Case Study – Love
Love is the project of one man development team Quel Solaar. After a career in graphics programming Eskil Steenberg decided to design and build his own “not so massively multiplayer mmo”. Love is an ambitious project and an extensive undertaking (especially) for one developer, the usage of procedural and emergent systems is a key element to making the project both feasible in production and unique in gameplay. The following article was produced after an unstructured interview I carried out on the 26th June 2010.
Almost everything in Love is generative. The world, AI, player interaction and objectives are all driven by a growing number of complex rulesets that constantly drive emergent gameplay. One of the key features of this approach is the construction of procedurally generated worlds for each game server. These landscapes are procedurally generated and then slowly rebuilt over time via ongoing algorithmic remodelling. Radical shifts in tectonic type or altitude give the game a fantastic almost ‘prog rock’ aesthetic where new regions overwrite previous terrain with new outcrops or plateaus.
A hillside meadow
When asked if these emergent processes ever get ‘out of control’ Steenberg said that they only become ‘overly’ chaotic when he wants them to be, in order to explore possible unpredictable outcomes. “I am in control of what I want to control”. He is happy to reliquish a level of stability in order to allow the expressive capacity of the code to do something suprising. The resultant range of possible environmental configurations does mean that occasionally players can find themselves trapped in pits or facing unreachabe plateaus. Steenberg sees these incidents as inevitable consequences of any emergent system, and is reticent to restrict software processes to remove such events (and therefore limiting its expressive range). He suggests players should equally understand and embrace such situations as part of the living identity of ‘their’ own unique worlds. Even so, the game doesnt abandon players in these circumstances, but allows them to teleport to more geographically safe environments. This compromise avoids both the ‘problem-solving sanitisation’ of the world and removes players potential frustration. In fact, Steenberg sees such emergent instances as directly reflective of real world experience and is opposed to how he believes many gamers and designers expect a game world to be built by “intelligent design” where every item is hand placed and every goal is signposted.
Although the emergent systems in Love could potenitally forgeground the mathmatical techniques involved Steenberg hopes the players primarily experience these features as an immersive aspect that guides their collaborative development of the gamespace. Steenberg does suggest that players have to ‘get it’ in terms of their understanding of the game-world as a malleable entity (something unusual in most games). This ability to effect the structure of the game world is probably the most important element of Loves mechanics. Eskil doesn’t expect this engagement to be immediate but he does see the players as ‘part authors’ in the larger evolutionary ‘story’ of their game world. Players are given tools to ‘edit’ the world, allowing them to build settlements and alter the geography around them. Ai bots will pursue the same ambitions, building their own cities and stealing tools. Alongside these interactions the world itself is slowly evolving.
A section of the world grid map
Unlike the majority of generative techniques in game design Steenbergs proccesses are active during gameplay. The entire landmass of each Love world is under constant flux. Functions equivalent to tectonic development are applied to the environment throughout the life of the game world. The underlying heightmap has specific zones which dictate the overriding features of that region (Desert,Ice-Cap, Meadow, Cliff…). These zones are patches of land with a detailed local awareness of neighbouring territory and their own structural rules. Meadows for instance average out the height of their individual cells to create a gently rolling surface, rocky zones have more radical terrain, and ice-caps present sheets of surface punctuated by chunky outcrops.
The zones are always under gradual renewal. At timed intervals the nature of a zone could switch to a different type (Desert to Rock, Forest to Meadow). Once flagged for transition the cells of a zone are modified from the server over a specific time period. (Not too fast so as to avoid rapid ‘popping’). Steenberg estimates that the entire surface could be re-zoned and re-configured in approximately 4 weeks.
The terrain evolution processes in Love avoid the standard applications of erosion (hydraulic,thermal etc) as Steenberg feels that these functions often result in landscape structures that are essentially boring as both gameplay spaces and as aesthetic models. Loves world will never essentially ‘grind down’ but is rather in a state of constant measured genesis. It is entirely possible that players could explore the entire terrain without locating a beach, only for a beach to emerge from the re-zoning process a week later.
A nearby settlement
In producing these zone types Steenberg experimented with multiple functions and iterations in order to locate a wide enough range of variation and style. In mathematical terms, each design function would initially explore a large search-space of potential expressions which Steenberg then trimmed back to suit the appropriate mechanical and aesthetic needs. He likens this to a google search session, where a broad initial enquiry needs to be refined into more specific terms, but even this process suggests additional opportunities and interesting variations. Similarly to Brian Enos approach to generative music production, Steenberg tends to test and improve systems in an accelerated timeframe, before finally committing the processes to a live software model. Neither does he have interest in recording or freezing the development of any of Loves server worlds, a decision itself that is an understandable extension of the conception of Love as a ‘living’ system.
Steenberg cites the iterative expressive range of generative functions as vital in determining the potential development of his terrain systems. In identifying possible locations for bridge structures the mathematical search function revealed more opportunities than a human eye would ever presume, even if many of them were aesthetically wrong. Steenberg speaks of these discoveries as invaluble insights in the process of finding the expressive possibilities contained in the underlying data models. He is insistant that when working with generative processes you should never try to assume the results. In his opinion many of the problems with generative design is the overzealous guidance of top-down design where the programmer already has a strict outcome in mind. Perhaps the willingness to let digital processes evolve interesting form independent from pre-conceived expectations is what gives Love is unique ‘fantasy’ style landscape.
The world as a sphere (3rd party viewer)
When discussing inspirations or precedents Steenberg refers more to mathematical innovators such as Ken Perlin than actual games. His reflections on procedural technology in historical or current games is almost dismissive. He outlines three forms of generative game technology.
A- Procedural methods as compression solutions:
8 bit games frequently used forms of procedural generation to build complex data within tight memory and storage constraints. The major example is Elite, a space trading game with hundreds of mahtematically generated starsystems.
B- Procedural methods as timesaving solutions:
Contemporary graphics engines support tremendous detail in the context of 3d modelling. However much of these elements are arduous and timeconsuming to produce by hand. Most modern engines generate details such as grass, dirt, foliage and atmospherics using procedural maths.
Both of these implementations are essentially static, pre-production techniques. Even when used in real-time (grass placement etc) they have no impact on gameplay or aesthetics.
C- Realtime generative systems:
Procedural systems can provide an active framework for the exploration of gameplay and design. This is Steenbergs intention with Love and an approach he sees rarely in any other game projects.
Steenberg admires the detail of simulation in projects such as Dwarf Fortress, but believes that the decision to focus on simplistic graphics and interface is both an advantage (in terms of preserving time to focus on gameplay simulation) and a restriction (in terms of inviting player engagement and exploring aesthtic form)
Ultimately though, like most game designers he considers the players experience to form the guiding principle of how far emergent behaviour should be allowed to express ‘itself’. Although digital art contexts frequently prioritise the ability of a procedural system to generate a wide expressive range for aesthetic purposes, commercial game design necessarily focuses on the usability of the system from a players point of view. Like many other designers he sees this as the flaw with Will Wrights ‘Spore’, a game with impressive generative possibilities undermined by lacklustre gameplay mechanics. Steenberg also beleives that game mechanics enable the designer to ‘gate’ progress and exploration in the game and by doing so make the experience more potent, both rarity and restricted access adding to the value and mystique of content (as in a economic model)
In the same manner Steenberg expresses little interest in the artisic context that could frame his work. He doesn’t see his projects as having philosophical intent and like many talented programmers is inspired more by the potentials of code than the nature of its cultural reflections. Even financial reward seems unlikely to persuade Steenberg to explore more abstract, less game-like versions of his vision. Perhaps this is a reflection of how gameplay itself represents a creative and emergent space of possilibity that mirrors the exploration and multiplicity of generative systems.
Eskil recently posted a short article where he discussed the difficulty of managing the emergent problems of complex procedural systems. In my opinion Love exists more often to satisfy the curiosity and experimental aesthetic of the developer rather than the requirements of a stable gameplay system. In any form of user friendly gameworld/simulation the purity of the unbounded emergent world will have to be tempered for the sake of consistent space to play in.