Procedural generation has had a major impact on game development in recent years. It allows developers to produce vast landscapes like in The Elder Scrolls V: Skyrim (2011), galactic simulations as in Elite: Dangerous (2014), and infinite replayability in roguelike-inspired games such as Spelunky (2009). Procedural generation is a means for major developers to break restrictions on what games can achieve, whilst also being a cheap and effective tool for indies to add more content to their games.
But has it changed the games themselves? Procedural generation is often used as a replacement or extension to the traditional level designer. It’s a shortcut to greater variety or a mechanism to expand boundaries. What impact does it really have on the gameplay, and can we be doing more with it?
Procedural content was used in the early games industry as a way of bypassing physical memory limits on machines. The original Elite (1984) was able to contain over two thousand stars and planets, each with individual economies, all in the BBC Micro’s 22 kB of memory. Rogue (1980) randomly created its levels anew every game because there was no space on computers at the time to store detailed level designs. In the modern age these technical restrictions are no more, but procedural content generation remains an invaluable tool.
For decades roguelike games have consistently relied on procedural generation for their gameplay to work at all. Since the players start the whole game anew after each failure they need a new level designed each time they play. Developers cannot hand-design thousands or millions of levels for the content needed, so they teach the program how to make the levels. As roguelike elements spread into many indie games, we see the same techniques now widely deployed. Yet how different are these computer-generated levels from those the designer would create by hand? Some games like The Binding of Isaac (2011) simply randomise pre-designed levels, or games like Spelunky stitch together pre-designed segments instead of designing from base elements. Playing a single instance of this is not so different from playing a non-procedural version.
Though on the surface it seems like the gameplay in Spelunky is similar enough to traditional platformers, at the mastery level they are worlds apart. Experienced Spelunky players know where the exit will be once they enter a new cave. They are so attuned to the generator, to the way levels are created, that they gain an intuition for how the game elements are laid out. A speedrunner for Super Mario Bros. (1985) will have learned the precise jump points and power-up locations for the perfect run; a Spelunky speedrunner learns the underlying rules and systems which make up the game itself. It is the same across many roguelike games—the gamers learn the rules inside-out, and though they can never predict everything they can weigh the odds and game the system. The procedural generation changes the way the player thinks about the game and plays with its mechanics.
Can we go further than this? Some games pull the player deeper into their procedural systems, requiring mastery of the procedural content to work at all.
Spore (2008) and Black & White (2001), for all their flaws in implementation, both created new avenues of gameplay for players to explore. The procedural generation of creatures, down to their looks, behaviour and culture, becomes an active part of the gameplay for the player, far more engaging than choosing features from a list. Directing the evolution of your creature or society becomes the gameplay, and through this we see procedural content become the output of play, not the input.
Mushroom 11 is an innovative indie game still in development where the control system relies on the manipulation of a set of cells. The cells grow and expand based on the player’s manipulation, in a pattern known to many procedural designers as cellular automata. The game is surprisingly intuitive to play, pushing the cell group into different shapes to fit through tunnels and solve puzzles. At its heart the game invites the player to experiment with a procedural system, and the player will come to experience the procedural generation as a natural part of the gameplay.
Some board games also draw the player into playing with procedural systems. Hey, That’s My Fish! (2003) and Hive (2001) both involve players changing the game grid with their actions, making the players act like cycles in a procedural algorithm, modifying the board according to the rules and system of the game. For the players this is the gameplay: they are at the heart of the generation. Board games in particular show this well, as they require the players to act out the rules of the game—there is no computer to move the pieces for them. They need to understand the systems at work in order to play at all, and in learning the rules they learn how to game them. This is where the new gameplay arises, the change in play based on dynamic generation which we as players can affect.
We see snippets of similar player manipulation of game elements in other games: the control of lava flows for gameplay purposes in Dwarf Fortress (2006), or the creation of reactive systems and even computers within the world of Minecraft (2011). Both require the gamers to know and play the game at a new level of understanding. Yet we still have few games fully incorporating procedural generation, making it central to the gameplay, and we have especially little from AAA developers.
There are wide and open spaces to explore here. As procedural content becomes more common, and in particular as indie developers experiment more freely, we will see bolder and more unique games embrace procedural generation as core to the gameplay. And in doing so they will draw gamers to be far more invested in the gameplay and the mechanics of the game.
Procedural generation can be so much more than an extension of traditional game design. It can be a new branch of gameplay and a new way to game.
Join the conversation! Comment on this article at This Old Neon’s community forums.