My informal definition of procedural generation is a term for a program with a system of play or use with constraints where each time you run the program it results in significantly altered gameplay, mechanics, and/or visual output. The goal is to produce surprising outcomes that were not 'hand-coded' by the designer. Not surprisingly, this field is full of programmer-artists or programmer-gamers as part of the joy of these systems is that even the designer can be surprised by and challenged by their own programs; in fact, they are usually the best judge or player for their own systems built on their own goals and constraints.
In mainstream games the most commonly cited examples of procedural generation to produce surprising outcomes, mechanics, and environments tend to be: roguelites, generally a 2d realtime platformer; or top-down realtime strategy game with randomly-generated levels, mechanics and challenges (example: Faster Than Light); 2d partially-generated platformers like Spelunky; or generated seemingly-infinite worlds like the 3d space exploration game No Man's Sky. Other procedurally-generated games include a crop of card games of many types such as Slay the Spire.
These games are probably fine, but I'm less inclined to be interested in game mechanics especially as they relate to fighting or space exploration. The following are some games that signficantly feature procedural generation that I've recently tried.
I learned about this relatively recent game from the Eggplant podcast (previously: Spelunky Showlike). It's a deckbuilding narrative-driven speculative / magical realist game. You are the owner of a shop in a fringe desert community, the child of a matriarch that built up the shop and was part of a caravan of traders that wandered between towns and cities to find items to sell in her shop back home. Your mother has passed and you're now in charge. You take up your mother's position, joining the caravan's regular travels on the desert roads.
The mechanics of this game are truly singular. The game is vividly beautiful, with digitally-rendered tableaux, almost like generated cutscenes. The main mechanic of the game is that you select a person to talk with, you play a card matching game, and while the game is played you engage in a conversation. Depending on how well you and the person you converse with match, your conversation goes smoothly or can devolve. You may be rewarded in your successful matching/conversation, or not. And the conversations drive the strong story arc. Signs is a deckbuilder. As you complete a round of conversation/matching you select a previous card to destroy and choose a new one to add to your hand. The matching works somewhat like dominos, with some additional special effects cards.
The mechanic and strong narrative are compelling but the pace at times felt lethargic. I found the matching games challenging and tiring, but you are able to save your progress automatically, and come back later. So I played the game over approximately 4 sessions of 45 minutes to an hour. Even when I was consistently 'losing' the matching/conversations I found the narrative engaging enough that I did want to see what would happen at the end of the game. Would the caravan move on from my town? Would I build up my little shop? Would I find out more about my mom's journeys? There was enough there, and the procedural-generation was significant enough that I think I may try playing again sometime to see some other outcomes and try some other strategies.
One other thing worth mentioning: this was one of the few games to prominently feature women, non-binary and POC characters that I've played, and this played well in the story as well.
My big takeaway was how I loved the compelling story arc that changes based on gameplay actions, and I started wondering how this can be featured in other games and stories that make use of procedural generation. I highly recommend listening to the Eggplant podcast interview with the creator Dyala Kattan-Wright.
Moon Hunters is a 5-year-old game by Kitfox games led by Tanya X. Short, editor of the procedural generation tome Procedural Storytelling in Game Design and Procedural Generation in Game Design, with Tarn Adams of Dwarf Fortress. In other words, she literally wrote the book(s) on procedural generation!
Like Signs I thought this game was pretty unique. Kitfox describes it as a '1 to 4 player co-operative RPG personality test RPG in a rich, ancient world that's different every time you play.' This is a top down action roguelite. You pick one of four characters to play and it generates a mythology. The main goal is to travel and explore the world, in search of a missing Moon goddess. You can play solo, but I played the game perhaps 3 or 4 times of an hour or two each with a friend over the Internet in the co-op mode. I was pleasantly surprised by how well this worked.
The main use of procedural generation in this game was to generate the levels. This generation created rough-hewn outdoor worlds of enemies and loot, and often included coming across a special character who advanced the loose story. How you respond to that NPC (with kindness, with anger, etc) alters your character in some way, and advances the story. You also had a choice of options at the end of these levels, like crafting using food items or resting to regain health.
The pace of this game was much faster than Signs. It's not turn-based. In other words, enemies would attack and you needed to keep fighting, look for power-ups / health items or else. I died a lot, but the mechanic of bringing your fellow player back to life (like in Contra!) was really helpful.
Like Signs I finished one complete story-arc game at the end of 3 or 4 sessions. In a certain way I enjoyed the procedural generation but didn't always think it was significant enough. As I mentioned at the beginning of this post I'm not always the most excited about 'killing enemies' type games, and even though this game featured some story, procedurally-generated levels and strong female characters, it wasn't always enough to make the game feel significantly more interesting to me than other roguelites. For example, I have no interest in playing popular roguelite game The Binding of Isaac, so perhaps this says more about my tastes than anything else. But I was glad I played it. I don't think I'll necessarily play it again though, which for me is a sign that the procedural generation was interesting but ultimately the mechanics and genre wore me out.
Caves of Qud is an 'early-access' prominent Roguelike game begun originally in 2014, and though it uses tiles, it still feels the least accessible to new players of any of the games here. There are several dozen key commands, and I've played it for perhaps 6 or 8 hours in many smaller sessions, but I don't feel like I'm much of an expert at this game. There's many layers of strategy and knowledge that I have still to learn.
Unlike the previous two games, I'm still playing this one. But also unlike the previous this one really requires quite a bit of research. You have to read and watch some tutorials to even really get started. So you have to be committed. It's not a game for casual players in that sense. It's compared favorably along with Dwarf Fortress, which itself has long been described as 'the most complex video game ever made.' So comparatively Caves of Qud is also incredibly complex, though not as infinitely complex, and maybe half as difficult to learn! :)
Getting Started with Dwarf Fortress
The story in Caves of Qud feels a bit looser than Signs, though that may be because I didn't advance far enough and because you can wander the world rather than complete the quests. Essentially, you are in a magical realist far future 'after the fall' type game where broken technology litters the landscape and dangerous mutant animal creatures roam. There is some handcrafted content like pre-made dungeons, villages and quests. And that's mixed with procedurally generated graveyards, NPC's, items (like books), personalities, weapons. This game is the most like a traditional oldskool roguelike game. As a roguelike fan, perhaps that's why of the three I find this the one I'm most able to return to even though it's also the less groundbreaking.
The overworld is the main part of this game. You wander the map, occasionally running into snapjaws or other mutant creatures, looking for dungeons, where you descend on quests. I've never made it that far. You can try to do the quests or just wander around and enjoy the open world.
In addition to the game there is a CoQ discord server that is highly active. So much so that immediately after signing up I left, never to return.
This is a game I'll continue to play as each time I do play I find more to explore and enjoy and I'm getting a bit better with each play. I love crafting my body using the myriad mutation options, and seeing how I can do a bit better.
The devs are highly active, posting about new features on Fridays. Most recently they added more of an 'adventure mode' that is focused on exploring the world instead of pursuing quests. I haven't tried downloading this new version yet, it's a slight hassle for more to update, but I'm looking forward to trying it soon.
One other note: I found adding Caelyn Sandel's The Qud Survival Guide to be the (simple) add-on that really helped me get into the game. It's minimal but provides a nice scaffold into beginning the game. Highly recommended for your first time or two playing.