Large Sandbox Design

Iuliu-Cosmin Oniscu
4 min readJan 8, 2023
The visual form of Los Angeles as seen in the field (Kevin Lynch — Image of the city)

A simple formula

This is going to be a more solution-oriented post. This page aims to define better a simple way of organizing content in a large-scale map. It leans heavily into my previous posts (primarily subdivision of space and Kevin Lynch’s Image of the City).

Large Sandboxes tend to suffer from content organization issues.

This happens because without a well-established context associated with that volume of content available, scattering them randomly often leads to uninteresting results. This also leads to confusion among the players because, without carefully balancing these tropes, the players will get lost.

Getting lost equates to difficulty in how fast the participant can construct a mental model of the space it inhabits.

There is a tendency to try to design space from the bottom up, starting with the first location the player begins in, be it an introductory onboarding area, and then slowly expanding the reach into larger open spaces.

This seldom works and is hardly how the real world is designed or planned.

The critical world is planning.

Hardly any inhabitable spaces start being conceptualized from the basic room outwards.

Instead, the better approach, and if you have been following my blog, you will know this, is to define space from macro to micro.

Defining Borders

If no other elements have been defined, the best way to start is to define borders and entry points.

Borders equate to barriers that are visible to the player. They block progress but act as paths allowing the player population to explore alongside them.

The space in the middle (without any distractions) will either be avoided or engaged with a higher degree of preparedness.

Entry Points

Entry Points are essential at this stage since they will imprint entry vectors that players entering the area will use.

The shape of a space imprint direction by default. Long spaces with no obstructions in the middle tend to have directionality on their long sides.

We usually do abstraction and assume that directionality comes on the long horizontal and vertical ledges in the constraints of the borders.

Travel Vectors

Suppose the entry vectors are not aligned with the axes defined by the shape of the zone we are working on. In that case, we could adjust the vectors to match them or assume that, eventually, these vectors will converge, allowing for some form of organic connectivity (again, in the absence of any other obstacles).

Attractor Landmarks

This is all nice, but the outcome is that all these entry points and intersections might look the same regardless of the chosen entry vector. The player might need to find out where they are in 3d space.

Adding landmarks to key locations will create some variation and make the entry vector easier to identify in the long term.


Concentrating POI (Points of interest) around vector nodes and entry points leads to clusters of interest that will allow the player to identify where they are faster.

Proximity to each other makes the groups look like individual formations with identities.

Districts and Negative Space

The space in between these paths and clusters of interest can be defined at this point as filler content since it’s out of the intended player path (represented by travel vectors)

This is where you want to play with biome variations. You will want to ensure that the travel vectors are the biome borders. If you can have at least two biomes, this will facilitate the district separation better.

Rare POIs and Corners

Since districts in this context serve as separators of content and you will put most of your generic content in intersections and around entry points, depending on how deterrent these are (as in dangerous and unexpected), you consider populating these areas with rare POI or Resources, given they are far away enough from everything else.

The same goes for corners.

Just ensure this does not become a rule that the players can expect from the area. You still want that area moderately deterrent (since gameplay will be concentrated in the intersection and nodes).



Iuliu-Cosmin Oniscu

As a level designer I am not a creator, I am a facilitator. Senior Open World Designer. #Leveldesign #Open World #Design @notimetoulose.