Considering this article is four years old, a lot of this didn't pan out. Spore failed to come remotely close to its hype, Borderlands' gun system turned out to be unpopular (too many guns were completely useless). Diablo 3 uses procedural dungeon generation, just as 1 and 2 did, but with plenty of predesigned set pieces scattered throughout.
The problem still seems to be that the set of things that you can generate on the fly and all result in something fun is fairly small. The other interesting thing is that we're seeing the opposite trend in terms of the growth of user-created content as a central game component (thanks Minecraft).
Procedural generation will certainly be used to keep games replayable (D3's dungeon and loot generation, L4D's director AI, etc), but to claim that it will replace designers in any way is silly.
Huh? Borderlands gun system was unpopular? Its as good as the loot system in any other Hack&Slay: some things don't really combine or do not fit your style, but technically, it worked really well. I see the procedural gun system as one of the biggest achievements of the game, especially because it yielded guns that were shooter-worthy and totally nuts at the same time.
That said, I think the problem is not only level layout, but designing interesting architecture. D3s levels could be played on white squares with properties, but a custom-built staircase makes the game interesting.
I thought people generally liked the Borderlands item system. They're keeping it for the sequel that's coming out later this year. There were some useless items, but that tends to be the case even when designers make the weapons if the game is the sort that boasts a huge number of them. Crap finds go hand-in-hand with big item catalogs, because the point of that kind of system is to create a sort of lottery.
The effect you're noticing is that it is a lot easier to get variety into random content when you don't have to provide graphics for it. It's also easier to make a world feel real (w.r.t. the baseline the game is establishing via its interface).
As a matter of fact, I'm in the process of writing an exploration and combat based space game, with an astronomical number of locations. (Much, much greater than billions) These locations will also contain procedurally generated alien technologies that can be incorporated into ship designs, resulting in emergent ship capabilities and gameplay.
Such a large universe also means that there will always be a frontier for someone to strike-out into and establish their own industrial infrastructure/empire. (Once certain thresholds are reached, such factions become visible in "Hyperspace" to other large factions.)
I know that generally, the cost of generating a thousand levels is roughly the same as generating a billion of them, but I just don't see this as being a huge selling point. Never mind that no one lives long enough to visit a billion things period, most people don't stick with a game for more than a couple years. In that time, how many locations will they visit? How many of them will turn out to be duds? And if e concentration of duds is high during a gamer's first 10 hours, will he/she stick with the game for even one month?
I know it's not a situation in which you can say, well, maybe I'll hire a dedicated designer...because cost is the issue with intentionally designed levels. Im just saying that the games I adore the most are the ones with just even a few levels that thrilled me, likely because of e human genius that pored laboriously over every detail.
I just don't see this as being a huge selling point.
The point isn't for someone to visit all of the worlds. One problem I experienced in Eve Online is that even with thousands of worlds, with a well populated MMO with powerful and well established factions, the new territories are gobbled up pretty quickly. I want to make sure there's always room to start building from scratch.
Im just saying that the games I adore the most are the ones with just even a few levels that thrilled me, likely because of the human genius that pored laboriously over every detail.
I aim to create a system where players can encounter such efforts, but generated by fellow players, not by a pool of devs.
Programmers tend to argue in favor of procedural generation, artists and designers against it. But the truth is, it benefits everyone, and does not necessarily replace anyone. I hate to break it to anyone, but almost every single game uses procedurally generation, either at runtime or pre-compile time. Did you use a filter in photoshop to produce that texture? Procedural generation. Did you perform a lathe operation on that curve? Procedural generation. Anything that takes a simple user action and produces a new (potentially complex) result via code is procedural generation. The more you empower artists and designers, the more they can do. Even if a game is completely procedurally generated, you need someone to tweak all of the parameters (material colors, building styles, character attributes, etc).
That bit aside, I am still waiting for a game to do more on the AI/plot/puzzle side of things in terms of procedural generation. As I have mentioned a few times before, emergent gameplay is key. Still, there is not one game with strong mechanics that has successfully employed emergent gameplay (aside from very basic puzzle/board games).
I never saw the point of procedural generation in games like Diablo. Seems to me something only programmers would be interested in. What would have been wrong with getting a level designer to do it? Then gamers could talk more about shared experiences.
It adds to replayability by preventing players from memorizing the shortcuts without making it feel like grinding. Many classic RPG's have a shortest path that's somewhere between 1/4 and 1/40th the average play-though and once you get close to it the game becomes a platformer vs. an RPG. But, randomize locations and you get to relive the hunt the hunt on the next play though. It also tends to make a wider range of things useful, vs letting people optimize the effort of finding stuff vs it's usefulness.
It is interesting that more games have not used this in re-release of older games. I can think off the top of my head 3 games that I want more content to extend the replay-ability.
-Mario-Kart with procedural generated race tracks after you beat all the levels and gain all the trophies.
-Multi-Player Halo. An option for one of the random games to have a procedural generated map. With a player vote at the end of the game. Top maps added to the rotation.
-Burnout paradise- Use the city generator to create random games. Again players vote at the end of the game. Top maps added to the rotation.
Bonus-Left for Dead and make it through the random city.
Notice the publishing date: Monday, 14 January 2008
Getting rid of designers sounds great from the point of view of the guy footing the development cost of a game. But it doesn't work. It makes for a very generic gaming experience.
The focus on nowadays is making a designer's job easier so they can be more productive.
I was playing games with "procedural content generation" back in the 80's. Heck I was making computer games that did it. Sigh. One of the joys/annoyances of ignorance and new generations is that everything old is suddenly new again.
Given that the entire website this article lives on is about roguelikes, and the author is also the developer of a roguelike (Unangband), I imagine he is likely aware of the history of randomly generated levels.
Considering this article is four years old, a lot of this didn't pan out. Spore failed to come remotely close to its hype, Borderlands' gun system turned out to be unpopular (too many guns were completely useless). Diablo 3 uses procedural dungeon generation, just as 1 and 2 did, but with plenty of predesigned set pieces scattered throughout.
The problem still seems to be that the set of things that you can generate on the fly and all result in something fun is fairly small. The other interesting thing is that we're seeing the opposite trend in terms of the growth of user-created content as a central game component (thanks Minecraft).
Procedural generation will certainly be used to keep games replayable (D3's dungeon and loot generation, L4D's director AI, etc), but to claim that it will replace designers in any way is silly.