Thursday, April 05, 2007

The Digital Game Canon: StarCraft


Perhaps the most definitive RTS game, ever. Consistently considered one of the best games of its genre ever. Played as a professional sport in Korea.

Obviously, Blizzard was doing something right with that one.

Basically, this game is a testament to how to balance a game with multiple distinct sides. Chess and Go are "automatically" balanced because each side has the same pieces and takes the same actions (though some suggest that the first player in Chess has an automatic advantage). StarCraft, by contrast, deals with 3 distinct sides.

The beauty, design-wise, of StarCraft isn't just that it managed the feat of almost perfectly balancing 3 sides. The beauty is that it did so in such disparate ways; that each side is so very distinct from the other and yet still is balanced.

On the small scale, take the first two units that each side gets, for example. One ranged and one melee. However, ranged units (according to the game's mechanics) have an automatic advantage over melee units. By the very nature of how they attack, lots of ranged units can all be attacking a single target, whereas far fewer melee units can gang up on a single target. Therefore, it was designed, on an implementation level, to make the early-game melee units more powerful per cost than ranged units.

Because each side is so distinct, however, they went about it in three different ways. The Protoss Dragoon is powerful, but hideously expensive. For the cost of a Dragoon, you can get anywhere from 2 to 5 melee units (except for Protoss melee units). And sheer numbers makes this a forgone conclusion. And while a Dragoon can kill a Zealot one-on-one, a Dragoon costs a lot more than the Zealot, and it will take more damage per-cost than the Zealot did.

Terran Marines are relatively cheap, and the first Terran unit, and yet two Marines will be mercilessly slaughtered by a single Protoss Zealot. A single Marine cannot hope to survive against a pair of Zerg Zerglings. And a Marine will fall to a Terran Firebat (not quite equivalent in cost, since they require gas to produce, but close enough).

Zerg Hydralisks are often considered the most cost-efficient unit in the game. Even so, a single Firebat will destroy them. And don't even consider sending them one-on-one up against a Zealot (not exactly the same cost, but close enough). And 3 Zerglings will likewise own them.

But look at how each individual melee unit achieves its dominance. The Zealot, despite having it's hugely powerful Psi Blades (16-damage per strike), doesn't win due to its damage output. It attacks far too slow for that. It wins because it can sustain massive quantities of damage. The Dragoon doesn't win in a cost-analysis because it doesn't shoot fast enough to kill the Zealot before it meets out substantial damage. A pair of Marines may not even penetrate the shields of a Zealot before it has its way with them. And a Hyralisk likewise can't kill it fast enough.

The Zergling wins through sheer numbers and awesome damage-over-time output. For the cost of a Dragoon, you can get 5 Zerglings, which deals 25 damage per blow. A Dragoon can kill a Zergling in 2 shots, but that's still plenty of time to get in 50-75 damage. And there's still 4 Zerglings left. A Marine costs 2 Zerglings, and Marines, despite their high rate of fire, can't kill two Zerglings before they exhaust the meager 40 Hp of the Marine. And a Hyralisk costs 3 Zerglings (or 4, depending on how you count gas costs), and they can deal damage much faster than the Hydralisk.

The Firebat wins through its brutal damage output. Though it has the same damage as a Zealot per-attack, its refire rate is far superior. It doesn't have the toughness of a Zealot, but it's not exactly weak either. One-on-one, it's a tossup as to whether a Hydralisk would win, but the costs aren't equal; a Hydra costs 33% more than a Firebat. Two on one, and it's no contest. The Dragoon's cost means that you can get fully two (2.5 by some reckonings) Firebats, and that's enough to tear down the Dragoon's defenses and kill it. And a Marine's damage output can't possibly cope with the Firebat's output.

Each unit is balanced in a way that suggests an overall strategy. They are all balanced, all following the implementation rule, "Melee beats Ranged," but each does it using entirely different mechanics.

And yet, the balance in this case isn't absolute.

Zerglings rely on numbers to be brutally effective. With a line of Dragoons, one long enough so that the Zerglings can't get around it, a problem arises. Only 3-4 Zerglings can attack a single Dragoon, but if the Dragoon line has more Dragoons behind it, then numerous Dragoons can fire at a single Zergling. Even if the costs are kept equal, 4 Dragoons in a line are better than the 20 Zerglings they would go up against. Thus, Zerglings lose their advantage.

Except, Zerglings can be upgraded. With appropriate armor upgrades, a Zergling can require 3 (unupgraded) Dragoon blasts to kill. At the very least, this requires that the Protoss player upgrade the Dragoon's attack to achieve parity again. But Zerglings also have an Adrenal Gland upgrade. This, coupled with their full attack power upgrade, can allow them to more than double their damage output. Thus, one fully-upgraded Zergling becomes two unupgraded Zerglings. To fully counter the attack power upgrade, the Protoss player must upgrade both shields and armor. And nothing can be done to counter the Adrenal Gland upgrade, so the greater rate of attack is still available to the Zergling.

In the end, Zerglings can be made less effective by a Protoss player's actions. But that player must then devote time and resources to this upgrade, thus forcing the Protoss player to slow down.

So, in the early game, melee units are better than ranged units. With time, upgrades, and numbers, however, this changes. And thus, a playthrough of the game changes from moment to moment.

Single-unit tactics eventually become meaningless, because they are easily countered. However, remove 10 of those Zerglings in the Dragoon example and replace them with 3 Hyralisks, and everything goes badly for the Protoss player.

On a large scale, one can see that each side is balanced by it staking out a particular set of player strategies/ideologies. As long as each side remains ideologically distinct, and each side can reasonably exploit a weakness of another's ideology, then the game can maintain a semblance of balance.

The Zerg ideology, for example, is on cheap units and fast expansions. One weakness of this is that these cheap units are not always monetarily efficient compared to others. The Terrans, due to their ability to quickly repair anything they make, can be more efficient in terms of resources. At the higher levels of play, it is said that for a Zerg player to beat a Terran, he has to stay an expansion ahead of the Terran player. The Protoss ideology of quickly moving powerful forces for a strike (the infamous Protoss drops) easily can counters Zerg expansion, simply by quickly eradicating expansions. This forces the Zerg player to substantially defend any expansion, which makes attacking more difficult.

The take-home lesson for StarCraft is about perfecting game balance in entirely unique ways. It's about developing a good set of mechanics that allows for deep gameplay, yet using implementation to maintain balance between different sides in unique ways that exemplify the ideology of each side.

Monday, April 02, 2007

Game Design: Knowing the Unknowable

Earlier I considered the question of what game design was. This was partially in regard to the Situation Model, but it also branched out more into the relationship between what the player sees (the set of possible inputs for a given Situation) and what the game designer creates (a list of rules and so forth that determines what a given Situation is and how the player can act therein).

The secret of game design, the purpose of having a coherent theory of game design in the first place is to answer the following question: how does a particular set of rules create a specific set of Situations that the designer would feel that a player considers interesting and potentially entertaining? That is, a full theory of game design answers the vital question of Chess. How do these seemingly simple rules create such a degree of complexity, depth, and challenge?

However, we may not want to use game design to create "complexity, depth, and challenge." We may want to do something else. Game design can be used to create other emotional reactions (fear, suspense, etc) or to do things we haven't really thought of yet. So a greater, more general formulation of the question is this:

Given a specific goal for a game's design, how do you create the appropriate elements to achieve this goal?

One problem with this question: it may not be answerable.

That being said, the Situation Model is a good tool for answering this question. Using it as a model, a game designer can test bits of gameplay (whether as a thought experiment or as a live prototype) and attempt to experience the game using particular bits of Knowledge and/or Experience. As a design tool, its most useful purpose is making sure that the decision making portion of the game is without pathologies, difficult to the specific degree the designer wants, and has the specific pacing and advancement that the game designer wants.

The Situation Model is a priori knowledge. That is, it is derived entirely from logic based on premises. You could derive the Situation Model before there were videogames, or even before there were games.

As we come to understand the decision making process, we can also see how certain aspects of game design work to create Situations. For example, the primary difference between Chess and Tic-Tac-Toe with regard to decision making depth is that Chess is, thus far, not solved. That is, we do not know the single path that leads to complete victory (or guaranteed stalemate). Tic-Tac-Toe is solved; there is a single best answer which, if taken by both parties, is guaranteed to lead to a tie game.

One of the principle differences is that Chess has a wealth of decisions available to the player from the start. Tic-Tac-Toe, even in its opening move, only really has three (corner, center, side). So, as an ad-hoc rule, one could say that the number of decisions available increases the depth of the game.

Of course, looking at other games can show that this only holds when those decisions are actually important. If some decisions are clearly wrong, then there are fewer actually viable decisions.

This kind of information is an example of a posteriori knowledge. That is, it requires evidence from the world to understand. It requires a degree of experimentation, rather than purely bound to logic.

Until we find an a priori mechanism that can directly answer the fundamental question of game design, until we have a way to know the unknowable, we are going to have to develop rules based on a posteriori knowledge. We're going to have to look at specific instances and develop a theory of how game design becomes what the player plays.

Articles in the "Design Details" section will explore this kind of a posteriori knowledge in an attempt to deduce some kind of theory of game design.