Starfield has been a wild success. Like, objectively: it was the best-selling game in September and has since become the 7th best-selling game for the year. And those stats are based on actual sale figures, unmuddied by Xbox Game Pass numbers. Which is astounding to think about.

[Fake Edit] A success… except in the Game of the Year department. Yikes. It’s at least nominated for Best RPG, along with (checks notes) Lies of P? No more sunlight between RPG-elements and RPG anymore, I guess. Doesn’t matter though, Baldur’s Gate 3 is going to continue drinking that milkshake.

Starfield having so many procedurally-generated planets though, is still a mistake. And its a mistake that Mass Effect: Andromeda took on the chin for all of gamedom a decade ago.

Remember Andromeda? The eagerly-anticipated BioWare follow-up to their cultural phenomenon trilogy? It ended up being a commercial flop, best remembered for terrible facial animations and effectively killing the golden goose. What happened? Procedurally-generated planets. Andromeda didn’t have them, but the (multiple) directors wanted them so badly that they wasted months and years fruitlessly chasing them until there was basically just 18 months left to pump out a game.

You can read the Jason Schreier retrospective (from 2017) for the rest of the story. And in total fairness, the majority of the production issues stemmed from EA forcing BioWare to use the Frostbite engine to create their game. But it is a fact that they spent a lot of time chasing the exploration “dream.”

Another of Lehiany’s ideas was that there should be hundreds of explorable planets. BioWare would use algorithms to procedurally generate each world in the game, allowing for near-infinite possibilities, No Man’s Sky style. (No Man’s Sky had not yet been announced—BioWare came up with this concept separately.)

[…] It was an ambitious idea that excited many people on the Mass Effect: Andromeda team. “The concept sounds awesome,” said a person who worked on the game. “No Man’s Sky with BioWare graphics and story, that sounds amazing.”

That’s how it begins. Granted, we wouldn’t see how No Man’s Sky shook out gameplay wise until 2016.

The irony though, is that BioWare started to see it themselves:

The Mass Effect: Andromeda team was also having trouble executing the ideas they’d found so exciting just a year ago. Combat was shaping up nicely, as were the prototypes BioWare had developed for the Nomad ground vehicle, which already felt way better to drive than Mass Effect 1’s crusty old Mako. But spaceflight and procedurally generated planets were causing some problems. “They were creating planets and they were able to drive around it, and the mechanics of it were there,” said a person who worked on the game. “I think what they were struggling with was that it was never fun. They were never able to do it in a way that’s compelling, where like, ‘OK, now imagine doing this a hundred more times or a thousand more times.’”

And there it is: “it was never fun.” It never is.

I have logged 137 hours in No Man’s Sky, so perhaps it is unfair of me to suggest procedural exploration is never fun. But I would argue that the compelling bits of games like NMS are not the exploration elements – it’s stuff like resource-gathering. Or in games like Starbound, it’s finding a good skybox for your base. No one is walking around these planets wondering what’s over the next hill in the same way one does in Skyrim or Fallout. We know what’s over there: nothing. Or rather, one of six Points of Interest seeded by an algorithm to always be within 2km walking distance of where you land.

Exploring a procedurally generated world is like reading a novel authored by ChatGPT. Yeah, there are words on a page in the correct order, but what’s the point?

Getting back to Starfield though, the arc of its design followed almost the reverse of Andromeda. In this sprawling interview with Bruce Nesmith (lead designer of Skyrim and Bethesda veteran), he talked about how the original scope was limited to the Settled Systems. But then Todd Howard basically pulled “100 star systems” out of thin air and they went with it. And I get it. If you are already committed to using procedural generation on 12 star systems, what’s another 88? A clear waste of time, obviously.

And that’s not just an idle thought. According to this article, as of the end of October, just over 3% of Xbox players have the “Boots on the Ground” achievement that you receive for landing on 100 planets. Just thinking about how many loading screens that would take exhausts me. Undoubtedly, that percentage will creep up over time, but at some point you have to ask yourself what’s the cost. Near-zero if you already have the procedural generation engine tuned, of course. But taking that design path itself excludes things like environmental storytelling and a richer, more tailored gaming experience.

Perhaps the biggest casualty is one more felt than seen: ludonarrative. I talked about this before with Starfield, but one of the main premises of the game is exploring the great unknown. Except everything is already known. To my knowledge, there is not a single planet on any system which doesn’t have Abandoned Mines or some other randomly-placed human settlement somewhere on it. So what are we “exploring” exactly? And why would anyone describe this game as “NASApunk” when it is populated with millions of pirates literally everywhere? Of course, pirates are there so you don’t get too bored exploring the boring the planets, which are only boring because they exist.

Like I said at the top, Starfield has been wildly successful in spite of its procedural nonsense. But I do sincerely hope that, at some point, these AAA studios known for storytelling and/or exploration stop trying to make procedural generation work and just stay in their goddamn lane. Who out here is going “I really liked Baldur’s Gate 3, so I hope Larian’s next game is a roguelike card-battler”? Whatever, I guess Todd Howard gets a pass to make his “dream game” after 25 years. But as we sleepwalk into the AI era, I think it behooves these designers to focus on the things that they are supposedly better at (for now).

We learn from our mistakes eventually, right? Right?