Tuesday, November 29, 2005

Expectation and PAC

I wanted to write this essay yesterday, but I decided I had swamped the blog enough. That's good, because some of the commentary has made me refine the concept slightly.

Players get bored with the same challenges. That's pattern adaptation control's (PAC's) purpose; although in truth it is about finding the most efficient method of keeping players intrigued rather than simply "not bored".

The majority of my writings on PAC focus on changing the game level to create variations in the pattern of experience. For example, to keep your players interested, you offer a variety of challenges. Every level has a different architecture and decoration - but within the "overall pattern" of the game.

One of the things which is usually left out is expectation. Not because I don't know about it, but because it's a somewhat more advanced concept.

When a player is feeling eager and pulled forward, he is at his most interested. The most important factor isn't what the player is doing right now so much as what the player will be doing. However, this is also somewhat inaccurate, because what the player is doing now could be what he once will have been being doing. IE, this moment could be the moment he was looking forward to a little while ago. This lends it a kind of apotheietical power, to make up a word in addition to a grammar.

Think of a player's game experience as a squiggly line on a graph. One axis of the graph is time, the other is player interest. The line squiggles it's way across time, going up and down on interest.

Now, if you were to think about the really cool play moments - the moments when the level explodes, the boss fight, the jumping-off-a-building. These moments are only cool for one reason: they've been built up. Or will have been built up, if they are built up after the fact. (In this case, time can be considered to run in either direction.)

So, what you see is these blots on the graph representing these awesome play moments. Usually, the squiggly line spikes at and either just before or just after these moments, steadily tapering down in that direction. Like this:

Depending on whether you build up the event before or after it happens. You can do both, I suppose, but that doesn't make my purpose clear.

Looking back, I see my purpose hasn't been made clear, yet.

The pattern of experience isn't what we're manipulating here. We're not changing the game at all. What we're doing is telling the player about things to come (or building up things that have already happened), such that their interest in our game is peaked. Even if they're playing through part of the game they don't really care for, they'll continue on because they know what lies ahead - or what impressed them earlier.

This is why you'll play through that level you hate in that game you love - to a point. If the level is worse than the memories/foreshadowing is good, you'll ditch the game.

So, when you're designing a game's plot, you'll want to think about how you can inject the future and the past into the present.

A simple way to do this would be to put up signs pointing the way your next level and your previous level. For example, walking into an intersection that's labeled with arrows pointing "Science Lab" and "Armory". These are meaningful labels which give the player something to look forward to.

More usefully, clawmarks could be scoured on the side of the corridor saying something dire, like "I can feel it changing!" or simply "Mine. Stay out." This gives a much richer impression of what's ahead - spatially and chronologically - than a simple sign. The player feels a thrill of apprehension.

There are a million ways to do this. Overhearing guards talking. Finding a mutilated corpse. Reading a love letter. Seeing a new color of paint on the walls. It's simply foreshadowing. Building expectation, often only semi-consciously.

Except it runs backwards, too.

Some sudden events are merely to break up the pattern of play. For example, a little ambush to keep the player on his toes. There's nothing "behind" it. Right?

Wrong. Sure, you can have that isolated ambush. But it's much better - and not very hard - to write it into the story. In the simplest version, it can be used as a kind of foreshadowing: run into the zombies, you know there are zombies ahead.

But it can be rigged such that you "build expectation" backwards in time. You have an experience, and then learn why it happened and what its significance was. This is pretty common. In "Castlevania, Symphony of the Night", you start with a lot of powers and - bam - all your items and levels are sucked away. The rest of the game is spent recovering from it and learning why it happened. The first makes the event "ludologically" important, the second "narratively" important. Both are simply building expectation in reverse.

Imagine playing the game backwards. You start off with all this awesome shit and gradually lose it over time. Bosses alude to your past (future) as that time when you were badass. So you know this losing of weapons is only temporary, and sure enough, Death comes along and gives you some of the best equipment available just when you've lost your last, crummiest equipment. Functionally, it's the same as foreshadowing moving forwards, except it goes backwards.

It sounds arbitrary, as if it would be better covered under a different subject. But it's the exact same method, simply reversed in time. To a large extent, psychologically, people operate independently of the flow of real time.

I think it's interesting, and an important point. So, when you think about your design, think: "Where is the player? Where is he going? Where has he been?"

Because you don't leave the first part of your game behind. You build on it. Use it as foreshadowing and "echo" backwards to build it up. Like sloshing in a bathtub, these foreshadowing ripples forward and backwards in the game build on each other. Sploosh!


Patrick Dugan said...

This is where the quantum mechanics analogy becomes useful, I'd wager, its like non-locality can take place, connecting two events which don't have a direct causal relationship.

I wish I had something more insightful to say at the moment, I'll have to think about this some more. Very sharp.

Craig Perko said...

I tried to decide whether to put a more direct quantum physics analogy in, but I decided that would be rather too much geek.

You are, however, absolutely right. There's a lot of similarities between this and Feynman-era quantum physics, extending deeper than what I covered. :)

Patrick Dugan said...

Hey, remember that stuff I was telling you about taking the integral of a challenge function between two intervals, such as save points? Well thats just like taking the sum-over-histories in the Feynman multiple histories model, where the quantum system is the game's discrete parameters.

Craig Perko said...

You can theoretically use a system similar to Feynman diagrams to determine the way "play particles" interact, thereby determining the chances that a given player will play a specific way over any length of time.

However, I haven't had any time to work on it, so I don't know that theory resolves to fact.