What is an (Good) Open World Really ?
An open world genre study, written from the designer's chair
I have spent most of my working and student life building spaces that people walk through, and a good chunk of it watching them refuse to. What follows is not a review round-up. It is what the genre looks like from behind the glass in a playtest room, and from the meeting where someone asks how many square kilometres we are shipping.
I want to start with the number, because the number is where the trouble begins.
---
The square kilometre is a lie we tell procurement
Every open world project I have worked on or consulted for began with a size discussion, and every one of those discussions happened before anyone knew what the game was about. It happens early because size is the one thing on the pitch deck that a non-designer can evaluate. It is legible. It compares. It sits next to a competitor's number in a grid and either wins or loses.
So the map gets drawn, and then the map has to be filled, and now you have a content budget that was set by a marketing comparison rather than by anything you learned about your own game. Everyhing that follows inherits that mistake. The pacing inherits it. The traversal system inherits it. The side quest count inherits it, which is why so many games ship with exactly the number of side quests that fit, rather than the number that were worth writing.
The industry knows this. Far Cry 3 is the reason. Not because it was cynical, it wasn't, but because it demonstrated that a map plus a marker system plus a repeatable activity set produces a game you can plan, staff, and ship on a date. That is an enormous achievement in a business where most things do not ship on a date. Everyone who calls it "the Ubisoft formula" as an insult should sit through one production meeting before deciding it was a moral failure. It was a scheduling solution that got promoted into a design philosophy, which is a different and more interesting problem.
Then The Witcher 3 landed and every greenlight conversation for six years ended with "like that, but ours." We inherited the checklist and the ambition together and mostly delivered the checklist.
Genshin Impact to state a more recent and ongoing problem of this type. Far the more sucessfull then others games on this article, suffered massvely with retention by the massive cheacklist and markers.
---
What playtests actually show you
Here is the thing I did not believe until I watched it maybe 20 or so times.
Put a player at the top of a hill. Give them a horizon with nine things on it. Do not tell them anything. Time how long they stand still.
With two or three points of interest visible, they move in under five seconds and they move deliberately. They pick. They point at the screen and say "that one." With nine, they stand there. Then they open the map. Then, most of the time, they follow the golden objective marker, because it is the only choice the game has told them is correct.
A former Rockstar designer gave this a name : exploration anxiety. Stappled on my wall since I learned about it. Too many landmarks produce analysis paralysis, and the player resolves the paralysis by beelining. The team's instinct is always to read beelining as disinterest and add more content to fix it. That makes it worse. You are treating a legibility problem with volume.
The fix is composition, not subtraction for its own sake. From any given vantage, you want a small number of clearly differentiated invitations, each one legible in silhouette and each one implying a different kind of promise. A ruin says secrets. A smoke column says people. A hostile silhouette says risk. If three things on your horizon all read as "generic structure," you have one landmark rendered three times, and you have taxed the player's attention for no return.
I have never regretted removing a point of interest. I have often regretted adding one. Zelda Breath of the wilds is the Prime Example of this done right. Likes of Journey, Hallow knight and Eldern Ring Deserve a mention of this done right.
---
The traversal tax
Every metre between two pieces of content is a tax the player pays for your ambition. Whether they resent it depends entirely on whether the crossing is itself an activity.
We used to measure this crudely: pick fifty random pairs of adjacent content beats, time the walk, and then, separately, ask testers afterward to describe what happened in between. If they can't tell you anything, that time is dead. It doesn't matter that the foliage is beautiful. Nobody has ever finished a session and said the best part was the grass.
This is why fast travel is such a diagnostic. Fast travel is not a convenience feature. It is a confession. The moment you add it from the start, you are telling the team, and the player, that the space between places has no value. Sometimes that is the right call and you should make it early and design around it honestly. But watch what happened to FromSoftware's own geometry when it arrived. Dark Souls without early fast travel had to make the world fold back on itself, and so it did, vertically, with Firelink at the root and every path spiralling out and eventually wrapping around to touch something you already knew. You don't find shortcuts in that game. You find out that two places were secretly adjacent, which is a feeling almost nothing else in the medium produces.
Later games in that lineage are stretched forward and backward instead of up and down, and the connective revelation mostly goes away. Not because the designers got worse. Because once you can teleport, the topology stops paying for itself.
---
Density is a craft problem, not a budget problem
The comparison I use in every internal argument is Prague against eighteen quintillion planets.
Deus Ex: Mankind Divided's Prague has four train stations, a few hundred buildings, and a couple of districts. It is too small to call an open world with a straight face. It is also more convincing than nearly anything with a hundred times its footprint, because every apartment has an authored way in and every vent goes somewhere a person decided it should go. You can feel the hand.
No Man's Sky at launch had every planet and one planet. Same structures, miles of nothing between them, caves with lovely geometry and nothing inside. That is the genre's real question stated at maximum volume: there is only so much a team can make, so how far apart do you space it and how often do you repeat it. One of those games answered "as little as possible" to both.
When someone tells me the small dense world isn't feasible on our budget, my answer is that it is the cheaper option and always has been. What isn't feasible is the big map at that density. The choice was made when you drew the boundary.
---
Grounded worlds have a higher maintenance cost, and nobody budgets for it
The more fantastical the setting, the more you get away with. In the Lands Between nobody blinks at a fire-breathing anything, because the world announced up front that it writes its own rules. It is also easier to populate, because you get to invent what belongs there and everything you invent is automatically canon.
Realistic settings do not give you that. Once you assert a real place with real physics, immersion becomes an ongoing maintenance obligation, and interactivity is the only currency that pays it. This is why Red Dead Redemption 2 keeps turning up in arguments it was not invited to. Track a wounded animal by its trail. Skin it. Start a fight. Antagonise a stranger. Play poker. Rob a train. Lasso a man and drag him around because you can. Sit still and listen to your own gang talk. That is a staggering amount of unglamorous, unshippable-looking work, and it is the entire reason the world holds.
The grounded games that skip it feel like a photograph you are allowed to shoot at. I have shipped one of those. The postmortem line was "the world felt like a backdrop," and it was correct, and I knew it eighteen months before launch, and there was no line item to fix it.
---
Structure is a design decision, and we keep making it by default
The most useful thing I ever did on a project was ask, in month three, whether the game needed to be open at all.
It usually doesn't. Open world is one option. There are open zones. Semi-open worlds. Open sandbox pockets inside an otherwise linear spine. Each gets you the exploration feeling without restructuring the game you actually wanted to build. Yakuza has been quietly perfect at this for twenty years by taking a few city blocks and stuffing them past capacity. Ghost of Tsushima would have lost nothing and gained a lot from compact zones. Horizon might have been a better semi-open game than it was an open one.
Look at how the linear games use open space as a spice. Resident Evil 4's remake has a boat. Ragnarök has an optional area built around wrecking Freya's wedding gifts, which is deranged and delightful and contains a dragon. The Last of Us Part II opens up just long enough for a guitar and a Take On Me cover. Each of those hits the exact note the open world genre promises, in a fraction of the footprint, with none of the traversal tax.
Metro is the honest case. The first two games were tunnels, tight and linear and openly indebted to Half-Life 2. Exodus opened up because the story left Moscow and the tunnels could no longer hold it. Structure followed fiction. That is the order it is supposed to go in, and it almost never does, because fiction arrives in month nine and the map was sized in month one.
---
The tools will not save us and may make it worse
I have spent enough time in Unreal's procedural content generation , Geomentry node based generation in blender, to have a view. It is graph based and rule driven and thoroughly authored: artists break areas down, tag elements, define how they combine, and then the system honours those rules. Move a rock and the terrain reworks itself, the log finds its resting place, the trees adjust. It is genuinely good and it is going to end up in nearly every project.
It also lowers the cost of surface area, and surface area was never our bottleneck. The bottleneck is authored intent. No graph is going to decide why the player should walk over there. What PCG will do is remove the last practical excuse for a small map, right when a small map is the thing most of these games need. I expect the next five years to produce the most beautiful empty worlds ever made.
---
The three questions I actually use
I stole this framing from a talk about competitive multiplayer maps, of all places. The designer stopped asking how to build the greatest map ever made and started asking what needs the map serves. Competence, autonomy, relatedness. It survives translation to open worlds better than anything that came out of our own side of the discipline.
Autonomy means options that differ, not options that repeat. Gothic 2 let you into Khorinis six ways: work for a landowner and wear his clothes, take a pass off a man who will want a favour later, bribe the guard, find the wall you can jump, claim you are gathering plants for the alchemists. That is autonomy. Two hundred question marks is homework in a nice font.
Competence means the world teaches you to read it and then pays you for reading it correctly. A skyline in Elden Ring is a competence engine. So is a Dark Souls shortcut. So, honestly, is a well-placed Ubisoft outpost, and the formula's defenders are right that the loop is satisfying. The problem was never the outpost. It was the fiftieth one.
Relatedness means the world answers back. It means the gang has conversations you can sit down inside of.
And then the four blunt production questions, in the order I ask them:
1. Does this setting actually want to be explored, or do I just like how it looks in a wide shot?
2. Can we author, not generate, enough worth finding at the size on the whiteboard? Be honest about the writer count.
3. Is the crossing an activity or a tax?
4. If the answer to any of the above is shaky, what does this look like as zones?
Nobody has ever been fired for shipping a smaller, denser world. Plenty of teams have been quietly destroyed by a big one.
---
Conclution:
The genre is not exhausted. It is being sorted, and the sorting criterion is embarrassingly simple. The games people still talk about placed every landmark on purpose. The games people bounce off placed the boundary first and then spent four years trying to justify it.
We drew the map before we knew what the world was. That is the whole postmortem, on almost every one of them, including mine on Punk_Noir.
Comments