I wait for this trend to grow further. Procedural generation offers higher replayability, requires players to adapt their strategy every time, and shifts the focus from map design to core gameplay, which distinguishes games from books.

Procedural generation has nothing to do with multiplayer. There are games that have competitive multiplayer and procedural generation, e.g., Heroes of Might and Magic III: Horn of the Abyss. Interestingly, this game offers both human-created and procedurally-generated maps, but players compete nearly exclusively on procedurally-generated maps.

I fundamentally disagree - games are simulations, and games that limit the simulation in favor of a planned “story” lose a lot of interesting nature of games, especially as they try to get closer to the hardcoded mediums - film, lit, etc. Procedural generation naturally ties into that simulation — though a lot of procgen games misunderstand its purpose, using procgen to “make the game for them”, rather than to produce various environments to interact with the simulation (they end up focusing on the procgen rules as an end to itself, rather than the simulation rules which is actually of interest). Borderlands is probably the most obvious failure in this — they entirely misunderstood the appeal of procgen/sim reducing it almost entirely to loot-gambling for % modifiers, with little to no real impact on the gameplay itself, except by accident (eg the difference between 1-shot and 2-shotting enemies). No man’s sky is another recent game in the same boat — they essentially had no rules to the simulation, and bet almost entirely on sufficient amount of “content” to be generated in an almost purely aesthetic fashion. Spore was also a failure in many, many ways, but the lack of real “rules” to its systems is one of them. It is the simulation rules that one wants to explore, not the visual rendering of it.

That is, Dwarf Fortress is the one true video game.

Multiplayer games on the other hand are like a game dev cheat code — you really don’t need to do much at all to produce a “fun” multiplayer game, because the player-interaction will bring all of the complexity and emergence facilities you need. All you need for multiplayer to work really is to not stop players from doing interesting things with the system, apparently a herculean difficulty for most companies.

I'm hard on story-driven games as well, I don't think games are very good at telling narratives because the narrator's desire to control what happens next is at odds with the player's freedom to choose what happens next.

The ideal game is somewhere in the middle where the designer focuses on mechanics which combine to produce more than their sum, and then anchor it in handcrafted content meant to show the best angle of the mechanical emergence.

One way to split the hair would be to say the mechanics are procedural (I would rather reach for the programming metaphor of composition, however) while the environments are hand crafted.

Nintendo is the king of this, but I also mentioned multiplayer games because they intentionally focus the designer on the mechanics and the levels while leaving the challenge generation to the players (and a whole conversation about how best to matchmake them.)

The key point I had meant to make is that procgen allows wide exploration of the simulation space -- which ofc is only interesting if the simulation is itself interesting. There's nothing wrong with that, but it only works with a sufficiently enabled sim -- Dwarf Fortress is the target here.

In a more shallow simulation, procgen falls on its face, and mostly wastes time as it fails to note any interesting aspects of the sim (or repeats itself, or only points it out once every so often). Hardcoded levels wins out here: its more concise.

That is, as simulation complexity increases, the amount of procgen can be increased. The scaling between megaman (weak sim, all hardcoded) to roguelikes (decent sim, mixture of hardcoded/procgen) to DF (sim porn, procgen porn).

Of course, you can also have a complex sim with hardcoded levels, but this is unsatisfying as it leaves a lot on the table -- you don't play a grand strategy for the campaign mode.

Nintendo games are intentionally fairly weak sims (few general behaviors, usually explored throughout the full game), and thus hardcoded levels works well -- there's little of the simspace remaining to explore when you finish a title -- so procgen would just be a noisy, verbose explanation by comparison

> But hand generated content is always superior

It's very much not a binary choice between procedural vs handcrafted. Procedural content creation is truly great at some stuff, while the same is true for handcrafted creations. The real trick is to blend handcrafted and procedurally generated in such a way that both tools get maximal utilization in the domain they excel in.