How to make an exciting Open World: the POIs Diversity Rule
Learn my simple rule for making an interesting, engaging open world level, and get the full context with a detailed look at the different types of game levels, including linear, non-linear, and open worlds.
Hello everyone, my name is Vasiliy Skobelev (aka Baz), and I’m a Lead Level Designer at MY.GAMES (Pixonic). I’ve been working in game development for 9 years, and for most of that time, I’ve devoted my work to Level Design. Today, I want to share my thoughts on how to design an exciting open world. This article won’t be a strict set of rules or a step-by-step manual, but rather, it will consist of my observations on this matter, coupled with a bit of personal experience. I hope it will be helpful for your game, as well.
Examining video game level types
In order to properly delve into open world solutions, I’d like to first take a step back and talk a little bit about the various types of video game levels in general. For the sake of simplicity, let’s divide them into three major categories: linear, non-linear, and matrix (spoiler, this is the one we’ll be aiming for).
With linear levels, players will follow the critical player path or the golden path, and this basically means all the available navigable space is mandatory to progress through the game; these levels have clearly defined start and end points. Although we refer to these levels as “linear”, this doesn’t necessarily mean they can’t include optional content: we can implement dead-ends and looped spaces, but we have to keep in mind that these can’t be an equal alternative to the main level path. In 2022, we saw a perfect example of such a design: The Callisto Protocol.
That game keeps level gameplay streamlined from the very beginning to the very end while also introducing forks (usually with two options) which may lead to a dead-end or a loop returning the player to the main playable area. We can also find similar design solutions in modern installments of Doom installments or 2013’s Tomb Raider.
Non-linear level gameplay differs significantly from its linear ancestor. And it’s also much more expensive to produce. The reason for that is simple: while still maintaining the same play session time to complete a level, you double, or even triple, the available options within this time. An intended player experience is not set in stone and can vary depending on the chosen playstyle. These levels usually introduce multiple (seemingly equivalent) critical paths which most of the time end at the same point. A good example of this mechanism can be seen in immersive sim games that emphasize player choice, like Deus Ex, System Shock, Dishonored, Thief or Prey. (When you face a door which can be broken with Strength attribute upgraded up to level 8 you can always find a vent nearby or lock that will demand a certain level of Hack/Lockpick skill.)
It’s important to keep in mind that even non-linear games don’t necessarily feature non-linear level design. For example, The Bureau: XCOM Declassified has multiple endings, all of which introduce either different protagonists for the final mission, or which feature completely different levels. The trick is that all of the final levels are still linear. The same goes for the Call of Duty: Black Ops Cold War: there is a narrative choice at the end which determines the player’s final level; both of those possible levels are linear in nature, but this does not contradict the non-linear structure of narrative.
A matrix-like (or simply matrix) level structure underlies open world maps. In the world of mathematics, a matrix is a row of rows. What about the level design of such spaces? Well, there is indeed a row of rows, but usually every cell is a square part of the landscape. Nowadays game engines let you design the world in less strict forms but still, the underlying logic is about grid-like distribution of objects.
Open worlds usually consist of outdoor environments with the occasional interior space. The critical path is no longer a predefined trajectory, but a list of coordinates a player has to visit in order to progress through a game. We can no longer direct (i.e. predict) the majority of their experience. Instead, the level designer has to decide how to distribute points of interest (POIs) on the map. There are some basic rules like don’t repeat the same gameplay twice in a row (we don’t need to follow a combat situation directly with another combat situation) or design multiple entry points for each POI but there are also not so obvious solutions. We will return to them in the next section.
For now, let’s decide at what point a level can be called “open world”. In terms of games, we can think of open world titles like The Witcher 3, The Elder Scrolls V: Skyrim and Red Dead Redemption. There is no argument here: all of these present vast spaces with lots of things to explore and an adventure where players can choose how their individual experience unfolds. But there are also “semi-open worlds”; a good example of this would be the recent Dead Island 2.
In that game, you can’t traverse from Santa-Monica beach to Monarch Studios in a straight line on foot: you have to use fast travel, at some point encouraging a few loading screens in the process. But most of these in-game regions present a space of roughly one square kilometer with no predefined path but these levels still have a matrix structure under the hood. That automatically makes them open worlds but little ones. The game consisting of such maps can be called a semi-open world game. And that also suggests a specific approach to skill checks. But what is a skill check? Let’s take a closer look.
Skill Checks and the “POIs Diversity Rule”
We need to define a few terms so we can all be on the same page, even for those readers who have never developed games.
Video games have what is known as a core loop; it’s an in-game cycle of skill checks the player goes through while playing the game. Meanwhile, a skill check is an in-game challenge that presents a task to let the player show how well they can use a mechanic they were taught earlier in the game.
So, one could describe a core loop as a cycle of in-game core mechanics, or groups of mechanics. But, in my opinion, it’s more correct to refer to this as an order of skill checks, since it’s not so important which mechanic is used as much as how the player uses it.
For example, let’s take a look at Tomb Raider (2013). It was the first installment in the reboot trilogy and it introduced one ingenious solution that brought much higher sales than previous titles — puzzles became simple within the critical path while all the hardcore challenges were moved to optional areas. So, why did they do this, and why is it relevant at all?
If you can recall the core loop of the old Tomb Raider games it consisted, basically, of three groups of skill checks: traversal, battle, and puzzle. This was fun for a time but stopped being fun at some point. So the team concentrated their efforts on battles and acrobatics while all the difficult brain twisters became side content, and so story progress is not bound by time a player could be stuck doing side content.
This way, players could now clearly anticipate how the game would challenge them. If you stuck to the main story, you would be challenged with different battle encounters and overcoming obstacles between them. If you explored the world outside the golden path, you could encounter everything the game could offer: battles, acrobatics, and puzzles.
The important thing here is to consider the right order to introduce different types of gameplay. If you juggle only fights and movement, this doesn’t mean you can have three fights in a row or a 40-minute long obstacle course. We must diversify the player experience as often as possible, while at the same time, ensuring it remains immersive and fluid. To put it simply, you can’t place 3 boxes and ask a player to make three contextual “climb” animations one after another — that’d be boring as hell. Instead we can place a box, then a curved slope, and maybe a small gap to jump over.
I think the idea is clear enough — in a linear experience, we have to keep in mind the order a player will face skill checks. But how can we do this in an open world where the player themself defines the order of events, i.e. skill checks? I’ve thought about this for years, and I have tried to define a simple rule based on successful games of this genre. In the end, I came up with something I previously called The Triangle Rule. But, since there is another rule with a similar name from the The Legend of Zelda: Breath of the Wild developers, I’ll call mine the POIs Diversity Rule to avoid confusion. Take a look at this picture:
POIs from Breath of the Wild lead to different experiences: the tower on the left presents a climbing challenge with stamina management, the foggy forest in the middle is a sort of a navigational puzzle leading to a shrine, and the volcano on the right introduces survival elements related to its hazardous environment. (By now, you might have already figured out what this rule is all about.)
POIs Diversity Rule states: in order to keep the high quality of the core loop in an open in-game environment, a level designer should populate the horizon line with at least three POIs each offering different gameplay, no matter where the player goes.
Now, let’s break that concept down even more: we can’t control the order of locations the player will visit, right? Right. But we can control the diversity in terms of the distribution for these sub-levels. Of course, we can’t actually guarantee diversity for all possible coordinates in the open world, but most of the time, we can keep at least 2-3 POIs visible to the player by using landmarks and composition in general. After we’ve done that, we should also verify that those locations actually offer different experiences. That’s it! It’s not exactly rocket science, but it’s not an easy feat to achieve by any means, either.
Sometimes you can cheat though. You can actually place similar POIs close to each other if you have a clear divider between them which will guarantee the player won’t experience them one right after the other. A good example comes, somewhat unexpectedly, from Resident Evil Village, and the village itself to be precise. While it’s not a typical open world, it has the same underlying distribution logic under the hood. But this time, the logic applies to game objects placed in the village. Once the player has all mechanics unlocked, they can go through all the spaces in this level. The catch is, as I mentioned above, dividers: walls, buildings, fences, etc. There is not a single object type consecutively repeated on the map except those with places divided by some static object.
An Open World Level Designer should also remember some fundamental metrics specific to outdoor gameplay, which aren’t that obvious. These include:
- The average time to introduce a new event or point of interest. For the aforementioned Breath of the Wild title, this is roughly 40 seconds. The game would play differently if most of the time players went 2 whole minutes before noticing something of interest.
- The effective distance of attack, which is crucial for proper implementation of battle encounters.
- The maximum line of sight. You can’t just make the whole map visible from one corner to the other. This wouldn’t work performance-wise and it would also take away the intrigue of exploration;I can’t stress enough how important that is. You can see huge space dividers like mountains in all open world games from TES V: Skyrim to Just Cause.
- Elevation limit. It’s crucial to understand how aggressive the verticality can be in your game while keeping the gameplay comfortable. This must be considered alongside the fact that most modern engines still don’t support vertical level streaming. That’s why games like Days Gone don’t have it at all: zombies in the cave below still react to your sounds and actions even when they don’t see you, because from a top-down perspective, Deacon is actually close to them.
- And there are many, many more.
All the things I’ve mentioned are easy to imagine in terms of outdoor design, especially when we’re talking about natural environments like forests, fields, mountains, and so on. But what about cities? While there are many similarities, this question demands a separate chapter.
Urban spaces
While the general level design rules for outdoor spaces are no doubt useful, they are much harder to implement in spaces with lots of strict limitations, like towns or cities. You don’t just have to think differently, you need to do this from the very beginning.
While the POIs Diversity Rule is something I came up with myself and, that I’m, well, proud of, in this chapter, I’d like to refer to the work and analysis of my fellow level designers: Iuliu-Cosmin Oniscu (Watch Dogs series, Assassin’s Creed: Rogue) and Mikhail Kadikov (Might & Magic Heroes VII, The Climb, Hunt: Showdown). To be specific, I will reference Large Sandbox Design and the concept of Player Possibility Maps.
When you design a city, you have to start with macro design: squares, main streets, landmarks. These elements will ultimately define the navigation and the places that attract the player’s attention. Then, you will need to emphasize the main entry points to this urban space (and actually lead the player inside). Yes, this is the exact kind of environment in open worlds where all the popular navigational signifiers like leading lines, bread crumbs, framing, objects of relativity and landmarks have extremely high value. (Not as high as in linear levels and interiors in general but still significant.)
When you get to micro level design and start working on streets and props, it’s important to understand the concept of the Possibility Map. To put it simply: this is the number of available paths and their respective weights. Keep in mind that this applies primarily to the critical points of the player route: crossroads, door frames, gates, corners, points of no return, and other places where the player has to make a navigational choice.
For example, the cities in Titanfall almost never exceed 4 available paths in forks and crossroads. Meanwhile, starting with the tutorial, games like Dishonored lead the player via one available path, slowly adding forks of 2, then 3, to allow the player to become comfortable with local navigation.
Additionally, to control the amount of stress a player experiences, you have to think about the weight of these paths — this basically means the probability of the player choosing a specific path to follow. The main path should be clearly defined and emphasized with a broad entry point, light, leading lines, open doors, or something else. To contrast, optional content off to the side may be less attractive due to an obscure entrance, curved trajectory, or other design solutions.
Take a look at this particular scene in Dishonored 2. This is literally the first time a player faces a fork of 3 possible paths.
This scene was deliberately placed in a low-stress environment with no time pressure and options 1 and 3 are clearly less attractive to avoid unnecessary stress; this effectively communicates where players should go to progress the game and where they can go to explore side content.
The negative and positive effects of fast travel
In these closing chapters, I’d like to share some discoveries I’ve made along the way while working on open worlds myself and consulting numerous projects.
Firstl, never underestimate the impact of fast travel mechanics. These are much more important for your level design than it may seem at first glance. Fast travel points are places that players will visit the most, whether we intend it to happen or not. This mechanic is easy to use and shortens time spent on navigation.
In games like The Legend of Zelda: Tears of the Kingdom, Fallout 4, or Star Wars Jedi: Survivor, there is a less strict solution: you can travel to fast travel points from anywhere at any time. This eliminates unnecessary backtracking, and it also gives players the feeling of freedom and control, and it motivates them to fully explore the map without the need to keep in mind how far they’ve traveled from the last city or continue point. Most importantly, fast travel greatly increases the chance that the player will visit most locations on the map because they can travel at any moment as they wish.
Games like The Witcher 3 or Cyberpunk 2077 bring backtracking to the equation. This is good if we want players to face random encounters and different content which changes on the way back, but it will frustrate the player if the path back doesn’t provide a new experience.
In terms of level design (or even in terms of analytics and heat maps), this makes fast travel points hot spots, and this means players will visit them the most. At the same time, the places directly between the nearest two fast travel points won’t see as much player traffic.
Additionally, in these games it’s not advantageous for the player to go particularly far from a fast travel point, especially on foot: the further you go, the longer the backtracking will be. This mechanism also makes life harder when you try to analyze metrics. Did the player visit this point of interest because they needed to get back to the fast travel point or because they wanted to? You’ll never know.
There is also a third approach. Games like Red Dead Redemption 2 and Horizon Forbidden West add another restriction — fast travel is not free. You have to spend resources to use this system. Level designers should remember that this mechanism will motivate the player to visit special places to gather these resources.
If you really need this then go for it, but I’d strongly recommend against design solutions like this since it goes against a true sense of freedom and brings a lot of repetitive gameplay when the player just wants to explore the world you’ve developed. Introducing places a player will need to visit (but doesn’t initially want to) is always a risk.
The Three Cs
At the D.I.C.E. Summit in 2002, Mark Cerny presented his “Method” of production, and it included a really crucial concept for Level Design — the Three Cs: Character, Camera and Control.
How is this concept connected to open world level design? It’s directly connected. I’ve seen many games in development where a playable character wasn’t finalized but the world map was already deep in development.
Or even worse: they were dealing with a vehicle, but at some point, the developers decided to let the character leave the vehicle and go on foot.
Or the opposite: the metrics were configured for a bipedal humanoid and then, all of a sudden, the character gets a car. After that, obviously, the world map stops working as intended, and it demands a lot of time and money to glue everything back together.
If you’re faced with the task of bringing a world map to life in any form, the first thing you should do is to finalize the character and all mechanics tied to it — including all the possible controls and camera movements. I can’t stress enough how much money and effort this will save down the road.
You can help the engineers and mechanic designers by assembling special levels with the range of possible gameplay setups, like gaps to jump over and ropes to climb. Only after you’ve set your mechanics in stone should you start to develop levels to let those mechanics truly shine.
Non-Euclidean Spaces and other tricks
Sometimes you’ll encounter significant obstacles while developing large maps. Most of them will be tied to performance issues and tech limitations. I’d like to highlight a couple of issues and how to avoid them.
First and foremost, you’ll have to somehow separate safe zones from hostile environments and stream a lot of NPCs in cities — which is not necessary when you play near the city but not inside of it. It’s a whole different story when you go inside though. Studios like Bethesda tend to load these spaces separately via loading screens and here’s why: almost all the cities in the 3D Fallout games are bigger on the inside than the outside. Yes, that’s right. This solves many problems and it’s not really noticeable during gameplay; you just prepare a static placeholder which looks like the city. Then, do the same for the outside panorama when you load an actual city.
Many engines don’t support vertical streaming of levels out of the box. The recent Tears of the Kingdom makes it look easy, but even the most popular software like Unreal Engine and Unity are still figuring out how to do this properly.
So, this raises the question: how do we load levels if we have elevators? The short answer is this: make the map wider, not deeper. The elevators won’t actually move, and instead, they’ll teleport the player to the right spot then open the doors. The second level will be at the same height in the overall world, just outside the general area. Incidentally, this is the reason why old games all have elevators without windows. Of course, now you can play games with elevators that actually move like Cyberpunk 2077 or Atomic Heart.
Thank you very much for reading this article to the end. I hope you found some useful takeaways for your future level design and game analysis. Of course, this is just the tip of the iceberg and I have so much more to share on the topic but let’s take one step at a time. Stay tuned for more to come!
