This is something I thought up while tossing around a few plans for Down Under, the Abuse-like platform shooter I mentioned. I’m forcibly keeping myself off game development until A/Ls, but that doesn’t mean I can’t dream.

Procedural generation is not a new thing.  It’s been around almost since the first video games, when systems didn’t have enough storage space to hold that many levels. And I’m frankly obsessed with it, even though I know very little about the complex world of terrain generation algorithms.

So I was in class, sketching out a long-winded terrain generation alg. for sidescrollers [read: Down Under], thinking about how this one would play out. I had been thinking of using triggers and scripted events to maintain “a stream of mercenaries” jumping on the player, guns blazing – and I quickly realized that if I implemented this new process, out went almost all the control I was expecting over the player. It wouldn’t be a speedrun campaign anymore.
So my next question is, why not both? Consider a game where the main level, say a path from A to B, is one created personally by the level designer. Now this path, this level, has many offshoots, which are other levels spawned by a terrain generator.

And these levels, of course, are much higher-risk, higher-reward than  A-B path; this would satisfy both speedrunners, explorers, not to mention introducing a voluntary and more natural difficulty system compared to the “Easy” “Hard” “Insane” selection. It’s logical – stick to the main level if you want to pay it safe; head out and explore if you want – but on your own head it is. Here there be dragons.