often against a specific other class, but in an open PvP situation where all the classes skills and abilities come into play, all the classes should have a role and the ability to contribute to their team.”
From Todd: "For us, really the balance equilibrium that we’re working towards is effectiveness toward winning a team mission, not one-on-ones. That does mean that there are a lot of cases that are situational, where a class just got caught in the wrong spot or at the wrong time or with the wrong off-hands on cool-down, that sort of thing."
Brian added that one of the fundaments for the PvP team was to make sure that every single power had something impactful to offer in PvP. “We put a strenuous amount of time into all our powers, making sure that they are useful in PvP. A classic example is “taunt.” Taunt is a power that in most MMOGs is useless in PvP. Well, we didn’t want to do that. So taunt is something you buy, it’s not a power in and of itself - it’s an advantage you apply it to another existing power. We tried to make it be similar in its effect. If you throw taunt on a PvE enemy, it works as you expect - it forces him to attack you for a few seconds. If you use it on a player, it has two effects. It breaks their block if they’re using block (which everyone has natively and is extremely effective), and it puts a debuff on them which causes them to do reduced damage to anyone but you.
“When players do PvE - maybe because it’s best practices, maybe it’s because it’s what they’re used to - they tend to self-select into familiar roles. When players do PvP, they tend to go for a very broad spectrum of powers as opposed to really centralizing. Frankly, I don’t mind. I want PvP environments to be as organic as possible. I’m a firm believer that attempting to impose order on that would be to defeat all the purpose in doing it.” The balancing act in PvP is, therefore, something of a disappearing act. “The active theory is that by making sure that the actual contest stays as visibly skill based as possible, then winning is bloody well its own reward.”
With any game offering players substantive control over the powers, spells, and abilities they can equip, are you bound to have two extremes: the gimped and the overpowered? Brian seemed to think so, so I asked whether or not it was possible to create a totally gimped character in Champions Online. His answer? “Absolutely.” He noted that if you stay within the frameworks, you will have a more than viable character. If you think of a bell curve, frameworks occupy the fattest part of the curve (more or less useful depending on the situation - e.g. gadgeteering framework is excellent at raid support and PvP roles while the high DPS martial arts or fire framework is excellent at soloing), while custom user frameworks have the potential to occupy the long tails on either side of the bell curve, for good or ill.
That said, Brian noted that a gimped character, in actuality, would probably be one that hasn’t spent all of his or her training points, and if he or she really had, they’d have to make a fairly concerted non-random effort to gimp themselves.
If all this talk of bell curves is giving you community college flashbacks, Todd Harris explained that his team uses straight math to know that the classes of Global Agenda have achieved something like perfect balance: “We’d know that we nailed team balance if the team composition was two assault, two medic, two robotic, two recon and two 'other' - where agencies are choosing those other two based on strategy. I think we’ll see a lot of other variations besides that; we’ll hopefully see ten robotics in one case or eight recons and an assault and a medic. That’s where a lot of the fun is for us as designers and a dev team. Because it’s such an emergent game, we do hope to see some surprises there.
“As far as what we’re trying to get, though, since we’re trying to have all these classes become viable and contributing, the “by the book” answer is that we’d like to see at least two of each class on every team of ten at the point when players are choosing for themselves what gives them the best chance of winning.”
As for a final, more chaotic word on achieving balance, I turned to Brian, who isn’t afraid to use the n-word. “[PvP in Champions Online] is a great laboratory for me. If any combination [of powers] becomes too prevalent, I immediately nerf the hell out of it. I want to make other stuff be as appealing.” So a bit of discord indicates that you’re approaching balance, then? “If all the players come to agreement, then I have failed.”
Aside from analytics, formulas, spreadsheets, and all the numerical methods and rules of thumb that designers use, the panelists also had some helpful hints and lessons learned for dealing with the human side of the equation. These might be geared more towards fellow designers, but players will note how many successful MMOs follow most or all of these precepts and how many unsuccessful ones don’t, didn’t, or simply went through the motions without any real conviction.
#1 - Communicate with (and challenge) the players
#2 - Embrace the “Flavor of the Month,” then move past it
#3 - Mix It UP
#4 - You have to play to win
#5: Fun First