Archive for May 19th, 2009

Some people remember the T-Rex or the lawyer from Jurassic Park.  What I remember most is the mention of Chaos Theory.  I understood the concept easily enough since I’d been tinkering with fractals since I was 12 or so, but before that film, I didn’t realize that there was a big “T” Theory behind it.  (Yes, that’s me, the art geek, figuring out how they did the effects and ruminating on science when any “normal” high school student would just be cheering the dinosaurs and wondering why Samuel L. Jackson was so quiet.)

Channeled Chaos

Channeled Chaos

What stuck with me the most was the notion of “bounded chaos”.  That’s more or less what fractals are, since they are recursive and iterative, generally staying within certain conceptual boundaries (most Mandelbrot “pieces” share similar structures, as do Julia sets, and so on).  They have bounds within which their recursion makes sense, so they aren’t completely irregular and chaotic (totally random), yet they have an element of randomness to them.  They are noisy (plenty of things going on, some that look random), but ultimately very patterned.  That dichotomy is fascinating to me, as a professional artist with deep interest in science.

To this day, Photoshop uses the Cloud and Difference Cloud filters, which bear a striking resemblance to some of the early 80s fractals that I experimented with.  Such semi-random (but repeating) visual effects are vital to making games, since we don’t have infinite texture storage space, but we don’t want things to look too patterned (since it would look too manufactured and obviously fake).  Repeating semi-random patterns are useful to add visual noise to a texture that repeats, breaking up the repetition and making for a more plausible final output.  The human eye is very good at discerning patterns, and it’s one of the game artist’s more interesting tasks to subvert that penchant.

Most computer generated art and “randomness” is actually very patterned (just by the nature of the medium; computers deal with bits with clear on/off states… true “randomness” isn’t a computer’s forte, though cryptographers make use of their computational power to fake it sometimes).  As it happens, so is a lot of nature, though it looks random on a macroscopic scale much of the time.

As visual beings, we’re used to this sort of “variable pattern” because we see it nearly everywhere.  From a distance, a brick wall may look very regular, but upon further inspection, each brick is different.  Our mind almost accepts this as a sort of visual shorthand, which is why we can cheat it often enough with repeating textures in games.  The mind is used to things looking regular sometimes, and assumes that there’s more detail than there really is.

One of the key derivatives of these notions is that variation does not require true randomness.  We’re actually OK with patterns, and in a lot of ways, they are how we survive.  (Witness the stress that comes when a daily routine is jarred, or big changes happen in life.)  We’re not actually built to handle a lot of randomness; it can paralyze us, and if life were completely random, we’d never really learn much, since cause and effect wouldn’t be predictable.

Short story long, as an artist and scientist, I’ve dealt with randomness and patterns for a long time now.  I see the strength in both (and the interplay between them), and am continually finding ways to use both for my projects.

This, of course, has ramifications for Game Theory.  The very nature of unpredictability is key to pretty much any game.

Mark Rosewater has a catalog of great articles, and a recent one caught my eye:

Kind Acts of Randomness

Mr. Rosewater argues that games require an element of randomness, but that too much randomness is detrimental.  Magic The Gathering is built on unpredictability, but many of the tools that players have are expressly designed to minimize or manipulate that unpredictability.  Some of the most powerful cards make things more predictable.  (Future Sight is one good example, as is Sensei’s Divining Top.  These are very powerful cards, and they deal with only a fraction of the unpredictability.)

Let me quote from the article:

Anyway, in the book, Jesse Schell defines numerous game design–related terms. His definition for fun was “pleasure with surprises.” His definition for play was “manipulation that satisfies curiosity.” And his definition for a game was “a problem-solving activity, approached with a playful attitude.” Can you see the connector between these three definitions?

The answer is the unknown. Surprises are things that are unexpected. Curiosity is a desire to learn about things you don’t know about. Problem-solving is finding solutions that you are unaware of when you begin. At its crux, gaming is about discovery of the unknown.

I can’t recommend the article enough, and this is the heart of why.  This is why I keep coming back to the notion that players need to have the ability to make varied and significant choices in their games.  They need to be able to try different things, experiment a bit, and not be punished for it.  A good game gives players plenty of things to explore and ways to approach the content, and lets them loose within parameters.  Bounded chaos works for gamers as well as planetary rings.

Put another way, there’s the popular weaselspeak phrase “think outside the box”.  This is completely irrelevant if you don’t know where the box is.  You have to know where the boundaries are and what the parameters of a situation are before you can take significant action.  Complete chaos isn’t all that fun since there isn’t any measuring stick of progress, and no way to learn.  Players can derive their own meaning, but that can be onerous.

This is why even “sandbox” games still have a box around them.  There are rules and regulations, predictable functions within the game world.  Players are cut loose to go crazy within those parameters, but truly random chaos is not given free reign.

I tend to argue that too much constriction in game design can be stifling.  A game that is completely predictable isn’t fun for long.  That’s why I rail against class-based, level-heavy, gear-centric DIKU MMO design.  To me, the bounds they set for their chaos is too tight.  I’ll never advocate for a complete free-for-all mess of a game, though.  That would break things just as surely as a game on strict rails (a movie with a pause button?).

Players have fun with variety and experiencing the unknown… but they also want to be able to master it.  Setting clear bounds on the game design, even if it’s just internal (not given to the players), will go a long way to making the game work.

Another book I’ve read on game design reminds that changing player “verbs” is dangerous; if a Holy Hand Grenade +5 does 5D4 of Holy damage, it should always do that much damage. Changing how things work arbitrarily is the epitome of “gotcha” design, and it’s usually very poor practice because it takes away the player’s ability to predict things in a largely unpredictable world.  (Though, notably, horror games take advantage of just that.  It’s just annoying in most games, but “gotcha+mood=horror” in that genre, and the subtle psychological manipulation of player expectations is key to most successful scary games.)

While we’re at it, even the xDy notation of dice rolling implies bounded chaos.  You know that a 3D6 die roll (3 six-sided dice) will always give you a total value between 3 and 18.  You can plan around that.  A 4D4 roll will give values between 4 and 16, a narrower band, which might be more preferable than the wider variation of the 3D6.  Similarly, a 1D10 vs. a 1D4 + 6 will function very differently, despite the upper potential being identical.

This is the backbone of D&D, and countless derivatives and sister designs.  The ability to tune effects using this sort of bounded chaos is a great shorthand for the risk/reward notion that makes choices interesting.  Is a 1D30 spell better than a 3D10 spell?  Statistically, a 1D30 spell has an equal chance to do high damage as it does to do low damage.  On the other hand, a 3D10 spell’s statistical distribution follows a rough bell curve, with an aggregate higher chance to do decent (midrange) damage than either low damage or high damage.  (And it won’t even do less than 3 natively.)  The choice between the two is significant enough to be interesting, giving power to the player, but not chaotic enough to make game tuning and balancing a headache, or annoy the player with a string of unlucky and underwhelming “rolls”.

So when I call for a classless skill-based system in MMO design, I’m not calling for chaos, I’m calling for more choices for players within somewhat wider “bounds” than the mainstream DIKU strain.  Not coincidentally, more player choice in a multiplayer game means more interesting unpredictability (within bounds) for other players, for better or worse (I think it can be better, but it needs to be considered carefully to channel it properly).  Also, when I write about Autopilot Character Development, I’m talking about letting players take the reins a bit and control some of those bounds.  In a system that has a great deal of freedom, ACD would be a way for players who crave structure to impose it on themselves (and to shrug it off later, if they so choose).  MMOs in particular can take advantage of such a variable system, since players have different play styles, and even a single player might want to change their approach at different times.

It’s certainly a balancing act, finding a sweet spot of effective bounds to give freedom without invoking pure chaos, and every game will have its own set of bounds relevant to its game design, but it’s important to understand why you’re doing what you do as a designer, and to realize that both randomness and control have hugely important repercussions.  Players need variability (not true chaotic randomness), and they also need power to control it.

Other articles that I didn’t quite work in, but are relevant:

Randomness and Replayability

Replayability 1

Replayability 2

(Controlled randomness is important for initial play, but maybe even more so for replay value.  It’s also important for keeplayability, despite the generally higher level of player control that I think keeplayability needs.)


Read Full Post »