Most strategy games have fixed units and possible behaviours.

However, think of a game like Magic The Gathering : each card is a set of rules. Regularly, new sets of card types are created.

I remember that the firsts editions of the game have been said to be prohibited in official tournaments because the cards were often too powerful. Later extensions of the game provided more subtle effects/rules in cards and they managed to balance the game apparently effectively, even if there is thousands of different cards possible.

I'm working on a strategy game that is a bit in the same position : every units are provided by extensions and the game is thought to be extended for some years, at least. The effects variety of the units are very large even with some basic design limitations set to be sure it's manageable.

Each player choose a set of units to play with (defining their global strategy) before playing (like chooseing a themed deck of Magic cards). As it's a strategy game (you can think of Magic as a strategy game too in some POV), it's essentially skirmish based so the game have to be fair, even if the players don't choose the same units before starting to play.

So, how do you proceed to balance this type of non-symmetric (strategy) game when you know it will always be extended?

For the moment, I'm trying to apply those rules but I'm not sure it's right because I don't have enough design experience to know :

  • each unit would provide one unique effect;
  • each unit should have an opposite unit that have an opposite effect that would cancel each others;
  • some limitations based on the gameplay;
  • try to get a lot of beta tests before each extension release?

Looks like I'm in the most complex case?


5 Answers 5


I'd throw away the first two of your bullet points. It might be a good idea to get some units designed, but isn't going to really help you for balancing.

Really what you're going to have to do is just play the game a lot and keep a analytical mind (or set of minds) looking at the problem at all times. Design is a very soft art, there is no iterative way of saying whether or not a game is balanced.

On the other hand, these are the kinds of things genetic algorithms are pretty good at weeding out. It won't replace having a large beta test and human feedback, but it might catch some edge cases you won't normally think of.

Speaking of Magic, here are som other things it does to help its balancing

  • Cost based balancing. You may get some cards, but playing them or using their abilities requires a tweakable amount of mana
  • The luck of the draw. You don't start with your perfect set of 7 cards, and you can have at max 4 of the same card in your deck (sometimes less depending on the card). Keeps people honest.
  • General purpose counters. Even if you can't fight a unit or ability directly, you might be able to use your finite resources to prevent something from happening against you in the first place.
  • Constantly updating ruleset. You might want to go ahead and implement some kind of extensible rule system to ban or limit certain units
  • \$\begingroup\$ Thanks for your advices. The way they manage their game balancing for Magic always seem obscure to me. \$\endgroup\$
    – Klaim
    Commented Dec 7, 2010 at 16:24

Actually, M:tG has gone through ups and downs where some expansions were overpowered and some were underpowered. For example, most of the expansions around 3rd and 4th edition were fine but the Tempest expansion made a lot of earlier cards redundant. Maybe they're more careful now but they were still making significant balance 'mistakes' (or deliberate changes to encourage card purchases, if you're cynical) quite some way into the cycle.

The key thing here is to watch for obsolescence. If 2 objects have the same effect but one costs less than the other, the more expensive one is 'dominated' in game theory terms and thus worthless. To prevent domination of one object or strategy over another, each needs to have a situation where it is worthwhile. For example, one unit might be stronger than a second at the same price, but maybe the second has an extra ability, or more subtly, maybe the second can be used in conjunction with a third in a way that the first cannot.

Often people suggest some sort of 'rock/paper/scissors' approach here, but beware - although this does mean that no strategy dominates the others, it also means that no strategy has any particularly interesting features either. M:tG improved on RPS by having qualitatively different approaches for each of the 5 colours instead of the typical circular relationship of X beats Y, Y beats Z, Z beats X, etc.

So it's usually easy to ensure that none of your objects are completely dominated and worthless. But it doesn't necessarily mean that they are going to be judged as 'balanced', especially if the situations where they shine are quite limited. To do this, you probably need to ensure that you understand each of these potential situations. In M:tG, at least back when I played, you had a few standard types of strategy: direct damage, lots of small creatures/speed, a few large creatures, control/countermagic, etc. If you have a list of strategies like this, it is possible to see how each of your game objects might fit into one or more of them. eg. A medium-sized creature of average cost is not interesting for the two creature-based strategies above, but if it happened to use the same resources as the direct damage or countermagic approaches did, it would help round out that strategy.

Testing is a good idea. But player-driven beta tests should be your last resort of testing, not your main method. Firstly, you should be numerically analysing your game objects. You control the rules, so you can know exactly how many resources are available and therefore how quickly attacks can be launched, etc. You can work out maximum bounds and averages for these, compare them between strategies, and adjust objects to reduce massive disparities.

Secondly, from a very early stage you should have AI systems that are able to play many games against each other at a rapid speed and learn from their mistakes. You can start them with strategies the designers devised, or have them pick random ones, and evolve between them by seeing what works and what doesn't. Leave them grinding overnight, gather the statistics, and see what they're telling you. You'll probably find they converge on one or two strategies that always win, which gives you something to adjust.

  • \$\begingroup\$ Thanks for the advices, that's helpful. About the AI thing, I thought it would be too hard to write several AIs for this type of game because it suppose that a generic AI can understand the forces and weakness of each unit types and use the combinations in a smart way, or make severa AI having different strategies and prefering specific units combo... so it's a lot of work. \$\endgroup\$
    – Klaim
    Commented Dec 7, 2010 at 17:45
  • \$\begingroup\$ Making a game is a lot of work, AI or not! I'm sure that if a human can understand the game, then an AI routine can too. If you can encode a rule in code, then you can understand a rule in code. \$\endgroup\$
    – Kylotan
    Commented Dec 8, 2010 at 1:07

Normally the term is "expansion" not "extension", FYI.

Best way to do this is to understand the cost curve for your game. If you know that all of the cards (or units, or whatever) in your game have the same cost/benefit ratio, and that it is balanced, then you can theoretically come up with as many other things as you want, and as long as they fall along the same cost curve they will (theoretically) be balanced with what has come before. This isn't how Magic does it, but most other CCGs I've worked on have done something like that. As a bonus, knowing the curve significantly reduces the required time it takes to playtest for balance in future sets, so it saves you development time/money.

Here's the down side: it's really hard to know the cost curve up front. Usually this is something best discovered after the release of the base set and maybe another set or two after that... at which point, you've already got a bunch of (unintentionally) unbalanced stuff. With computer games you can always release a patch that balances the old stuff later, provided you can handle the fallout from the player community (they tend to get uppity over any change to their beloved game, even if it's a positive change, unless they are carefully managed).

Of course, this won't get you ALL of the way there. Sometimes you make cards or other effects that are just so far out there that they can't really be directly compared or balanced with anything else, so you just have to go by feel with your instincts. But if there's only a few of those, you can spend the bulk of your time working on balancing those by hand, and leave the rest to the automated math.

There is one other problem you will run into eventually which is not balance related, and I think this (and not balance) is the main reason why Magic started using "Type 2" and other limited formats: eventually the card pool just gets too large to manage. Sure, players who have been there from the beginning who are incrementally adding one set at a time to their knowledge of the game, they can manage indefinitely... but for NEW players, coming into a new game and suddenly realizing you need to memorize a thousand cards is a massive barrier to entry, and you'll find that without some kind of restricted card pool to draw from, your game is effectively closed to new players entirely. So have a plan in advance for how you're going to handle that, irrespective of balance issues.

If you're interested in cost curve balance, I've written up a tutorial here: http://gamebalanceconcepts.wordpress.com/2010/07/21/level-3-transitive-mechanics-and-cost-curves/ (The rest of that blog is devoted to other topics on game balance that you might find interesting as well, but this particular post is directly relevant to CCGs.)


It's going to be hard to balance things when the set of things you're balancing is open-ended. That's kind of the nature of the thing, and why we call it "balance". Adding one new unit can have a dramatic effect on other ones.

If your game is already multiplayer and socially-oriented, one possible solution is to crowd-source it: let players rate the relative strengths of different units, and then aggregate that to give each unit its score. Then, when building their troops, set a cap for total unit strength of all troops.

Or you could otherwise leave it up to the players to have control over this by letting them define game requirements that allow or disallow certain units.

You can't shirk all of the work onto the players. No one will bother to spend the time ranking things if the game is so unbalanced that it isn't fun to play. But you may be able to give them the ability to fine-tune.


I have a strategy game in a similar situation: superhumans with a diverse array of powers. I'm not explicitly planning on expanding the character list, but it's an obvious way to expand on the game if it's well-received.

The term I've used to describe the game rules is "exception-based", which I'm pretty sure I've heard elsewhere in reference to M:tG. Powers mostly operate by acting as exceptions to a set of common rules rather than by introducing new rules. Where new rules are introduced, they are typically coupled directly to the powered character itself.

My game is neither finished nor released, yet. That said, here's what I've applied to balancing so far:

K.I.S.S. - Each power has just one effect, expressed as simply and concretely as possible. As the powers are themselves exceptions, I keep them free of exceptions and special cases. Where characters interact, it is always clear whether their powers interact or not and if so, which one dominates. I created a large spreadsheet comparing each power to each other to map out unclear interactions. Although this design work wasn't really 'balancing', keeping things clear let me identify different ways of grouping the powers to look for common issues and solutions.

Exceptions to what? - Every character is modeled on a base set of attributes and actions. Together the basic attributes and actions define what I call a 'pawn-level' character. When characters lose their powers, they still have this to fall back on. Also, given the limited focus of the powers, any given character is 90% pawn-level outside of their specialty. Regardless of how powerful a character is, each one has at least one pawn-level Achilles heel.

90% Dominance - The KISS and pawn-level aspects come together to create my 90% Dominance plan. Each power exception maps to one or more of the pawn-level Achilles heel aspects. Given the distribution of powers, this means that each character can dominate 90% of the other characters within their own area of expertise. In many cases this also limited the number of powers that needed to be balanced against each it: only a subset of all the powers would interact with it. Only a subset of those interactions didn't have a clearly dominant power. With around 100 unique powers, each one required direct balancing against around 5 others.

The game is designed about as far as it can be without a lot more playtesting. Playtesting is super important, especially for a game with as many combinations of things as mine (and yours from the sound of it) has. There is no degree of solo designing and balancing I can do that doesn't show up all sorts of inadequacies once people actually play it.

Leaving you with a link to a site I haven't seen linked-to around here yet. Balancing Multiplayer Games, Part 3: Fairness by David Sirlin. There's plenty of useful info there, but be warned that he is dogmatically committed to the idea of Playing To Win (might not seem like something worth warning about, but just keep it in mind).


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .