This entry in ‘AI and Games’ is not only discussing existing work by academics in the field of level design, but also a rare occasion in which I will actually highlight research of my own. This topic of my own research is an area which this series has seldom explored, given I wish for ‘AI and Games’ to stand on its own merits and celebrate the work of our larger research and development communities and not act as my own personal podium. This blog article is an adaptation of a research paper I have written on this topic, re-written for a non-academic audience. This work has been ratified for academic publication in its original form and takes on board the comments and criticisms from the peer-review process. While more speculative comments may be injected throughout, these are reflections of personal opinion and may not be expressed in the original publication, given they may lack any empirical basis.
Video games often suffer an interesting conundrum: a reality driven by the complexities, finances and timescales of development and how these issues clash with the needs of a creative medium. Game development is a risky business; one in which a misstep by any company can result in their subsequent closure. If a game does not resonate it with players or the market at large, it can subsequently fail to meet projected sales targets. With substantial amounts of money lost, a developer may be forced to close doors which results in many a talented developer looking for further employment. Many of the developers I’ve met over the years – be they indie or AAA – are very creative and innovative thinkers: with a keen eye on creating new and interesting experiences that attempt to buck the existing trends of the market. But in order to do that, you need not only to recognise the state of that market, but also how to build your ideas from it.
This is readily apparent when dealing with major intellectual properties in games. Where sequels are churned out with alarming precision, with entries in game series being released almost a year to the day of one another. But in order to do that, you need to give players a reason to care. It’s a fine line that all major IPs face: the need to continually craft product that not only respects that which preceded it, but also innovates on that base template. These innovations can be major or subtle, but they must be communicated to players sufficiently such that they acknowledge their merits and relevance, then dissect the more subtle nuances of these changes. Failure to achieve this, can lead to disinterest from the previously loyal fanbase which has lead to failure for many a previously successful IP. This can be achieved through a variety of means, be it franchise fatigue (Guitar Hero), a failure to respects the merits of that which preceded it (Sonic the Hedgehog) or changing a formula to the point the innovated product no longer reflects that which it sought to enhance (DOOM 3).
With this in mind, we look at the most successful video game franchise of all time, Super Mario: a series that not only has driven the sales of entire consoles lines, but is still being published after 30 years. Now the Super Mario series (and Nintendo in general) is often criticised for not innovating: with yet another Mario game being published. But how do these games continue to innovate? How do they ensure that Mario continues to be just as important to gaming now, as he was in 1985? In this piece we discuss how the 2D Super Mario games continue to innovate and one means identified in academic research that helps to to quantify this practice.
Super Mario Bros.
For those not familiar, we will bring you up to speed quickly! Super Mario Bros. is the first of a series of 2D platformer games released on the Nintendo Entertainment System (NES) in 1985, with subsequent releases across all of Nintendo’s hardware portfolio. The player typically takes control of the titular Mario and must navigate him through an environment filled with hazards and enemies, typically found by progressing right along the horizontal axis. Mario will complete each fixed segment of play, or level, by passing a final obstacle in the map, typically a flagpole. This obstacle, while non-negotiable, is easy to complete; with a bonus given based on how well the player navigates it. Levels in Super Mario games are grouped together into ‘worlds’: a collection of four or more levels that often share some visual aesthetic or core design principle. Each world culminates in a boss battle in a castle, whereby the player must face off against the antagonist Bowser or one of his minions.
Since the original games release, the Super Mario series has became the biggest property in video game history. Establishing not only Mario himself as a icon of popular culture (ironically in America, despite his Japanese origins and Italian ethnicity), but also many of the established tropes of Super Mario gameplay and level design.
Design patterns are a body of academic research that started in the mid-2000’s, with an emphasis on establishing how to express game concepts. A body of work by both Björk and Holopainen culminated in a textbook on the subject (Björk and Holopainen, 2005), which largely acts as a reference material for the many ways in which design patterns can be identified or conceptualised within games. The key part of this process is establishing what relationship exists within this recurring structure within the game and the implicit understanding that players will build from it by acting appropriately. While we are interested in how these patterns are exhibited in Super Mario Bros., this is of equal relevance to all aspects of gameplay, with perhaps the easiest to understand patterns coming from user interface design, given we have these common conceptions of how to navigate through them.
Design Patterns in Super Mario Bros.
The Super Mario Bros. series continues to be a popular area of research, given not only the interesting problem space it provides, but the continued popularity of the franchise. While there is a significant body of work exploring the series from a range of angles – with AI bots and PCG discussed in our Mario AI Competition article – design pattern analysis is a new area to explore. An interesting piece which can be found in (Dahlskög and J. Togelius, 2012), identified over 20 recurring patterns of gameplay when assessing all of the original Super Mario Bros. with the exception of castles and underwater levels. These can be broken down into five categories:
- Gaps: Gaps in the ground that present a hazard for Mario to cross.
- Enemies: Groupings of one or more enemies together such that it can cause some issues.
- Paths: Use of blocks to create multiple paths for Mario to traverse, either focussing on items or avoiding enemies.
- Stairs: Well… stairs, I guess.
- Valley: A grouping of environmental obstacles that place Mario within a small valley.
These patterns are identified to exist in particular ‘beats’ of play, which is singular chunk of gameplay, a term first really explored in (Smith, Cha and Whitehead, 2008). Note that these beats are not a reflection of the physical length of the level, but the number of individual segments of play. The patterns identified provide an interesting lexicon through which to express Super Mario level design: a language in essence. This allows us to encapsulate and observe interesting aspects of each level and conduct comparisons and evaluation.
The Legacy of Super Mario Bros.
Given this work by Dahlskög and Togelius, I was inspired to explore the design patterns of subsequent games. Does the original research from the 2012 paper stay true? Perhaps more importantly, can we encapsulate how Mario games innovate by looking at the design patterns that exist in subsequent games? This research, detailed in full in (Thompson, 2015), looks at 10 different games from the Super Mario Bros. series that retain the core princinples of Mario level design. This list of games is shown above, with three games ignored given they deviate too far from the original works (Super Mario World 2: Yoshi’s Island, Super Mario Land 3: Wario Land and Super Mario Bros. 2/Super Mario Bros. USA, instead relying on the original Japanese release).
This paper (and indeed this article) reflects solely on the first world of each game, while annotating the rest of each game is in progress there are some interesting findings from this first set of data.
- Beats Per Level Fluctuate: The total number of beats has continued to increase gradually over the last 30 years in the first world, until the New Super Mario Bros. series, which has continued to decrease that number (instead increasing the density of patterns per beat).
- Consistent World Design: Each world largely follows a pattern in the order of level types, with a minimum of one land, underground and highground level per world and typically in that order. While these types (underground, highground etc.) are largely aesthetic, they do have an impact upon the level design.
- New Patterns Aplenty!: Future games introduce a myriad of new patterns, with our research in the first world identifying roughly 25 more (check the paper for the full list). Interestingly, all of these patterns originate from one or more patterns that precede them.
- Pattern Popularity and Throwaways: Some patterns from the original Super Mario Bros. seldom return in existing games. Meanwhile games such as Super Mario World and Super Mario Land 2 are notorious for introducing patterns that seldom appear in subsequent games (if ever).
- The Silent Tutorial: Nearly all Super Mario Bros. titles exhibit this same grouping of patterns in the first beat of the first level of the first world. This acts as a tutorial for players to help acclimatise to the core mechanics of gameplay.
- Single-Enemy: The use of one enemy, typically a Goomba or KoopaTroopa.
- n-Path (n ≥ 2): Blocks or platforms used to establish multiple paths for the player to navigate across.
- Risk-Reward: A scenario where the player can opt to take a more treacherous path – one where the enemy is placed – with the potential to acquire a power-up item
- Pattern Crossover & Subversion: The New Super Mario Bros. games are notorious for combining otherwise incompatible patterns together, to create weird hybrids. However, what’s really interesting about these hybrids is that one pattern remains dominant, such that you can figure out how to traverse it without giving it too much thought. Arguably, the second pattern is there to add some flavour to the proceedings. Meanwhile, there are some instances where an existing pattern is deliberately subverted: it retains the original premise, but the physical manifestation of it changes such that it forces you to change tactics.
As intimated previously, there is a significant body of work in this area and plenty more still to be done! Firstly, there is the continued analysis of other games through design patterns, not to mention the remaining 7 worlds of Mario games to explore! In addition, there is a body of research that has began to explore how to use patterns as part of the process for procedural content generation techniques. This article does little justice to the significant amount of work in this area. If anything, we hope it merely highlights its existence to those unfamiliar with this area of research as it continues to expand!
This article and video are based on numerous academic publications in the area of design patterns, with some specifically exploring Super Mario Bros. We identify a number of key publications in this area that should be explored if you are interested in this work. Naturally if you are aware of further publications that merit being added to this list, please get in touch and I will amend accordingly!
- S. Björk, & J. Holopainen. Describing Games – An Interaction-Centric Structural Framework. In Copier, M. & Raessens, J. (Eds.) (2003) Level Up – CD-ROM Proceedings of Digital Games Research Conference 2003, Utrecht, The Netherlands, 4-6 November 2003.
- S. Björk, S. Lundgren, & J. Holopainen, J. Game Design Patterns. In Copier, M. & Raessens, J. (Eds.) (2003) Level Up – Proceedings of Digital Games Research Conference 2003, Utrecht, The Netherlands, 4-6 November 2003.
- S. Björk and J. Holopainen. Patterns in Game Design. Charles River Media, 2005.
- S. Dahlskög and J. Togelius. Patterns and procedural content generation: revisiting mario in world 1 level 1. In Proceedings of the First Workshop on Design Patterns in Games, page 1. ACM, 2012.
- S. Dahlskög and J. Togelius. Patterns as objectives for level generation. In Proceedings of the Second Workshop on Design Patterns in Games. ACM, 2013.
- S. Dahlskög and J. Togelius. Procedural content generation using patterns as objectives. In Applications of Evolutionary Computation, pages 325–336. Springer, 2014.
- G. Smith, M. Cha, J. Whitehead. A Framework for Analysis of 2D Platformer Levels. Proceedings of ACM SIGGRAPH Sandbox Symposium 2008, Los Angeles, CA, August 9-10, 2008.
- G. Smith, J. Whitehead, and M. Mateas. Tanagra: A mixed-initiative level design tool. In Proceedings of the Fifth International Conference on the Foundations of Digital Games, pages 209–
- G. Smith, J. Whitehead, and M. Mateas. Tanagra: Reactive planning and constraint solving for mixed-initiative level design. Computational Intelligence and AI in Games, IEEE Transactions on, 3(3):201–215, 2011.
- T. Thompson. The Fine Line Between Rehash and Sequel: Design Patterns of the Super Mario Series. 4th Workshop on Design Patterns in Games, Foundation of Digital Games Conference, June 2015.