Magic: The Gathering's Richard Garfield's strategies for game balancing
This weekend at New York University, an intimate audience had the opportunity to hear game balancing strategies from a master -- veteran designer and Magic: The Gathering creator Richard Garfield.
PRACTICE at New York University's Game Center is a now-annual observance of game design as art and practice -- a conference focused on the moment-to-moment experience of game design, and on building a community that supports game design as a process in culture.
Garfield may be one of the most prolific and influential living designers, responsible for countless physical games that have influenced design and business models. Yet balance is an ambitious topic, even for him.
Given the breadth of variety in games, it's challenging to talk generally about them. "An analogy you might draw is that if you're a biologist and you're talking about life, there's a crazy amount of life," he says. "It helps to take a subset of that and say, 'I'm going to talk about that.'"
Balance in orthogamesFor the purposes of his discussion Garfield chooses to focus on orthogames, which he defines as finite multiplayer games (two or more players) that result in players being ranked -- classic games like bridge and chess fall under this umbrella, while FarmVille doesn't.
"We define balance as strategic collapse," Garfield explains. "What this means is that games can be approached with a set of strategies and if there's a strategy that is regarded as not being viable that people think should be viable, then people will talk about this as not being balanced."
Strategic collapse occurs when one strategy is so preferable that all players essentially choose it all the time. In the Magic tournament environment, looking for a variety of deck types is a good way to check the health of the tournament. A wider variety of types keeps the game balanced, and it's less enjoyable when there's perceived to be just one viable deck type. On the other hand, too many deck types is too confusing for most players.
"There definitely can be too many strategies," Garfield says.
Sometimes strategic collapse has nothing to do with strategy at all. For example, the player that goes first in Tic-Tac-Toe will win much more frequently, such that an opponent might lose interest in playing the game simply by virtue of not going first.
Garfield also highlighted the concept of play style collapse, relevant to balance -- players all approach a game in different ways, and a game that has weaker support for your approach than for other approaches will feel unbalanced to you.
Two ways of viewing balanceHe views two types of balance: Holistic and componential. When a game has separate components that need to be balanced, it's componential -- in Magic, should a lightning bolt cost just one red mana? In Diablo, should a piece of equipment give you 152 intelligence? Holistic balance issues concern the game as a whole -- in Magic should you start with 20 life or seven cards? In Diablo, should you be able to sell equipment for real money?
Sometimes the distinction is muddy, he notes, but it's still useful in talking about balance to differentiate.
"Often designers will design the game to be balanced for the expert," Garfield says. "This is certainly the way we thought about it in the early days of Magic, and it took me a while to outgrow this mode of thought."
A game balanced to favor experts risks other types of gamers having an unbalanced experience -- and the game may lose most of its players before they ever develop the skill level to attain the well-balanced experience. Meanwhile the experts run out of people to play with.
Also, balancing for experts often fails to consider that there might be levels of proficiency even above what the game is desgned to contain. Designers aren't necessarily the best players -- most of the time, they're not, actually.
All kinds of games are patched with rules to accommodate players that outgrow their bounds. Balancing is helped by the fact that most of the time skill goes up logarithmically, and the benefit for performance tapers.
The important thing is to provide options so that every type of player has a good choice, versus grouping players into correct and incorrect ways to use skills or classes. This lets players compete even if they're not mechanically-focused -- like roleplayers or fans of storytelling.
Some of these players will choose gear for their character based on appearance or narrative suitability, and feel forced out of their play style by the fact that other players are more successful. That unbalances their experience.
Strategies for balancing"Balance is an art, not a science," Garfield says. Math helps in his game design, but never supersedes the level of complexity that a poker player would need to know.
"I came to the conclusion there is just no formula," he says. "If you haven't solved the game, it seems a tall order to actually come up with a formula to balance the game in the sense that people usually mean... and balance is different for different audiences. If you solve the game and then balance it, then you're only balancing for the expert, which we've established you don't always want to do."
And audiences are always changing: Beginners become intermediates become experts, younger players become older players. "It's a moving target,"" he says. "Because of all this it's really a matter of psychology, rather than math, so you shouldn't expect there to be an exact formula."
An iterative design approach with a lot of testing and flexible prototypes is one essential technique to find a game's natural balance -- Magic was playtested for two years. "I have tried many, many times to design games as a document... but it is really hard to internalize what the game is going to be like unless it's really closely modeled on a game you already understand."
One benefit of iterative design is that everyone is a beginner at first, and the scope broadens as development goes on. But a risk is that the more you test and iterate, you can lose track of your beginners and casual players -- so intentionally bring in new people regularly and ensure you have a good mix of players familiar with the game and people who know little about it.
These days, publishers are likely to iterate after a game launches -- which is fine provided it launches with enough balance for beginners and can rebalance fast enough to keep up with the growing player base and not lock out new players either.
Rock-paper-scissorsRock-paper-scissors' structure -- where every element of a game is strong against a different element -- is incredibly helpful to keep in mind while balancing, Garfield says. On the component level it appears in a broad range of games, from Stratego to Team Fortress 2. On the holistic level it works too -- take rush, defense and economic-oriented gameplay patterns for Starcraft, where each strategy is weak to one other strategy.
"Once that was recognized by Blizzard as being a healthy rock-paper-scissors relationship, it's designed with that in mind," Garfield points out. "If they make units or base states for the game which imbalances that and makes one of those strategies dominate the other, then they will tweak those back."
While rock-paper-scissors in itself can be regarded as a trivial game, there're options: Rock doesn't need to beat scissors 100 percent of the time, for example. As long as the odds are better than 50 percent, then the structure is sound, and in fact in many cases it's best to make it so that the entire game doesn't rest on a single choice. Choosing a strategy within a cycle simply improves the success chances of a player that prefers that strategy, rather than guarantees a victory.
Add costAnother useful idea for balancing is to generate and define a "cost" for components that need balances, where that cost can be multiple resources. The cost can be paid during the course of the game, and all players begin at the same base. If the game has in-session costs and between-session costs, they usually will be a different resource and sometimes the cost will be difficult to define or subtle.
"The important thing is that you have one, or perhaps a few numbers to tweak to balance your components," Garfield says. "With cost and score [you need] one number that your developers recognize as being the principal knob. And as you become better at understanding your system... you'll become experts at tweaking that knob. In Magic, once we got good at tweaking the mana cost knob, we got better at tweaking other knobs."
Non-dominationStart by setting a series of benchmarks for cost. After that, grow the system -- add new components in such a way that new ones don't dominate the old ones by being strictly better, he advises. With domination things are easy to value and thus games have less dimension; non-domination generally offers a wider array of viable choices.
If a component is too powerful, you can use a strategy or component that "hoses" it, or reduces its efficacy. Magic uses this a lot; for example, the hurricane card damages all players and specifically flyers; "if I'm running into problems with my opponents playing a lot of flyers, here's an answer. Now if they rely too much on flyers, I will beat them." Starcraft's observers hose invisibility. Adding a hoser creates a viable response for players to use against specific strategies or powerful components.
Often, hosers can create a rock-paper-scissors situation, where Strategy A is the ideal choice against Strategy B but is vulnerable to being hosed by Strategy C.
Players can also bid for access to elements and features as a method of controlling their growth arc and keeping balance intact. Sometimes variance also assists in balance -- a strategy that isn't viable most of the time may randomly become viable. This is often present in card games. To mitigate a very powerful approach, make that approach useful under rare circumstances.