Procedural content generation and design patterns could potentially be combined in several di erent ways in game design. This paper discusses how to combine the two, using automatic platform game level design as an example. The paper also present work towards a pattern-based level generator for Super Mario Bros, namely an analysis of the levels of the original Super Mario Bros game into 23 di erent patterns.

This paper discusses how to use design patterns in procedural level generation, with particular reference to the classic console game Super Mario Bros. In a previous paper, we analyzed the levels in this game to nd a set of recurring level design patterns, and discussed an implementation where levels were produced from concatenation of these patterns. In this paper, we instead propose using patterns as design objectives. An implementation of this based on evolutionary computation is presented. In this implementation, levels are represented as a set of vertical slices from the original game, and the tness function count the number of patterns found. Qualitative analysis of generated levels is performed in order to identify strengths and challenges of this method.

In this paper we present a search-based approach for procedural generation of game levels that represents levels as sequences of micro-patterns and searched for meso-patterns. The micro-patterns are "slices" of original human-designed levels from an existing game, whereas the meso-patters are abstractions of common design patterns seen in the same levels. This method generates levels that are similar in style to the levels from which the original patterns were extracted, while still allowing for considerable variation in the geometry of the generated levels. The evolutionary method for generating the levels was tested extensively to investigate the distribution of micro-patterns used and meso-patterns found.

Generating content at multiple levels of abstraction simultaneously is an open challenge in procedural content generation. Representing and automatically replicating the style of a human designer is another. This paper addresses both of these challenges through extending a previously devised methodology for pattern-based level generation. This method builds on an analysis of Super Mario Bros levels into three abstraction levels: micro-, meso- and macro-patterns. Micro-patterns are then used as building blocks in a search-based PCG approach that searches for macro-patterns, which are defined as combinations of meso-patterns. Results show that we can successfully generate levels that replicate the macro-patterns of selected input levels, and we argue that this constitutes an approach to automatically analysing and replicating style in level design.

Evaluation is an open problem in procedural content generation research. The eld is now in a state where there is a glut of content generators, each serving di erent purposes and using a variety of techniques. It is difficult to understand, quantitatively or qualitatively, what makes one generator di erent from another in terms of its output. To remedy this, we have conducted a large-scale comparative evaluation of level generators for the Mario AI Benchmark, a research-friendly clone of the classic platform game Super Mario Bros. In all, we compare the output of seven different level generators from the literature, based on different algorithmic methods, plus the levels from the original Super Mario Bros game. To compare them, we have de ned six expressivity metrics, of which two are novel contributions in this paper. These metrics are shown to provide interestingly di erent characterizations of the level generators. The results presented in this paper, and the accompanying source code, is meant to become a benchmark against which to test new level generators and expressivity metrics.

We shown that novel, linear game levels can be created using n- grams that have been trained on a corpus of existing levels. The method is fast and simple, and produces levels that are recognisably in the same style as those in the corpus that it has been trained on. We use Super Mario Bros. as an example domain, and use a selection of the levels from the original game as a training corpus. We treat Mario levels as a left-to-right sequence of vertical level slices, allowing us to perform level generation in a setting with some formal similarities to n-gram-based text generation and music generation. In empirical results, we investigate the effects of corpus size and n (sequence length). While the applicability of the method might seem limited to the relatively narrow domain of 2D games, we argue that many games in effect have linear levels and n-grams could be used to good effect, given that a suitable alphabet can be found.

This paper analyses dungeons, of the varieties commonly found in role-playing games, into several sets of design patterns at different levels of abstraction. The analysis focuses on mechanical patterns that could be either straightforwardly instantiated or recognized by a well-defined process. At the most concrete level a set of fundamental components were identified, followed by a long list of micropatterns which can be directly instantiated. Shorter lists of meso- and macro-patterns, which can be identified mechanically, are also identified. The direct motivation for this analysis is to find building blocks and objectives for a search-based procedural dungeon generator, however we believe the analysis can be useful for understanding this class of game artifacts in general. In particular, the constraints on patterns being instantiable or recognizable leads to a stricter pattern analysis than many other attempts at analyzing game design.