Monthly Archives: July 2007

Making and Breaking Thematic Consistency

‘Thematic consistency’ describes a sense of unity in a piece of art. It is very important in level design.

The worst and most amateurish type of thematic consistency is simple: no thematic consistency. These levels draw visual vocabulary from many different sources and mix them haphazardly. This is common among beginner level designers. These maps are easy to pick out. In order to avoid the no-consistency problem, level designers generally go through a ‘hyper-consistent’ phase where they run to the other end of the spectrum. Instead of everything being chosen randomly, everything looks the same. Take this example: the classic Counter-Strike map de_dust:


Dust’s visual vocabulary is limited to a very small number of design elements. Everything comes from the same narrow band of tan and yellow colors. To make a hyper-consistent level, you simply make sure that every part conforms to your single narrow design theme.

This type of design will not provoke a negative emotional reaction, it will not provoke a positive one either. It lacks the ‘pop’ of really great art. Also note that that it cannot look real, because the real world very rarely conforms to a totally consistent visual theme. There are very few places on Earth, save Antarctica, where everything fits together visually with this level of consistency. Hell, even on Antarctica you will find meteors from time to time.

Obviously hyper-consistency sucks. The solution is to come full circle and start breaking consistency again. There are several ways to do this.

The first way is ‘multi-theming’. This essentially means taking multiple artistic themes, each fully internally consistent, and applying them to different parts of a level. Note that this does not imply mixing the themes at any point. It means that different parts of the level conform to different themes. These different parts can be entire areas, or specific ‘pieces’. So, to make a multi-themed level, you just make sure that every part conforms to one of your narrow design themes.

For example, I was consciously multi-theming when I created DM-Lightfalls and DOM-Aphrodite. In DM-Lightfalls below, I used two themes: the airy verdant jungle, and the blue alien metal. Here they are quite cleanly separated, with a structure of blue alien metal surrounded by a verdant jungle backdrop. There are no parts of the level that do not firmly fit within one or the other theme.


In DOM-Aphrodite, I got more creative. There are three themes: Greek classical stonework, wooden and iron scaffolding, and the same verdant jungle look from DM-Lightfalls. You should be able to see that the elements are mixed more organically than in Lightfalls. Note, however, that there is no single piece that conforms to more than one specific theme. The trees are pure jungle, the statues are pure Greek, and the walkways are pure wood and iron. You don’t see any wooden walkways, for example, with white stone parts.


Multi-theming this way can get you very far, especially with non-realistic themes. Allow me to p1mp myself here and note that I won a very expensive copy of 3dsmax with Lightfalls, and Aphrodite was a finalist in the Make Something Unreal contest. This stuff works.

If you want to do really tight non-realistic maps, however, or just decent realistic ones, you need something else.

What you need goes beyond multi-theming the level. It almost comes full circle, back into thematic inconsistency territory. But instead of being haphazard, this type of inconsistency is interesting while still maintaining an underlying sense of unity.

Organically multi-theming a level requires more than pure visual artistic vision. It requires a narrative context for the level. This means that you need to have thought out the social and physical conditions that created the environment in the first place. If all of the artwork is consistent with this specific narrative, it can be as visually inconsistent as you like on a surface level, but it will still come together as a unified whole.

If you want an example of this, go outside in almost any city. The visual appearance of the landscape is composed of countless different types of visual elements. There will be natural parts in the trees, bushes, and grass. There will be loud, colourful advertisements. There will be people sporting a hundred fashions, and cars in a hundred styles. Even the buildings will have been built over hundreds of years and conform to radically different architectural schools. Furthermore, these elements will be mixed in various frighteningly complex ways: Men and women of several subcultures lounge on stone and wood benches in a grassy park. Beside them is a 100-year-old bronze monument, half-built-over by a modern art nouveau masterpiece.

This stuff doesn’t look good in abstract. It uses every color and every shape. On the surface, there seems to be no consistency. But it all fits into a totally coherent narrative relating to how that city was built.

I present here my finest example of this type of organic multi-theming. You’ll notice many more colors, themes, and elements than Lightfalls or Aphrodite. But it fits together because it has a rational story behind it.


To execute an organically themed level like this, you must make sure every visual element can exist as a natural outgrowth of the narrative context of the level.

We had a big discussion about this article on this BeyondUnreal Forum Thread.

Map: Europe Sniper

I’m just finishing up my latest baby: Europe Sniper.

I made this level on contract for Close Quarters Combat by Groove Games. Almost all of the assets are there, though there were really very few assets appropriate for a European theme in CQC. I was just getting so damn sick of industrial concrete jungles and sunset or overcast levels that I decided to do someplace where I might actually want to be in real life. Minus the constant gunfire, of course.

The first two shots are panoramas.




Predicting Fun: Why Mental Simulation Sucks

In order to design games, we need to be able to evaluate whether they will be fun or not. Given a description of a game, it is good to be able to know whether it will work before we create it.

This post addresses a common naive method that is often used for evaulating unproduced game designs. It’s something I call mental simulation, and it’s one of the most basic mistakes in game design.

Mental simulation is the process of imagining yourself playing the game in your head, then evaluating the game based on how that imagined play makes you feel. For example, in evaluating a first-person shooter, one would imagine an intense gun battle in which the player is victorious in some particularly hair-raising way.

Game previews and advertisements are designed to make you do mental simulation. They often describe particular gameplay scenarios in poetic detail. The goal is to make you imagine yourself playing the game and enjoying that experience. This is misleading, because the quality of a possible micro-experience in the game says very little about the quality of the game design as a whole.

The main problem is that mental simulation only allows evaluation of a very short snapshot of gameplay. A person will tend to evaluate the coolest possible moment of gameplay imaginable. This means that the rest of the game is being completely ignored. In almost all cases, the imagined experience cannot be extrapolated to the hours and hours of other gameplay. A game needs to be consistently fun across its entire playtime in order to be effective. Thus mental simulation will make games which are only fun for 5% of their length seem really good, even though the other 95% is very boring.

Mental simulation also fails to take into account learning curve. Since the game is in your head, you understand it fully. This is automatic. Unfortunately, there are countless possible game designs which are incredibly good after the player knows how to play the game well. Evaluating games by mental simulation can obscure the difficulty in learning to play the game.

Mental simulation also tends to produce design ideas that lack rigor and internal consistency. Going from mental simulation to code will almost always reveal gaping holes in game logic which cannot be elegantly reconciled.

There are two cognitive biases that lead people to do mental simulation. The first is that human beings respond very well to narratives. Games, however, are not narratives. They are systems from which a narrative can emerge. In evaluating only one possible narrative, we completely miss the quality of the system producing that narrative.

The second bias is the confirmation bias. This is a problem with the way people test hypotheses. Humans will tend to look for evidence which confirms their hypothesis, when in fact, searching for falsification is much more helpful. In this case, a person using mental simulation is looking for “evidence” for confirm their belief that the game in question will be good. They will almost invariably come up with some idealized scenario, and then extrapolate the resulting emotions across the whole game design. This is not useful.

Instead, try to falsify the game design. Don’t try to think of the coolest scenario possible. Try to think of the most boring scenario possible. Usually you will be able to come up with many examples, because most game designs cannot stand up to this sort of attack. If your design can, you know you have a real gem.

It is almost impossible to avoid doing mental simulation. Just understand that will mislead you if you do it naively.