Was procedural generation ever done right?

Yes. Some things work well with procedural generation like maps and depending on the games design even missions can be generated well like DF adventure mode.
The problem is incompetent devs shoehorning it in where it doesn't belong.

It doesn't work that way, take DF missions for example where 100% works fine but you wouldn't want every piece of armor procedurally generated as it would be impossible to keep track of what armor was best when it's all created randomly every time you start a new game.
So if the map, NPCs, missions, and factions are procedurally generated but the items aren't what % would you call that?

Both, they put a lot of effort into making it produce believable designs without doing stupid shit like making a single corridor the entire map.

Personally I love procedural generation when it's done right and I'm a little worried that with all the devs getting it wrong at the moment the money men might start to think the problem is with the generation itself instead of with the poor implementation.

As I see it, procedural generation is like a spice in a dish. Some dishes benefit having a bit of it, others benefit from not having it at all, but nobody would put a spice as the main ingredient in any dish, unless they're a shitty cook.
Procedural generation can be fun and be good for a game, bu only if you use it sparingly or where it works fine. The main problem, you got dumbasses who think that shoving a few assets into an algorithm would make a game for them, but that's far from the truth and we've seen it time and time again.

...

ARPGS ? Minecraft, that fucking nigger notch has X billion dollars to prove it so

It can be great. Problem is, it takes skill, and a shitton of effort to make good pieces, and enough pieces, and then layer them together properly, and then address the exponentially increasing pile of incompatibilities as more and more shit has to work together, and all of this has to be in the kind of game that benefits from being random gen.
This is a lot of work, only feasible for one person to produce if it's nograffix. In modern AAA vidya age, where shovelware is king and passion is dead, a good and good-looking procedurally-generated game cannot be made. So only turbo autists like DF's creator make good PG, and it always has to be nograffix.

Wasn't most of Daggerfall procedurally generated?

Leaks of Bethesda's Starfield said that it will be something like Daggerfall in space with space stations inspired by Deep Space 9. The thread was on neofag 8 months ago and deleted for no reason. Let's see how it will turn out like.

If it's touted as a main feature you can bank on the game being shit. Relying on procedural generation to make your game for you is the fatal flaw of it, or the gaping pit that lax devs seem to fall into.

If you have a good game structure then procedural generation can work a treat (like in DF), though if there's nothing to do (like in NMS) the player feels cheated and the universe, for all its baubles, is desolate and without purpose.

It has its place but it should be in the background not up front and center, procedural generation is best caught by the periphery of our senses.

im not sure i would say there is a golden ratio, it all depends on how its implemented and what the game is about. just like with hand made stuff you can be good or bad at making a 'random' generator.

Yes and you'd have to be a consolefag or underage to think otherwise.

It's quite a different application of procedural generation. Games like Just Cause or FUEL use it too - basically the map uses a fixed seed of sorts and everything in the game is built around it. It doesn't have the polish of a completely handcrafted map but saves space.
I wonder if you could potentially hack these games and completely scramble the map layout.

...

Speaking of which

Actual Roguelikes (not procedural death labyrinths) where you're expected to make the best of the crap hand you've been dealt.

It's not procedural generation that's necessarily the problem, it's the repetition and the endless sense of having been to almost the same place before.

See any of the 3d fallout/elder scrolls games. It's all done by hand as far as I can tell, but you see the same rooms and props again and again until it's almost disorienting.

I've always felt that procedural generation works best when applied to a hand generated map. Say you have a road network and want to place bushes randomly by the sides of the road everywhere according to temperature and wetness. Well, no need to place 30000 bushes by hand when an algorithm can do it for you.

Diablo1
you gotta a bunch of prefab rules before going into generation for it to be decent

As a general rule, procedural generation only "works" as well as hand-made content when you spend AT LEAST as much time creating the procedural content as you could have hand-crafting the amount of content a player would encounter over the course of a non-sppedy playthrough. Which is to say since devs are currently using it as a time-saving measure it's usually producing shit. If anything it would take far more time to produce a procedurally generated game that's good for tons of replays than a hand made one.

Basically it only works right now in autistic passion projects and a few indie games.

Toontown did procedural generation pretty well. Two of the cog facilities were done via that method, and both came out well, from a technical perspective. The Lawbot Minesweeper puzzle room was bullshit, though. I feel like part of the success was the fact that both facilities had relatively few items that needed to work with. Pretty sure the Lawbot facility involved like 10-15 types of rooms, whereas the Bossbot one may have featured even less.