Practice 2012: Richard Garfield on Balancing Games

To kick off the Practice conference, Richard Garfield, designer of games such as Magic: The Gathering, Robo Rally, and The Great Dalmuti, gave a lecture about the art of balancing games. 

Presentation

• We have a broad view of games. Games can mean anything from sports, board games, card games, game shows, and video games. For this talk, Richard is specifically talking about balancing orthogames, which are finite multiplayer games that result in players being ranked.
• Games can be approached with a set of strategies. If there is a strategy that isn't viable when players think it should be, people will think the game is not balanced. If there is one strategy that is a clear winner, the game won’t seem balanced. This leads to "strategic collapse."
• "Toy games" are not actually fun games, but are designed to illustrate a point. Imagine a toy game called Big Rock Paper Scissor. In this game, players can play a big rock (clasp both hands) which beats both rock and scissors. This would remove the viability of little rock as a strategy, but the game as a whole is still viable. If you as a developer don't think little rock should be a viable strategy, then the game is still balanced. However, if big rock is able to beat all the others, then all strategies collapse and the game is unbalanced. One strategy that is not viable is okay, but one strategy that overpowers everything else is not okay.
• A balanced Magic: The Gathering game is said to have 4 to 8 viable deck types. Less than that, the game would be considered unbalanced.
• Sometimes strategic collapse has nothing to do with strategy. For example, in Tic-Tac-Toe, the first player wins 90% of the time.
• Play style collapse is when a game doesn't fit a player's expectation of play style. For example, the player wants to be a builder, but the game only rewards players who are fighters. Common play styles include builder, collector, fighter, explorer, trader, and socializer.
• The MMO Earth & Beyond has three tracks (explorer, fighter, and trader). Richard tried to play the game as a trader but realized that the fighter strategy was way more viable.
• Balance is not an issue with all games, such as Bingo, but there can be stochastic collapse. In the toy game Ingo, you can also win if you dab any spot next to the center. The game loses all tension because a person who is very close to making five in a row can suddenly lose to a player who got one lucky number.
• Games can be balanced holistically (looking at the system as a whole) or componentially (looking at the components of the system). A componential balancing of Hearts, for example, is determining how many points the Queen of Hearts should be worth. A holistic balance would be determining how many points players should play up to in order to end the game.
• When balancing for different player types, you should not balance for the expert because it'll funnel out players. Be prepared to be frequently patching your game.
• Richard designed a game called Spectromancer, which launched with 6 classes. One class, the cleric class, was free to play. The cleric was considered balanced by the team, but because it was free, the players regarded it as the weakest. Later on, the expert players discovered that the cleric was the best class in the game.
• Investigating deeper into this, the developers discovered that the cleric had an okay win ratio for beginner players, a low win rate for intermediate players, and a very high win rate for expert players. Meanwhile, the necromancer class had high win rate for beginners, but a low win rate for expert players. All classes had varying success rates depending on the skill level of the players. This may seem unbalanced, but in practice, casual players play with other casual players, and expert players play with the expert crowd. This makes different strategies viable in different crowds.
• You should not balance for the role-player. You can create a weak sword and a powerful sword, and argue that role players would sometimes use the weaker sword because their character is meant to be weak in the story. However, they will soon realize this is not a viable strategy, especially when playing with non-role-players, and the game will become unbalanced. All types of players need balance.
• Balance is an art, not a science. There is no magic formula. You don't need advanced math, and although some knowledge of math helps with balancing, so do all other topics and subjects.
• While designing Netrunner, Richard tried to make a formula to balance the game. He ultimately failed because for every new card he introduced, he had to add a new rule to the formula. He realized that the formula can't predict any cards, it was only describing what's been done. He attributes his failure to two reasons: 1) games are complex and we cannot solve them, and 2) balance is different for difference audiences. The audience is also changing all the time, growing older and becoming more expert at the game.
• Balance is psychology more than it is math. Fortunately, balance is forgiving.
• Richard shares several techniques he uses to balance games.
1. Iterative design. Design, prototype, evaluate, and repeat. There is no substitution for trying out your design.
1. Magic: The Gathering was playtested for two years before it was released.
2. All quality products are iterated a lot such as Minecraft and Blizzard's games.
3. The benefit to playtesting is that you can't design for experts since everyone is a beginner. However, make sure that you are not using the same pool of players to playtest and that you are always adding new playtesters, so that you keep you design geared towards beginners.
2. Rock-Paper-Scissor technique. RPS offers an excellent structure for balancing and is used in many games. You can use this technique on the component level (ie. an archer beats a lancer who beats the calvary who can beat the archer) and on the holistic level (ie. in Starcraft, a rush tactic beats an economic build which beats a defensive build which beats the rush strategy).
1. In Magic: The Gathering, an aggressive deck beats a control deck which beats a combo deck that can beat the aggressive deck.
2. In Hearts, shooting the moon beats out a ducker (who tries not to take any points) that beats a sheriff (who prevents moon shooting).
3. Rock doesn't need to beat Scissors 100% of the time because you don't want the entire game to rest on one decision. An above 50% advantage is good enough.
4. RPS technique can also work with more than three components.
3. Component costs. Add costs to components you want to balance. This can be a single or multiple resources.
1. Costs can be paid in-game or before the game. Usually, you use difference resources for in-game and between-sessions costs. In Magic: The Gathering for example, the in-game costs for components are mana, but the between-sessions costs would be actual money to collect the cards.
2. Costs can be subtle such as using "mana" or "insanity." It doesn't have to be money. This is a number you can tweak.
3. Robo Rally doesn't have any costs but it has stats. Costs and stats are basically the same in that they're both knobs you can tweak.
4. Benchmark and non-domination. Set benchmarks (vanilla costs) and make sure any new component you add doesn't dominate the older components. Map out a network of strategies and make sure you have a range of viable choices. With less domination, there is more room for exploration and choice, but too many choices can also confuse and paralyze the players.
5. Hosers. If a component or strategy is too strong, add a rule or a component that hoses it. This is a response to anything overpowering and adds a RPS relationship.
1. In Settlers of Catan, robbers were added to hose players who hoard resources.
2. Magic: The Gathering has tons and tons of hoser cards -- ones that hose a certain color, ones that hose a certain ability, and even ones that hose players who like to use a lot of cards with the same artist.
6. Bidding. Bidding for access of elements or features can auto-balance them. Examples of this include auctions in Monopoly, Vegas Showdown, and Star Wars: Trading Card Game. In Star Wars, players bid to play a certain side (Light side or Dark side).
1. This can also be more subtle like a draft pick.
2. The benefit of bidding is that it adjusts itself to players. The problem is that it requires a lot of skill to participate successfully.
7. Variance. A strategy or component that is not viable most of the time might become randomly viable sometimes if the game has a lot of variance, like shooting the moon in Hearts.
1. In many card games, depending on what hand a player is dealt can impact what strategy is most viable to that player. 
2. Variation can also soften the edges of your design. For example, if a game has a problem where the player who goes first usually wins, variation will even that advantage out.
3. Luck in games is not like a panacea that can solve all problems, but more like an ibuprofen that can heal some slight pains.

Question and Answer

• Magic: The Gathering has some deliberate unbalance with useless cards. The developers wanted to give all players the pride of acknowledging bad cards and removing them from their decks.
• The more components in the game, the harder it is to balance. At Wizards of the Coast, the developers created two teams to work on Magic: The Gathering - one to work on overall design while the other focused on balancing the cards. The team was built around a method to balance games.
• Magic: The Gathering avoids balancing new cards against the entire library of old cards by retiring and expiring old cards. This limits the time frame and number of components during the balancing phase.