This text taken from zileas.com, now site moved to shambler.net Author - Zileas
Now i can't find it there, so i post text instead of a link.
Lecture 2: Rules and Game Dynamics: Game Design as Engineering
Game design is generally thought of as an art form, but in truth, only parts of it are. It is essentially a hybrid engineering/art like architecture for instance. Only some designers in the industry recognize it, or at least will say so, but those that do have produced some of the best games. I'm completely certain that Looking Glass Studios and Blizzard Entertainment both use this approach, among several other less known quality developers.
When making a game, its important that you have central objectives -- namely in the form of gameplay. What is it that you want your game to play like? How do you want to center the interaction? These concerns are very important to answer before you start addressing problems like "What sorts of tanks should I have?" or "What should the story be?". All too often designers will stumble into a project with some general idea, and start fleshing out the "fun stuff", like specific minor details of the story, and the abilities of various characters or units, or whatever, and then later on decide how the game rules work, while retaining things that were never thought through. In a nutshell, I cannot emphasize enough to engineer the framework first, THEN fill in the gaps and flesh out the details.
Generally, you engineer the framework of a game by making rules which create the desired game dynamics. Rules are obviously the mechanics of the game -- how far can an arrow fly, how does something die, what cards you can trump other cards with, etc. Game dynamics, which contribution to the creation of the gameplay in some way, are your ultimate objective, and come from the use of rules. A game dynamic might be something like "high ground is advantageous", which could come from a range bonus given to units being higher than other units, a downhill charging bonus to attacking downwards, or a penalty to climbing a hill (which makes ranged troops more effective). This would together with maybe other types of terrain-related game dynamics create a dynamic of terrain impacting strategic choices, which would be part of the key pieces of the gameplay. Putting together game dynamics results in your game focus, or gameplay. Also, game dynamics can, together, create other game dynamics.
You should try to figure out what game dynamics you need to get your desired gameplay, and then figure our rules that work to accomplish them. For instance, when making Quake 2, ID software no doubt decided that they wanted every weapon useful for a different situation. In order to do this, they thought up ways to accomplish this, and this ended up being a reason to add the grenande launcher (a weapon to shoot around corners), the rail gun (a sniper weapon), the rocket launcher (An area effect weapon for foes that are hard to hit), or the shotgun (A close range weapon). Although many of these effects are "realistic", ID tweaked them to bring out the desired characteristics.
Generally, if you look at a famous game, you should be able to pull out the game dynamics and the associated rules, and from there figure out what their goals were. A lot of the time you'll find that some things dont seem carefully designed, and a lot of the time you'll find a badly balanced game dynamic that would've been great end up breaking the rest of the game.
Examples of game dynamics:
An Example of a set of dynamics:
1) Myth: The Fallen Lords, developed by Bungie Entertainment, is in essence a game about tactics on a 3d battlefield. They chose the fantasy setting, which intrinsically makes a lot of things more believable and thus made things easier for them from a gameplay perspective. In order to accomplish the desired gameplay, they added the following game dynamics, based on certain rules:
a]Dynamic: High ground is advantageous
-Hard to climb
-Ranged units like archers shoot farther when shooting down
-Sight bonus (can see farther, allows better reactions
b] Dynamic: Units have clear counters
-Melee units can always outrun ranged units
-Short range artillery units (dwarves, fetch) extremely effective against
melee units, bad against other ranged units, as a result of their speed
and range (I.e. this is a dynamic supported by those rules).
-Clustered units prone to ranged units, non clustered units not prone.
-Skirmisher units (ghuls) are particularly effecitve against archers and
dwarves, but horrendously ineffective against anything else.
c] Reward mix-and-matched strategies that require creativity
-Melee units support ranged units well when ranged units placed properly
either as shields (delaying actions) or protection vs skirmishers.
-Short range artillery very good at softening up melee units, but can
seldom finish them off.
-Ranged units do well when melee units "lock down" other units by engaging
in hand-to-hand combat with them. Ranged units hit stationary targets
far better due to their realistic projectile movement. Also, their
arrows have scatter (like all ranged units), so they are again,
not as effective against non-clustered units (but clustered melee is
great against unclustered melee
There is more to myth than that, but that gives some idea as to how it becomes a game about tactics on a 3d battlefield.
An example of a slightly undermined set of dynamics:
1)Homeworld, by Relic Entertainment, is a great game, but some of its game dynamics are weakened slightly by a few problems. It was an extremely ambitious project, and its a miracle that it turned out as well as it did, and relic entertainment really proved itself in its creation. However, like all ambitious projects that cover a lot of ground, nothing was perfect. Three dynamics I am analyzing in this case are
a) The idea of a unit counterability circle (Complex rock paper sciscors basically), which I explain below
b) The idea of being able to choose strong counters to build against other unit mixes, as this is primarily what an RTSes strategy is about (that and tactical control)
c) The idea that all of the game, including the early game, should have strategic diversity, as that entails choices, and thus real interaction
Most of these were accomplished fairly well. However, they were weakened by either rules that were not thought out as well as others, or simple realities of the ambitious nature of the true 3d RTS they created. In general, some features and rules when not carefully thought out can sabotage your desired game dynamics. One of the biggest difficulties in game design is spotting these problems before its too late to easily correct them. This is what Relic fell into, but the game turned out being a groundbreaking title anyway

It is far easier to see the few problems in an otherwise perfect game, than to even begin to start a coherent discussion about the flaws of a really bad game. That is why I use Homeworld.
a] Unit counterability circle flawed
-Fighters do well against capital ships. Corvettes do well against fighters
Capital ships do well against corvettes. However, super-capital ships are
cost-effective against everything, and one of the super-cap units is
extremely effective against fighters. The result? Corvettes end
up being a bit less useful than other units, especially the multigun,
heavy and light corvettes (Salvage, repair and minelayer have special
functions that put them outside the true corvette dynamic, and into a
set of their own dynamics)
b] Production discourages choosing strong counters
-Your starting production capability (which is all you will have for at
least 30 minutes), can produce any one item at a particular speed. If
you produce more than one item at a time, they all produce at their
normal speed. The result? In order to spend your income, you produce
multiple units, sometimes when you only want a lot of one, or two. At
its surface this forces unit mixing, however, so its not ALL Bad --
but it DOES undermine the players ability to choose strong counters, and
unit mixing can be innately balanced (encouraged) into the game, which it
is in other ways in homeworld..
c] Early game static (usually)
-Map transverse times tend to be very long. Prevents many early game
gambits from being effective
-Because of dynamic b, in cases where you can apply early game gambits
such as various rushes, you cannot produce covettes (which counter early
fighters) in large enough numbers to make a difference... so you counter
with fighters... effectively resulting in no strategy choice, and a
static early game
Overall homeworld is an excellent game, even with the flaws I've mentioned. It is so new, and yet so playable, that such small errors can easily be overlooked. However, in future versions of the same sort of game, no doubt such errors could differentiate the title from other titles without such errors. Since homeworld is one of a kind, and those errors are far from severe though, it is still a good game.