Join our Discord!
Cart
Want it before X-mas? Order by Dec 13-15! (US estimate, not guaranteed)

Where Procedural Generation Fails

I'm in the middle of playing Subnautica and it struck me that they avoided a massive trap that recent games like No Man's Sky plunged head first into: Procedural Generation (PG) sucks at creating interesting content.


There's no doubt No Man's Sky works better as a screenshot creator than a game

Game developers have a problem in the PC space in the form of rising expectations. This, in short, means rising costs. If you want your world to be massive or you want a lot of content cheap, generating it with an algorithm seems to make a lot of sense. However, if you want to craft a cohesive world with interesting and unique places to explore, well, lets just say I haven't seen too much success here.

Ascent - The Space Game has 270 billion stars. Unfortunately few of them are worth visiting, which is realistic, but dull.

That isn't to say Subnautica used no procedural generation in its development. It probably did. The best games that make use of PG do so with a pretty simple formula: Craft interesting and unique areas and then connect them with PG content. My guess is this is how the underwater world was created. There wasn't some poor guy hand placing every rock and coral, but the big things, the hidden caves, the biome changes carefully placed at depths dangerous to a new survivor, and the hidden secrets of the world, those were undoubtedly crafted and placed by hand.

I still haven't been able to safely explore this area... I'll get there eventually.

I return to one of my favorite PG games, Dungeon Crawl Stone Soup (in my opinion the best true rogue-like). They generate a new map on load, but then place pre-set interesting locations throughout the world into that. While the endless corridors and open spaces are where you spend the bulk of your time, the first hundred or so play-throughs will constantly having you discover weird and unexpected events and places.


This is a legit castle with a moat and gnoll guards that may appear somewhat early in the game. Maybe.

The point of all this is to ensure we're looking for the right kinds of procedural generation. Beware of any game that makes too heavy use of generating content with an algorithm over crafting a cohesive world. Odds are the developer is cutting corners or just doesn't understand that changing a color from red to blue doesn't count as having crafted a new world.


Previous Post Next Post