Tobold's Blog
Wednesday, May 31, 2006
 
Content generation

A player of World of Warcraft can play through a zone, do all quests there, and "consume" all of the zones content much faster than in took to create the zone with all the content. That is a problem shared by all MMORPG, but different games have come up with different solutions to generate more content than their players can consume.

One exercise for programmers is to use Markov chains to generate random text, which at a surprisingly low order already much resembles really written text. In MMORPGs the equivalent is using Bryce-like software to generate random landscapes, and random quest generators to produce content for the zones thus created. Anarchy Online has both random landscapes and random quests, so does Star Wars Galaxies. City of Heroes / Villains and Anarchy Online have random dungeons. The obvious advantage is that once you have created the algorithm to produce a random landscape or quest, you have infinite supply and never run out of content. Unfortunately the result is often rather bland. A planet in SWG might have enough landscape to run around for hours, but unfortunately it all looks pretty sameish, and becomes boring to the player very quickly. The same is true for random quests and dungeons, they tend to be all rather similar, and just aren't interesting enough after a while to hold a players attention.

A second way to create content is to let users create the content. The extreme example of that is Second Life, where basically all the content is user-created. Games like A Tale in the Desert (which just started its 3rd "telling") work by mixing developer-created content with user-created content. Unfortunately user-created content suffers badly from Sturgeon's Law: 90 percent of it is crap. Also the number of players willing to create content is a lot smaller than the number of players wanting to consume content, as creating good content is actually hard work. The mixed approach works better than the pure user content approach, and user-created content tends to be better than random content. But neither approach ever produced a smash hit MMORPG.

So in the end the best method of content creation is to pay a lot of game developers a lot of money to create a lot of content, and then try to add content at a rate high enough that you don't lose too many players who have already seen everything. If there is one thing in which the original Everquest excelled even modern games, it is the amount of high quality, developer-created content, and the rate at which is was added over the years. It is no accident that Everquest dominated the market at that time, just as World of Warcraft dominates the market now, because both games are based on having a huge amounts of developer-created content of a very high quality.

Unfortunately that means that games without a triple A budget don't stand much of a chance to contain enough content to keep players happy for a long time. But it is to be hoped that one day game companies learn that infinite monkeys don't write good MMORPGs, and that user-created content can play a part, but not carry the whole load of content generation.
Comments:
The problem with random generated content also means that you're locked in to some form of tile/block based level design. One of the biggest draws of WOW has always been the pretty obviously hand-generated zones, not just in terms of art assets, but in terms of how the quest lines lead you through that space (The few unfinished areas are pretty obvious- Dustwallow is probably the most glaring, with areas that just don't have any discernable point to them).

User or algorithmically generated content also makes a heck of a lot more sense in the context of small games, I'll grant that. In fact, I'd almost say that you'll want to use it only if your game has modest technical aspirations (2d graphics is actually ideal- line of sight and pathing exploits seem much harder to root out if you've got height to take into account) And if you're looking to use either of these as a solution to content generation, don't kid yourself that developer verification can root out the bad or buggy stuff. You're implementing these solutions because you don't have the manpower to make your own content- what make you think you'll have the manpower to check your users' work?

(An additional note- map randomization is probably best in games with highly tactical natures- Diablo's random maps rarely mattered, because there were only a handful of situations where my location in the gamespace mattered- the only variables were the avatar's position, the relative position of the enemies, and a handful of chokepoint strategies. An online version of, say, a hex based combat game? Suddenly a map change becomes more than cosmetic, and actually introduces gameplay changes)
 
We still are years away from compelling random generated content. Problem is, that until today no one found a way to generate really procedural contents. Spore may do a big step into that direction though, we will see. As i see it, it takes more time to create a 100% procedural base, than it takes to craft handmade individual contents. So there is your risk, wich everyone hates. If Spore fails as a consumer product, there are years and millions gone to waste.

Handcraft will always be superior to machine made stuff, i hope. But lets take Lost dungeons of Norrath as an example. Sure those cut&paste dungeons were no Guk or no Chardok, but they offered huge amounts of fun too. I prefer a Scholomance wich is generated from a set of pieces and looks a little different, each time i enter it, than a Scholo wich is the same one i entered years ago.

I do not expect a complete original designed, itemized and quest added zone, on a daily basis, but some kind of difference is better then nothing. Tint the texture sets of Scholo, mirror the layout and change the NPC models and you have something, i will play with more fun, than the original. These are easy and cheap solutions with great payoffs.
 
Post a Comment

<< Home
Newer›  ‹Older

  Powered by Blogger   Free Page Rank Tool