Whenever a player is playing, the game he is playing is a specific subset of the WHOLE game he CAN play. In games like chess, once through is a COMPLETE subset - that is to say, it is an entire instance of the game such that there is nothing to continue playing, but it is not the ONLY version of the game which can be played.
A typical computer game features you playing through numerous subgames. Perhaps each 'level' of the game, or 'zone', or 'fight'. Each one of these is an INCOMPLETE SUBSET. That is, it is only a tiny portion of what the game has to offer AND there is more to play RIGHT NOW. Once he plays through all the levels, his experience is a COMPLETE SET - he has seen more or less everything that can be seen. Replaying the game is, by and large, an excersize in repetition.
Obviously, some games have this issue more severely than others. RPGs have this problem VERY BAD, whereas multiplayer games like Starcraft have no real problem - their single-player modes are a pretty minor part of the whole game experience.
The big thing here is, at first glance, OTHER PLAYERS. If a game is multiplayer, the experience will change each time, and therefore you'll be able to play it over and over for many unique complete subsets.
But it goes DEEPER than that. This is where I'm dipping a bit into uncharted territory - all of the above is pretty much as most people agree.
Some games are different each time you play them WITHOUT multiplayer modes. Tetris, solitaire, pinball, Oasis, whacking a penguin head across a minefield. The sorts of games we play idly, just for kicks. By introducing an element of chance, we essentially 'simulate' a second, unpredictable, capricious player.
But these games don't get anywhere NEAR the heights of bigger games when it comes to emotional investment. Why?
You could say that randomness doesn't have the punch of a multiplayer game. You could say that they play it for less time than an RPG, so there isn't very much time to build up a stable meme. Are those the reasons?
I really don't think so. I know people who've played Tetris or Lumines for longer than they've played any RPG, but when you ask them what game they have the fondest memories of, they're unlikely to whip Lumines out of their pants. And, frankly, I don't give one stale donut about how long a player plays my game, so long as they are impressed enough to want the NEXT one I produce.
The inherent problem with these randomized games - such as Lumines or pinball - is that their rewards are shallow. Their rewards are shallow. Sure, you can continue to get dozens and dozens of different variations in the pattern, but so what? Ten more points? Five more seconds? Sure, it's something to aim for, a reason to keep playing, but you're not going to keep pulling in intense interest with a reward system like that.
In a game using multiple players or an advancing story, they have deep rewards. Multiplayer rewards are pretty clear: it's in our blood to prove dominance over others of our species, and multiplayer games link to that directly. The deep reward is a deeply satisfying shout of "Who's yer daddy?" This reward rarely gets old, as ten thousand years of games and contests can attest. This may be a distinctly masculine reward - or, more likely, women simply have a different preferred vector.
Storylines ALSO give clear and deep rewards. That cut scene where the castle rips away from the ground? Yeah, that FREAKING ROCKED! Now we're in this flying castle and DRAGONS are attacking us! Hell yeah! Whoa, watch out!
The funny thing is that storyline rewards ALSO use the same, core domination buttons.
Let me demonstrate: perhaps you remember playing Halo 2? If you were one of the ones who DIDN'T hate it, what was your emotion upon cutting the cords holding the sky city in place? Unless I'm sadly mistaken, they were essentially "Whoa, COOL! I ROCK!"
Okay. Now I'm going to start to run. Remember the thing on action I keep harping on? Action is nothing more or less that a chance to prove your power (via your character). But in order for it to MEAN something, the enemy you've just defeated (whether it be a snot zombie or a puzzle) has to have a known power level. One of the major difficulties in 'easy' games with stories (such as 95%+ of games with stories) is to hide the fact that the enemy you're defeating SUCKS ROYAL BUTTOCKS.
This is usually done in two ways: by requiring you to have 'special' powers while fighting him or by 'proving' his power in events (both immediate and historical) during gameplay.
For example: I never had a real hard time with Kefka. But he's my favorite badass. He broke the world in half, he killed the general, he kept an army of robot slaves, and so on. When we got to fight him, we attacked him in WAVES in order to stand a chance! Of course, the fact that we've never fought in waves before means that this is REALLY IMPORTANT! right?
RPGs spend ALL THIS TIME building up these enemies and allies, solidifying their patterns and bouncing them off one another so that you can get this "Yowza!" fight-flight-I rock thing going.
Multiplayer games don't. They rely on the fact that all the players are very familiar with the patterns involved. The patterns ALREADY have power levels built in, so when the other player applies them as he sees fit, the first player can judge how much trouble he's in and what kind of player the other guy is. This is APPROXIMATELY the same approach as an RPG!
BOTH kinds of games create stable patterns and give the players their excitement by mixing established patterns in cool ways. The RPG builds these patterns step-by-step out of carefully placed building blocks. The multiplayer game builds these patterns by creating a game which can be replayed a zillion times, each time reinforcing the pattern.
The RPG uses COMPLETE SETS, whereas the multiplayer game uses COMPLETE SUBSETS. There is no difference in how long the player plays - the difference is merely in whether he plays it as ONE game or one MILLION games. Of course, a game which has a suitably wide possibility set will always be ABLE to be played longer than a game which has only one possibility set - but the player is unlikely to WANT to play them that long.
The question is, can this be done BY HEURISTIC rather than by the human hand?
In multiplayer games, the answer is NO. If you can map the whole of possibility space into your artificial intelligence, your possibility space is too small to bother with.
But can it be done with a STORY?
Okay, everybody and their pet dog has tried to make heuristically-generated stories. If they come even close to succeeding, it's always a painfully boring story. It never works. I'm NOT trying to do that.
I'm trying to make it so that an algorithm can build up the various characters and keep the player entertained along a story.
For example, you tell the game, "The princess has been captured by a dragon. You have to hunt down the dragon and kill him. But at the end you discover that the princess moved into the cave and kicked the dragon out."
The game would then extrapolate. It would have an engine, and graphics. It would build you a character and some abilities. It would create a series of levels and upgrades. It would use written story points, but the actual gameplay would be created on the fly, specifically honed for that particular player's play style.
The thing I like about this idea is that the player could specify a length. He could say, "only two hours, please!" or he could say, "give me a 40 hour game!" The game would then go off and doodle out some subplots and secondary characters.
Essentially, you teach the computer to play improv jazz.
Yeah, I know this is a really expensive and possibly impossible idea. Fortunately, there is nothing I am better at than reducing the scope of an issue.
For example, make it a futuristic racing game. You specify the 'length' of the game, which the system extrapolates out into a specific number of divisions (or planets) and stages (which can be created by heuristic). The computer also creates a large number of cars by heuristic - more cars for longer games.
The computer then creates a number of characters, either by accessing a character database or by heuristic.
Suddenly, the improv doesn't seem so far off. By keeping track of players and having a fairly large database of commentary, the computer can create rivals and friends for you, and doodle little stories about their troubles and triumphs around you. Simply by keeping track of who does what to whom and who wins over who, you can make characters 'learn' who to hate and who to trust. In a self-fulfilling prophecy, they are more likely to help their friends and hence become closer friends. A number of subplots - such as assaults, technological development, damaged vehicles - can be selected as needed to spice up the action.
While I would not presume this doodling would match a carefully planned out story, that's the cool part: it could COEXIST with a carefully planned out story! It could serve to lengthen the game to the correct length WITHOUT adding stupid crap. Indeed, even the carefully-planned story could have optional story points which only come out if the player is playing a long enough version of the game.
Of course, this would be an engine thing, not a particular game thing. You'd want to be able to reuse it for your next five or six games.
It's obviously total blue sky, but have you any comments?