Big Brother is watching you playPublication date: 16 July 2011
Originally published 2010 in Atomic: Maximum Power Computing
Last modified 03-Dec-2011.
How can you figure out what people like?
But it's not actually, usually, that easy.
The publishers of the magazine in which this column first appeared, for instance, have only a hazy idea of how many people read it, and how much attention readers pay to the ads. That last bit matters a lot to the publishers, and to the advertisers, but the ads aren't actually why people buy the magazine.
(Concealing this fact from the advertisers is vitally important. Thus far, no publisher's Circulation Department has let the terrible secret slip.)
You'll find similar weirdness in every other medium that requires ad money to survive. Commercial TV, newspapers, most Web sites. The question of what people really like - or, at least, what people can really be persuaded to pay for - is at the absolute neutronium core of the inescapable sucking singularity that is the advertising industry. Which has, by and large, ignored the issue, because ad agencies can make more money by just advertising themselves to potential clients.
Web sites have the great advantage that they don't need to poll some hopefully-representative sample of their readership to figure out whether the site should bother to stay in business. You can beat some sort of statistical rigor out of most poll data if you really try, but there may be surprisingly little connection between what people say they like and what they actually choose to do, just as people who buy lots of porn may also profess to be outraged by indecency.
If you run a Web site, though, you can just look at server logs and see exactly how many times each page, each picture, each ad, has been requested. Various levels of caching blur these numbers, but compared with the horrible guesses that hide the sausage factories where they make TV ratings and Top 40 charts, Web stats are as pure as Euclidean geometry.
Now, though, computer and console software have created an even better kind of user feedback: User behaviour tracking.
Lots of software phones home, occasionally with the user's permission. So the mothership knows that 43% of listeners use random play, or that nobody ever seems to use that amazing new word-processor feature that took months to develop. And now a more interesting kind of user tracking has emerged: Player tracking, in games.
Even if a game has no online multiplayer, it's now likely to be played on a computer or console that's connected to the Internet, and that makes it possible for the game to snitch about what the players are doing.
But player tracking has so much more to offer. The publishers can see, for instance, just how alarming is the percentage of people who never make it to the end of the game. They can see what kinds of characters people like playing, what kinds of decisions are most likely to end in the player restarting (or just quitting) in disgust, what things people do a lot, what things they almost never do, what special secrets almost nobody discovers, and so on. If we ever get truly episodic games, I wouldn't be at all surprised if the episodes had TV-like ratings, except with a lot more data than just one big dumb number.
Not many players are bothered about the fact that some games now collect this sort of data, unless they're worried that the anonymising isn't very good and other people might find out how much time you spend sniping old ladies, selling faithful companions into slavery or making amusement parks that no visitor ever leaves.
The concern, rather, is what effect all of this data will have on future games.
Suppose a game company discovers that almost nobody makes it to the end of their game, and that distressingly many players give up very early indeed. The light-side response to this information might be to, well, just make the next game more interesting, reduce padding and level-grinding, add a "casual" difficulty level, and so on.
The dark-side response is simpler. If few people played to the end of Generic Manshoot 1, clearly one should only expend any effort on the first half of Generic Manshoot 2, especially the first bit of the first stage, which is all you can play in the free demo. The end-game can be a bunch of repetitive corridors containing palette-shifted, hit-point-increased versions of previous baddies. Final cinematic? Nah, a wall of text'll do.
Some degree of the dark-side response is already on show in all sorts of RPGs, where there's a forest of interesting side quests in the first two-thirds of the game, and then it all focusses down into one Very Important but Not Necessarily Very Fun endgame path.
Fortunately, I don't think this is actually a major problem.
Well, not a new major problem, anyway.
Several big-name studios already make linear, samey games filled with nice safe slight modifications of stuff from previous games, and slightly better effects. Look at the endless parade of shooters that all, now, have regenerating hit points and single-player campaigns only slightly more interactive than Dragon's Lair. They're big, they're loud, they're all much the same; they're Interchangeable Blockbuster Movie Games!
(Done well, this strategy can work. StarCraft II may really only be StarCraft 1.5, but zillions of people - some of them not Korean - loved SC1, and SC2 gives them more of everything they've loved.)
But the existence of interchangeable blockbuster movies has not prevented the creation of more interesting films, just as the existence of top-40 zillionaire musicians has not prevented the creation of more interesting music. There's an absolute explosion of high-quality home-made music on the Web now, movie-making is way more accessible than it used to be too, and the outrageous horsepower of modern PCs means indie game makers no longer need to be godlike programmers to make a fantastic game.
(If you tried to tell me in 1996 that, 14 years later, a game written in Java would make its creator millions and millions of dollars before it even came out of beta, I would have immediately searched your pockets to see if you had any more of whatever it was you so clearly were on. And yet, there's Minecraft.)
We've already got international ad-hoc project collaboration, motion-capture hardware you can buy for $US150, synthesised orchestras, synthesised singers and moderately realistic synthesised speakers.
At the moment, stitching all of this stuff together into something that'll let you make a big action game in your bedroom is sort of like desktop-publishing a proper magazine was 15 years ago. You can do it, but it's a pain, and you'll be spending more time working around the limitations of your tools than actually making stuff.
This situation is changing fast, though, and we're already seeing some spectacular results. They're mainly in mods rather than whole games, but the better mods pretty much are a whole game already, and the line between mods and sequels is sometimes blurry anyway. Making a proper game will never be truly easy, unless we come up with artificial intelligences that can pretty much do the whole thing for us. But it still may not be long before creating a quality game will require no more investment of money and effort than an amateur dramatic society's production of H.M.S. Pinafore.
So even if the big game studios do want to market-research their way to games with all the wit and creativity of reality television and mainstream politics, who cares?
Pretty soon, some bloke down the street'll make something better for you to play.