No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Demonstrating the Feasibility of Automatic Game Balancing
Abstract
Game balancing is an important part of the (computer) game design process, in which designers adapt a game prototype so that the resulting gameplay is as entertaining as possible. In industry, the evaluation of a game is often based on costly playtests with human players. It suggests itself to automate this process using surrogate models for the prediction of gameplay and outcome. In this paper, the feasibility of automatic balancing using simulation- and deck-based objectives is investigated for the card game top trumps. Additionally, the necessity of a multi-objective approach is asserted by a comparison with the only known (single-objective) method. We apply a multi-objective evolutionary algorithm to obtain decks that optimise objectives, e.g. win rate and average number of tricks, developed to express the fairness and the excitement of a game of top trumps. The results are compared with decks from published top trumps decks using simulation-based objectives. The possibility to generate decks better or at least as good as decks from published top trumps decks in terms of these objectives is demonstrated. Our results indicate that automatic balancing with the presented approach is feasible even for more complex games such as real-time strategy games.
Supplementary resource (1)
... What makes this process even harder to solve is that there is no clear or accepted definition of what balancing means in the context of video games, or even what a well-balanced state or configuration of a game constitutes. In scientific context, balancing is predominantly covered from the angle of automatic or dynamic game balancing [5,83,85,103,107], which however plays only a minor role for industrials [84], as opposed to the predominantly similar skill is a promising strategy to lead to 50/50 win rates [27]. Yet, Claypool et al. discovered that even if winning chances can be closely approximated to 50%, such as in the highly populated matchmaking of League of Legends, players still subjectively perceive these as unbalanced [23]. ...
... In this respect, if perfect matches are not attainable, mental overload is still seen as producing higher enjoyment than boredom [54]. Andrade et al. claim that "game balancing aims at providing a good level of challenge for the user" [5], while Volz et al. keep it more general by defining the goal of the adjustment to make sure that "the resulting gameplay is as entertaining as possible" [107]. While this "modification of parameters of the constitutive and operational rules of a game" [93] follows technical procedures similar to the adjustment of parameters for adjusting the viability, symmetry or fairness of in-game choices or strategies, the underlying agenda and purpose considerably differs. ...
... This equally bears implications for recent scientific work that explicitly balanced towards symmetrical outcomes in the same or similar game genres [79]. Above that, in contrast to academic literature on difficulty balancing [5,103,107], we observed a clear trend in the community that sees difficulty not as the variable to adjust. Rather, they desire a variety of viable options that differ in difficulty (without being symmetrical in strengths, weaknesses and use cases) -while the adjusted metric is rather the damage potential (and variance) it can offer, based on that difficulty. ...
Balancing is, especially among players, a highly debated topic of video games. Whether a game is sufficiently balanced greatly influences its reception, player satisfaction, churn rates and success. Yet, conceptions about the definition of balance diverge across industry, academia and players, and different understandings of designing balance can lead to worse player experiences than actual imbalances. This work accumulates concepts of balancing video games from industry and academia and introduces a player-driven approach to optimize player experience and satisfaction. Using survey data from 680 participants and empirically recorded data of over 4 million in-game fights of Guild Wars 2, we aggregate player opinions and requirements, contrast them to the status quo and approach a democratized quantitative technique to approximate closer configurations of balance. We contribute a strategy of refining balancing notions, a methodology of tailoring balance to the actual player base and point to an exemplary artifact that realizes this process.
... Also, no publications seem to exist addressing game balancing for serious games and other types of games focusing on other goals than fun. Few scientific publications such as [1], [3], [4], and [2] focus on game balancing and especially automated or even dynamic game balancing, instead relying on definitions from practitioners or adopting simplified reinterpretations of the term relating to their respective goals (e.g., [2]). In section 3, we therefore present the practitioners' definitions first before continuing with the state of the art in scientific publications. ...
... Also, no publications seem to exist addressing game balancing for serious games and other types of games focusing on other goals than fun. Few scientific publications such as [1], [3], [4], and [2] focus on game balancing and especially automated or even dynamic game balancing, instead relying on definitions from practitioners or adopting simplified reinterpretations of the term relating to their respective goals (e.g., [2]). In section 3, we therefore present the practitioners' definitions first before continuing with the state of the art in scientific publications. ...
... In order to demonstrate the feasibility of automating the game balancing process, Volz et al. [2] treat game balancing as a multi-objective optimization problem, assuming that "game balancing is an important part of the (computer)game design process, in which designers adapt a game prototype so that the resulting gameplay is as entertaining as possible." Similar to Schreiber [18], they define game balancing "as the modification of parameters of the constitutive and operational rules of a game (...) in order to achieve optimal configurations in terms of a set of goals, i.e. a parameter tuning problem." ...
Balancing is commonly considered critical to any game’s success. However, there is no consensus on what “game balancing” actually means. This paper examines fourteen publications by experienced game designers and other renowned authors. The results of a formal semantic analysis of the respective authors’ reflections prove that no two authors share identical understandings of “game balancing.” Their differing concepts of the term are analyzed and discussed in order to identify key aspects, similarities, and differences. Contrary to what one might expect, concepts such as fairness, flow, or user satisfaction are identified to be important, but no central or fundamental concepts for the definition of game balancing.
... The two problems we will focus on have been published previously [21,22] along with their respective results. According to the framework proposed in [9], both problems can be classified as level generation methods with embedded input. ...
... As the problem requires the generation of a deck, the predetermined number of cards in a deck and/or the number of values on each card can be modified to create scalable problems (R-IV). The original publication [21] already contains diverse functions with differing numbers of objectives (R-I). Furthermore, AIs of different skill levels are already implemented (R-I). ...
... As expected in game optimisation problems, the included functions are noisy. However, the fitness for each solution is reported as the average of 2 000 simulations, which has been shown in [21] to produce an appropriate balance between computational effort (R-II) and resulting standard deviations. The only remaining issue is to create suitable instances of the functions (R-III), which on the one hand create fitness landscapes of similar type and structure, but on the other hand do not share the locations of e.g. ...
Despite a large interest in real-world problems from the research field of evolutionary optimisation, established benchmarks in the field are mostly artificial. We propose to use game optimisation problems in order to form a benchmark and implement function suites designed to work with the established COCO benchmarking framework. Game optimisation problems are real-world problems that are safe, reasonably complex and at the same time practical, as they are relatively fast to compute. We have created four function suites based on two optimisation problems previously published in the literature (TopTrumps and MarioGAN). For each of the applications, we implemented multiple instances of several scalable single- and multi-objective functions with different characteristics and fitness landscapes. Our results prove that game optimisation problems are interesting and challenging for evolutionary algorithms.
... Manually balancing a game, however, requires substantial work and testing in the development process of a game [2]. Automated balancing is therefore an important and active field of research [3]- [8]. Recently, we proposed a method using the Procedural Content Generation via Reinforcement Learning (PCGRL [9]) framework to balance existing tile-based game levels [10], [11]. ...
... Since the creation and balancing of game content requires a lot of manual effort including multiple human playtesters for instance, many works [3]- [8], [13] aim on automating this process using procedural content generation. Volz Similar to our work [10], [11], Karavolos et al. [5] use a PCG method for automated game level design using artificial agents to simulate game play. ...
Achieving optimal balance in games is essential to their success, yet reliant on extensive manual work and playtesting. To facilitate this process, the Procedural Content Generation via Reinforcement Learning (PCGRL) framework has recently been effectively used to improve the balance of existing game levels. This approach, however, only assesses balance heuristically, neglecting actual human perception. For this reason, this work presents a survey to empirically evaluate the created content paired with human playtesting. Participants in four different scenarios are asked about their perception of changes made to the level both before and after balancing, and vice versa. Based on descriptive and statistical analysis, our findings indicate that the PCGRL-based balancing positively influences players' perceived balance for most scenarios, albeit with differences in aspects of the balancing between scenarios.
... Game Analytics has emerged as an umbrella term for the various data scientific endeavors in industry and academia that deal with data analysis for game production, game performance, and player behavior understanding [2]. To achieve such goals, literature has proposed different tasks such as game balancing [3], churn prediction [4], and detection of failures during game design [5]. As a task-oriented and data-driven field, Game Analytics often requires telemetrybased approaches for game session data recording, whose outcomes can be submitted to a plethora of methods and B Sidney Melo sidneymelo@id.uff.br ...
... Through Game Provenance Graphs, one can visually interpret game sessions and identify bugs and unexpected behaviors in the game, both after release and during its development. It is also possible to use actual gameplay data for several pertinent tasks, such as player profiling and gameplay-related predictions, to improve the game experience [3,4,9]. However, games tend to generate a high volume of data, considering that multiple players can play multiple game sessions. ...
Game Provenance has been proposed and employed for Game Analytics tasks as they capture game session data in detail and allow exploratory analysis and visualizations. Games are highly heterogeneous models with several interacting agents and game-world environment elements. Game Provenance Graphs can accommodate the heterogeneous nature of such applications with different types of nodes and edges that tend to share information across themselves, enhancing cause-effect features rarely addressed by any other approach. On the other hand, existing Heterogeneous Graph Neural Network (HGNN) solutions disregard node feature information, overlooking shared features across distinct node types, and rely on naïve approaches, such as projecting each type of node to the same n-dimensional space. We conjecture that leveraging heterogeneous feature information is essential for tackling Game Analytics tasks, especially through Machine Learning based models. To achieve that, we propose a novel approach that allows HGNNs to leverage Game Provenance Graphs’ heterogeneous node feature information. Hence, we introduce in this paper three strategies for Heterogeneous Graph Representation Learning that encodes feature set information into the HGNN architecture and projects feature values leveraging similarities across such feature sets. We conduct experiments on two Game Provenance Graphs datasets, the Smoke Squadron and the Game Provenance Profile datasets, which gather game session data from different games. Our results show that encoding feature set information in the representation learning process improves the outcomes of GNN models in non-disjoint feature datasets.
Graphical abstract
... At its simplest, one could isolate one or a small handful of continuous values, and optimize those values so that the game achieves parity in winrate between agents. This is often referred to as game balancing or game tuning [24]. But approaches have been pursued that identify a larger number of game dimensions, demonstrating that search in a multi-dimensional game space can be used to find games that are distinctly different from each other but still playable [12]. ...
... Jaffe et al. [13] introduces a method for measuring aspects of game balance [13], and shows how the techniques can facilitate quick progress in balancing an educational card game. Volz et al. [24] introduces another balancing method for decks of the Top Trumps game. Here the authors compare the processes of automatic and manual balancing and assess the quality of computer balanced decks in relation to existing ones. ...
A significant issue in game balancing is understanding the game itself. For simple games end-to-end optimization approaches can help explore the game's design space, but for more complex games it is necessary to isolate and explore its parts. Hearthstone, Blizzard's popular two-player turn-taking adversarial card game, has two distinct game-playing challenges: choosing when and how to play cards, and selecting which cards a player can access during the game (deckbuilding). Focusing on deckbuilding, four experiments are conducted to computationally explore the design of Hearthstone. They address the difficulty of constructing good decks, the specificity and generality of decks, and the transitivity of decks. Results suggest it is possible to find decks with an Evolution Strategy (ES) that convincingly beat other decks available in the game, but that they also exhibit some generality (i.e. they perform well against unknown decks). Interestingly, a second ES experiment is performed where decks are evolved against opponents playing the originally evolved decks. Since the originally evolved decks beat the starter decks, and the twice evolved decks beat the originally evolved decks, some degree of transitivity of the deck space is shown. While only a preliminary study with restrictive conditions, this paper paves the way for future work computationally identifying properties of cards important for different gameplay strategies and helping players build decks to fit their personal playstyles without the need for in-depth domain knowledge.
... Fuzzy logic, which is frequently employed in adaptive systems such as educational games, effectively regulates equivocal data, thereby improving player satisfaction and adaptability [33]- [35]. The incorporation of fuzzy logic with other AI techniques, such as Q-learning, has resulted in substantial benefits [36], [37]. ...
Maintaining player engagement in serious management games is a challenge due to the repetitive nature of traditional predetermined difficulty levels. A dynamic difficulty adjustment (DDA) system is introduced in this study to address this issue by integrating fuzzy logic and Q-learning. Player ennui is frequently the consequence of static difficulty adjustments. In order to dynamically adjust game complexity in accordance with player performance and preferences, our DDA system utilizes a diverse array of performance metrics, adaptive narrative elements, and real-time feedback, as well as fuzzy logic and Q-learning algorithms. According to empirical assessments, players were 28% more effective overall, and play sessions lasted an average of 35% longer. Player satisfaction and involvement were also much improved. Customers played the game longer and were less bored because of the higher degree of difficulty and customization. The integration of fuzzy logic and Q-learning in DDA systems greatly enhances the ability to maintain long-term player engagement in essential management games. This approach offers a long-lasting alternative for creating constantly captivating gaming experiences by effectively reducing the repetitiveness of traditional difficulty adjustments.
... Evolutionary Fuzzy Cognitive Maps (E-FCM) ensure equilibrium by adjusting the weights in real time. Volz et al. [2016] applied AI agents to a set of mechanics to describe what makes a game balanced and enjoyable. Beyer et al. [2016] used machine learning algorithms to solve a development problem. ...
This work explores the balancing of an educational game to teach sustainable development in organizations by focusing on player interaction and employing strategies. Game success is a challenge that relies on balancing the relationships among its elements. Balancing is a complex process performed over multiple iterations, starting at game conception and continuing throughout development and testing stages. This work extends our previous case study, which did not consider player interaction for the game balancing. We built two models that contains all game mechanics using the Machinations framework. The first model includes elements that randomly produce, distribute, and consume resources, while the second model analyzes player interaction and implements four player strategies. We simulated these models in batch plays, analyzed game states, and adjusted game economies. The random model simulation achieved a victory rate of 40%, while the interactive model simulation with player strategies increased victory rates to values between 66% and 81%. These results show that player interaction and decision-making can be more decisive than randomness in achieving victory. Machinations contributed to enhancing the game, proved its usefulness for simulating complex models, and deepened our understanding of game dynamics, including player actions, potential deadlocks, and feedback mechanisms. This work supports other authors’ findings by demonstrating that balancing the game as early as possible in the development process, considering player interaction, makes the design feasible; and provides evidence that computer simulations, such as Machinations, benefit the game balance and improve the game design without the need to build a prototype and conduct extensive playtests.
... In this respect, if perfect matches are not attainable, mental overload is still seen as producing higher enjoyment than boredom [34]. Andrade et al. claim that "game balancing aims at providing a good level of challenge for the user" [4], while Volz et al. keep it more general by defining the goal of the adjustment to make sure that "the resulting gameplay is as entertaining as possible" [66]. While this "modification of parameters of the constitutive and operational rules of a game" [57] follows technical procedures similar to the adjustment of parameters for adjusting the viability, symmetry or fairness of in-game choices or strategies, the underlying agenda and purpose considerably differs. ...
Balancing is, especially among players, a highly debated topic of video games. Whether a game is sufficiently balanced greatly influences its reception, player satisfaction, churn rates and success. Yet, conceptions about the definition of balance diverge across industry, academia and players, and different understandings of designing balance can lead to worse player experiences than actual imbalances. This work accumulates concepts of balancing video games from industry and academia and introduces a player-driven approach to optimize player experience and satisfaction. Using survey data from 680 participants and empirically recorded data of over 4 million in-game fights of Guild Wars 2, we aggregate player opinions and requirements, contrast them to the status quo and approach a democratized quantitative technique to approximate closer configurations of balance. We contribute a strategy of refining balancing notions, a methodology of tailoring balance to the actual player base and point to an exemplary artifact that realizes this process.
... It does so by undertaking a learning path through a sequence of steps in which it picks two random cards from the deck and then analyses and compares them with random criteria. According to the winning result, the model iteratively updates its knowledge base in the same manner as a human, following the rule that 'practice makes perfect.' Hence the model will play, collect statistics, update, and iterate while becoming more accurate with each increment (Volz et al., 2016). ...
Heuristics are often characterized as rules of thumb that can be used to speed up the process of decision-making. They have been examined across a wide range of fields, including economics, psychology, and computer science. However, scholars still struggle to find substantial common ground. This study provides a historical review of heuristics as a research topic before and after the emergence of the subjective expected utility (SEU) theory, emphasising the evolutionary perspective that considers heuristics as resulting from the development of the brain. We find it useful to distinguish between deliberate and automatic uses of heuristics, but point out that they can be used consciously and subconsciously. While we can trace the idea of heuristics through many centuries and fields of application, we focus on the evolution of the modern notion of heuristics through three waves of research, starting with Herbert Simon in the 1950s, who introduced the notion of bounded rationality and suggested the use of heuristics in artificial intelligence, thereby paving the way for all later research on heuristics. A breakthrough came with Daniel Kahneman and Amos Tversky in the 1970s, who analysed the biases arising from using heuristics. The resulting research programme became the subject of criticism by Gerd Gigerenzer in the 1990s, who argues that an ‘adaptive toolbox’ consisting of ‘fast-and-frugal’ heuristics can yield ‘ecologically rational’ decisions.
... (Togelius and Schmidhuber 2008;Liu et al. 2017;Browne and Maire 2010;Cook, Colton, and Gow 2016a)). This approach has also been used extensively for game balancing, where the game is mostly fixed except for optimizable parameters (Beau and Bakkes 2016;Volz, Rudolph, and Naujoks 2016;Preuss et al. 2018). Another approach involves leveraging learning techniques that strive for optimal gameplay strategy as in (Yu and Sturtevant 2019), though these methods are likely too computationally costly to have in an AGD evaluation function. ...
Automated game design (AGD) research focuses on creating systems that can design entirely new games. This is often done by a genetic algorithm, with a fitness function that is used to find individual games that satisfy certain design criteria. However, it is difficult to tell if generated games actually have the desired emergent properties (such as balance), since the fitness function might not align well with human aesthetic judgments about such properties. This is particularly problematic when trying to automatically design balanced, fair, yet asymmetrical games for multiple players. In this paper we present an implementation of an optimization-based AGD system for brawler games, and present findings from a preliminary user study of generated games. We show that while the system successfully optimizes for our written fitness function during human play, we found that this optimization did not necessarily translate to our hypothesized human experience of the game.
... In [25], the authors present an approach for balancing games through the use of restricted play agents. In [61], the authors make use of a multi-objective optimization algorithm to demonstrate the feasibility of an approach for automatic game balancing in the context of the Top Trumps card game. In [5], the authors propose a semi-automatic process for the game balancing of a prototype of a commercial video game (Zombie Village Game by Blue Byte GmbH). ...
We present our Evolutionary Boss Improvement (EBI) approach, which receives partially complete bosses as input and generates fully equipped bosses that are complete. Additionally, the evolutionary algorithm and the new genetic operations included in EBI favor genetic improvement, which affects the initial partial content of the incomplete bosses originally provided. We evaluate our approach using Kromaia, a commercial video game released on PlayStation 4 and PC. EBI uses an evolutionary algorithm to evolve a population of bosses guided by duels between the bosses being generated and a simulated player. Our approach evaluates the quality, in terms of game experience, of both the bosses generated and those included in Kromaia using six metrics (Completion, Duration, Uncertainty, Killer Moves, Permanence, and Lead Change) from the literature. The results show that the quality of the bosses created by EBI is comparable to the quality of the original bosses that were manually created by the developers of Kromaia. However, the EBI approach reduces the time required to build the bosses from five months (of elapsed time as opposed to dedicated time) to just 100 minutes of unattended run. EBI enables developers to accelerate the creation of content, such as bosses, which is essential to ensure player engagement.
... Regarding balancing, scientific approaches often build on simulations that iteratively assess balance criteria and dynamically tune in-game parameters based on the former. Jaffe et al. [31], García-Sanchez et al. [32], Volz et al. [33], Zook et al. [34] and De Mesentier Silva et al. [35] applied this paradigm to board or card games, which was amplified by Mahlmann et al. [36] by introducing procedurally generated cards on top of these simulations. In other genres, Beau and Bakkes [37] utilized Monte-Carlo Tree Search for balancing units of Tower Defense games while Keehl and Smith [38] extended this to generic Unity games, Morosan and Poli [39] tweaked difficulty specifications in RTS and Arcade games after neuroevolution agents assessed these and Leigh et al. [40] dynamically balanced strategies though the coevolution of two competing agents playing a Capture The Flag game. ...
Video game testing has become a major investment of time, labor and expense in the game industry. Particularly the balancing of in-game units, characters and classes can cause long-lasting issues that persist years after a game’s launch. While approaches incorporating artificial intelligence have already shown successes in reducing manual effort and enhancing game development processes, most of these draw on heuristic, generalized or optimal behavior routines, while actual low-level decisions from individual players and their resulting playing styles are rarely considered. In this paper, we apply Deep Player Behavior Modeling to turn atomic actions of 213 players from 6 months of single-player instances within the MMORPG Aion into generative models that capture and reproduce particular playing strategies. In a subsequent simulation, the resulting generative agents (“replicants”) were tested against common NPC opponent types of MMORPGs that iteratively increased in difficulty, respective to the primary factor that constitutes this enemy type (Melee, Ranged, Rogue, Buffer, Debuffer, Healer, Tank or Group). As a result, imbalances between classes as well as strengths and weaknesses regarding particular combat challenges could be identified and regulated automatically.
... The core constraint is usually playability -aimed to ensure that a game can actually be used as such. One common approach here is to use AI gameplay [24] to test that a game can be won, or to determine basic imbalances [22,41]. Since generating content is usually fast and cheap [25], the aim is usually to test content automatically -but other approaches with a human in the loop exist. ...
There are a range of metrics that can be applied to the artifacts produced by procedural content generation, and several of them come with qualitative claims. In this paper, we adapt a range of existing PCG metrics to generated Minecraft settlements, develop a few new metrics inspired by PCG literature, and compare the resulting measurements to existing human evaluations. The aim is to analyze how those metrics capture human evaluation scores in different categories, how the metrics generalize to another game domain, and how metrics deal with more complex artifacts. We provide an exploratory look at a variety of metrics and provide an information gain and several correlation analyses. We found some relationships between human scores and metrics counting specific elements, measuring the diversity of blocks and measuring the presence of crafting materials for the present complex blocks.
... However, in the domain of games, academic literature is sparse. In [35], a static hand-crafted fitness function is used as a surrogate for gameplay; however, this function is not learned from observed simulation data. Morosan et al. [36] trained surrogate models in order to finetune parameters of a Ms. Pac-man (Midway, 1982) agent, a racing car in TORCS [37] and unit parameters in StarCraft (Blizzard, 1998). ...
This paper introduces SuSketch, a design tool for first person shooter levels. SuSketch provides the designer with gameplay predictions for two competing players of specific character classes. The interface allows the designer to work side-by-side with an artificially intelligent creator and to receive varied types of feedback such as path information, predicted balance between players in a complete playthrough, or a predicted heatmap of the locations of player deaths. The system also proactively designs alternatives to the level and class pairing, and presents them to the designer as suggestions that improve the predicted balance of the game. SuSketch offers a new way of integrating machine learning into mixed-initiative co-creation tools, as a surrogate of human play trained on a large corpus of artificial playtraces. A user study with 16 game developers indicated that the tool was easy to use, but also highlighted a need to make SuSketch more accessible and more explainable.
... However, in the domain of games, academic literature is sparse. In [35], a static hand-crafted fitness function is used as a surrogate for gameplay; however, this function is not learned from observed simulation data. Morosan et al. [36] trained surrogate models in order to finetune parameters of a Ms. Pac-man (Midway, 1982) agent, a racing car in TORCS [37] and unit parameters in StarCraft (Blizzard, 1998). ...
This paper introduces SuSketch, a design tool for first person shooter levels. SuSketch provides the designer with gameplay predictions for two competing players of specific character classes. The interface allows the designer to work side-by-side with an artificially intelligent creator and to receive varied types of feedback such as path information, predicted balance between players in a complete playthrough, or a predicted heatmap of the locations of player deaths. The system also proactively designs alternatives to the level and class pairing, and presents them to the designer as suggestions that improve the predicted balance of the game. SuSketch offers a new way of integrating machine learning into mixed-initiative co-creation tools, as a surrogate of human play trained on a large corpus of artificial playtraces. A user study with 16 game developers indicated that the tool was easy to use, but also highlighted a need to make SuSketch more accessible and more explainable.
... For them, Game Analytics fundamentally deals with the use of data analysis for game production, game performance, and-crucially-player behavior understanding [10]. The outcomes of Game Analytics tasks, such as game balancing [2] [37], player profiling [28] [35], and detection of failures during game design [40] [16], are employed to enhance the game development process or the game experience itself. In this context, Hooshyar et al. [20] points out that profiling and predicting player behavior is of the utmost importance when developing games. ...
... Outros autores destacam a importância de trabalhos que colaboram com a agilidade de um processo de design utilizando algoritmos genéticos para encontrar parâmetros ideais. Issoé feito no trabalho de Volz et al. [22], no qualé mostrada a viabilidade de automação do balanceamento dos jogos digitais. Os autores realizam a criação de baralhos de top trumps 2 utilizando técnicas de algorítimo genético multiobjetivos. ...
A criação de jogos do gênero RPG requer diversas etapas que estão relacionadas com os seus sistemas. No sistema de combate, a definição das habilidades dos personagensé uma tarefa importante e que muitas vezes envolve balanceamento manual por parte do game designer. Neste trabalho,é proposto um processo de balanceamento automático de atributos de per-sonagens em jogos de RPG para simplificar o processo de criação de personagens e otimizar o tempo utilizado no balanceamento. Para alcançar tais resultados, utiliza-se algorítimo genético para identificação de parâmetros da curva de crescimento de um jogador para que este alcance uma taxa de vitórias pré-determinada contra um inimigo previamente criado pelo game designer. A ferramenta de balanceamento proposta foi capaz de gerar os atributos do personagem para cada um dos níveis dentro da margem de erro definida. Em seguida, foram geradas as curvas de nível iniciais, que são suavizadas para gerar as curvas finais. Uma avaliação experimental utilizou dez níveis de um inimigo com uma taxa de vitória desejada de 80% e margem de erro de 5%. Esses resultados sugerem que o uso do algoritmo genético foi eficaz na geração de curvas de nível, sendo adequada como um processo de balanceamento automático para auxiliar o game designer. Keywords-balanceamento automático, sistema de combate de RPG, algoritmo genético I. INTRODUÇÃO Um dos elementos centrais dos jogos de Role Playing Game (RPG) estão as suas regras, os chamados sistemas [1], os quais definem parte do gameplay e das interações com o cenário. Esses sistemas costumam oferecer progressões de níveis e um certo grau de customização dos personagens aos jogadores. Uma dessas progressõesé a chamada curva de nível, que apresenta os valores das habilidades de uma entidade para cada nível. Definir essas progressões prova-se muitas vezes como um desafio ao game designer quando do planejamento das representações adequadas dos personagens dos jogadores e dos inimigos para cada nível [2]. Uma modelagem inadequada pode representar uma frustração na experiência do jogador, seja pelo tédio de ser muito fácil, seja pela ansiedade por um desafio muito difícil [3]. Um dos elementos cobertos comumente em um sistemaé o combate. De fato, boa parte do desafio dos jogos envolve o confronto do jogador com personagens controlados pela máquina ou por outro participante. Em jogos baseados em turno, essa experiência torna-se menos difícil de mensu-rar, uma vez que as representações de habilidade permitem simulações de batalhas e assim obter uma estimativa da taxa de vitórias e derrotas de cada um dos oponentes. Mais ainda, e possível determinar a melhor distribuição de recursos para a progressão dentro do jogo. Dessa forma, para manter um combate balanceadoé de grande importância que o game designer construa uma progressão adequada das habilidades tanto dos jogadores quanto dos inimigos. Dada a sua importância, a definição dos atributos dos jogadoresé um processo realizado com bastante cuidado. Comumente esse processoé realizado manualmente e por vezes com base na tentativa e erro [3]. Dessa forma, estratégias de teste e validação das representações de personagens são recursos importantes ao processo de game design. No en-tanto, elas podem ser bastante custosas, apesar de indicar se há problemas que gerem dúvidas, ambiguidades, estratégias dominantes, entre outras [3]. Trabalhos envolvendo jogabilidade e jogos digitais cos-tumam ter um apelo forte naárea da Educação, Saúde, Publicidade e Cultura. Estudos que buscam compreender ou melhorar os processos de design são relativamente novos e estão mais vinculados a trabalhos de designers experientes e consagrados em discussões publicadas como livros [3], [4] do que em revisões sistemáticas pela comunidade acadêmica [5]. Ainda assim, há trabalhos pioneiros recentes que exploram meandros das mecânicas, como Maranhão et al. [5], que buscam avaliações estéticas envolvendo elementos específicos dos personagens, como Islam et al. [6], ou ainda em Faria e Pereira [7], onde são gerados personagens para Tabletop Role
... The topic closely correlated with deckbuilding is how to design the cards in CCG to be balanced [16]. In [17], evolution is used to propose changes to the cards that will result in better balance. ...
In this paper, we evolve a card-choice strategy for the arena mode of Legends of Code and Magic, a programming game inspired by popular collectible card games like Hearthstone or TES: Legends. In the arena game mode, before each match, a player has to construct his deck choosing cards one by one from the previously unknown options.
Such a scenario is difficult from the optimization point of view, as not only the fitness function is non-deterministic, but its value, even for a given problem instance, is impossible to be calculated directly and can only be estimated with simulation-based approaches.
We propose a variant of the evolutionary algorithm that uses a concept of an active gene to reduce the range of the operators only to generation-specific subsequences of the genotype. Thus, we batched learning process and constrained evolutionary updates only to the cards relevant for the particular draft, without forgetting the knowledge from the previous tests.
We developed and tested various implementations of this idea, investigating their performance by taking into account the computational cost of each variant. Performed experiments show that some of the introduced active-genes algorithms tend to learn faster and produce statistically better draft policies than the compared methods.
... The topic closely correlated with deckbuilding is how to design the cards in CCG to be balanced [16]. In [17], evolution is used to propose changes to the cards that will result in better balance. ...
In this paper, we evolve a card-choice strategy for the arena mode of Legends of Code and Magic, a programming game inspired by popular collectible card games like Hearthstone or TES: Legends. In the arena game mode, before each match, a player has to construct his deck choosing cards one by one from the previously unknown options. Such a scenario is difficult from the optimization point of view, as not only the fitness function is non-deterministic, but its value, even for a given problem instance, is impossible to be calculated directly and can only be estimated with simulation-based approaches. We propose a variant of the evolutionary algorithm that uses a concept of an active gene to reduce the range of the operators only to generation-specific subsequences of the genotype. Thus, we batched learning process and constrained evolutionary updates only to the cards relevant for the particular draft, without forgetting the knowledge from the previous tests. We developed and tested various implementations of this idea, investigating their performance by taking into account the computational cost of each variant. Performed experiments show that some of the introduced active-genes algorithms tend to learn faster and produce statistically better draft policies than the compared methods.
... Acquiring understandable game metrics is essential to enhance a data-driven game design. This information may be useful for many purposes, such as game balancing [1] [2], players behavior understanding [3] [4], detection of failures during game design [5] [6], or even enhancing in-game monetizing strategies [7] [8]. ...
Game Analytics comprises a set of techniques to analyze both the game quality and player behavior. To succeed in Game Analytics, it is essential to identify what is happening in a game (an effect) and track its causes. Thus, game provenance graph tools have been proposed to capture cause-and-effect relationships occurring in a gameplay session to assist the game design process. However, since game provenance data capture is guided by a set of strict predefined rules established by the game developers, the detection of long-range cause-and-effect relationships may demand huge coding efforts. In this paper, we contribute with a framework named PingUMiL that leverages the recently proposed graph embeddings to represent game provenance graphs in a latent space. The embeddings learned from the data pose as the features of a machine learning task tailored towards detecting long-range cause-and-effect relationships. We evaluate the generalization capacity of PingUMiL when learning from similar games and compare its performance to classical machine learning methods. The experiments conducted on two racing games show that (1) PingUMiL outperforms classical machine learning methods and (2) representation learning can be used to detect long-range cause-and-effect relationships in only partially observed game data provenance graphs.
... Jaffe et al. developed a tool that calculates balance metrics of the gameplay between restricted and standard agents, and applied such to an educational card game [25]. Volz et al. used a multi-objective algorithm to create Top Trump decks, with win rate as one of the dimensions [26]. While these approaches are similar to the work we present, they are applied to games with lower complexity. ...
Balancing an ever growing strategic game of high
complexity, such as Hearthstone is a complex task. The target of
making strategies diverse and customizable results in a delicate
intricate system. Tuning over 2000 cards to generate the desired
outcome without disrupting the existing environment becomes a
laborious challenge. In this paper, we discuss the impacts that
changes to existing cards can have on strategy in Hearthstone.
By analyzing the win rate on match-ups across different decks,
being played by different strategies, we propose to compare their
performance before and after changes are made to improve or
worsen different cards. Then, using an evolutionary algorithm,
we search for a combination of changes to the card attributes
that cause the decks to approach equal, 50% win rates. We then
expand our evolutionary algorithm to a multi-objective solution
to search for this result, while making the minimum amount of
changes, and as a consequence disruption, to the existing cards.
Lastly, we propose and evaluate metrics to serve as heuristics with
which to decide which cards to target with balance changes.
... The comparative balance of specific material, rather than strategies, is analysed using autonomous agents and restricted game mechanics in [6]. The balance of the card game, Top Trumps, is optimised using both single-objective [7] and multi-objective optimisation [8], although the cards in Top Trumps are not orthogonally differentiated. Reinforcement learning (RL) techniques have been used for dynamic difficulty adjustment of single player games [9], [10], but RL has not yet been adopted as a tool to balance multiplayer games. ...
Probabilistic model checking can overcome much of the complexity inherent in balancing games. Game balancing is the careful maintenance of relationships between the ways in which a game can be played, to ensure no single way is strictly better than all others and that players are offered a wide variety of ways to play successfully. We introduce a novel approach towards automating game balancing using probabilistic model checking called chained strategy generation (CSG). This involves generating chains of adversarial strategies which mimic the way players adapt their approach during repeated plays of a game. We use CSG to map out the evolving metagame. The trends identified can allow game developers to identify strategies which will be too strong and ways of playing the game which a player may want to use, but are never viable for successful competitive play. We introduce a case study, a game called RPGLite, and use CSG to compare five candidate configurations for the game. We show how to determine which configurations of RPGLite lead to a more fair and interesting experience for players. We also identify unexpected trends in how the strategies evolve. Our approach introduces a new technique for improving game development and player experience.
... Another effort towards the study of automatic game balancing was put forth by Volz et al. [43]. and player skill [31]. ...
Collectible Card Games (CCG) are one of the most popular types of games in both digital and physical space. Despite their popularity, there is a great deal of room for exploration into the application of artificial intelligence in order to enhance CCG gameplay and development. This paper presents Fifth Aeon a novel and open source CCG built to run in browsers and two A.I applications built upon Fifth Aeon. The first application is an artificial intelligence competition run on the Fifth Aeon game. The second is an automatic balancing system capable of helping a designer create new cards that do not upset the balance of an existing collectible card game. The submissions to the A.I competition include one that plays substantially better than the existing Fifth Aeon A.I with a higher winrate across multiple game formats. The balancer system also demonstrates an ability to automatically balance several types of cards against a wide variety of parameters. These results help pave the way to cheaper CCG development with more compelling A.I opponents.
... http://hanabi.fosslab.uk/ 6 Tension has been described as the number of times the lead changed during the game as well as closeness in score[12]. ...
Hyperparameter tuning is an important mixed-integer optimisation problem, especially in the context of real-world applications such as games. In this paper, we propose a function suite around hyperparameter optimisation of game AI based on the card game Splendor and using the Rinascimento framework. We propose several different functions and demonstrate their complexity and diversity. We further suggest various possible extensions of the proposed suite.
... Jaffe et al. developed a tool that calculates balance metrics of the gameplay between restricted and standard agents, and applied such to an educational card game [25]. Volz et al. used a multi-objective algorithm to create Top Trump decks, with win rate as one of the dimensions [26]. While these approaches are similar to the work we present, they are applied to games with lower complexity. ...
Balancing an ever growing strategic game of high complexity, such as Hearthstone is a complex task. The target of making strategies diverse and customizable results in a delicate intricate system. Tuning over 2000 cards to generate the desired outcome without disrupting the existing environment becomes a laborious challenge. In this paper, we discuss the impacts that changes to existing cards can have on strategy in Hearthstone. By analyzing the win rate on match-ups across different decks, being played by different strategies, we propose to compare their performance before and after changes are made to improve or worsen different cards. Then, using an evolutionary algorithm, we search for a combination of changes to the card attributes that cause the decks to approach equal, 50% win rates. We then expand our evolutionary algorithm to a multi-objective solution to search for this result, while making the minimum amount of changes, and as a consequence disruption, to the existing cards. Lastly, we propose and evaluate metrics to serve as heuristics with which to decide which cards to target with balance changes.
... There are several approaches to evolutionary deckbuilding that this paper builds upon. Volz et al. [44] explore evolutionary deckbuilding for the game Top Trumps that optimized different objectives with a goal of expressing fairness in the game. However, in Top Trumps all of the cards in a given pack are distributed to all of the players, whereas Hearthstone players build their own decks individually and with intention. ...
Quality diversity (QD) algorithms such as MAP-Elites have emerged as a powerful alternative to traditional single-objective optimization methods. They were initially applied to evolutionary robotics problems such as locomotion and maze navigation, but have yet to see widespread application. We argue that these algorithms are perfectly suited to the rich domain of video games, which contains many relevant problems with a multitude of successful strategies and often also multiple dimensions along which solutions can vary.
This paper introduces a novel modification of the MAP-Elites algorithm called MAP-Elites with Sliding Boundaries (MESB) and applies it to the design and rebalancing of Hearthstone, a popular collectible card game chosen for its number of multidimensional behavior features relevant to particular styles of play. To avoid overpopulating cells with conflated behaviors, MESB slides the boundaries of cells based on the distribution of evolved individuals. Experiments in this paper demonstrate the performance of MESB in Hearthstone. Results suggest MESB finds diverse ways of playing the game well along the selected behavioral dimensions. Further analysis of the evolved strategies reveals common patterns that recur across behavioral dimensions and explores how MESB can help rebalance the game.
... Jaffe et al. developed a tool that calculates balance metrics of the gameplay between restricted and standard agents, and applied such to an educational card game [25]. Volz et al. used a multi-objective algorithm to create Top Trump decks, with win rate as one of the dimensions [26]. While these approaches are similar to the work we present, they are applied to games with lower complexity. ...
Balancing an ever growing strategic game of high complexity, such as Hearthstone is a complex task. The target of making strategies diverse and customizable results in a delicate intricate system. Tuning over 2000 cards to generate the desired outcome without disrupting the existing environment becomes a laborious challenge. In this paper, we discuss the impacts that changes to existing cards can have on strategy in Hearthstone. By analyzing the win rate on match-ups across different decks, being played by different strategies, we propose to compare their performance before and after changes are made to improve or worsen different cards. Then, using an evolutionary algorithm, we search for a combination of changes to the card attributes that cause the decks to approach equal, 50% win rates. We then expand our evolutionary algorithm to a multi-objective solution to search for this result, while making the minimum amount of changes, and as a consequence disruption, to the existing cards. Lastly, we propose and evaluate metrics to serve as heuristics with which to decide which cards to target with balance changes.
... There are several approaches to evolutionary deckbuilding that this paper builds upon. Volz et al. [44] explore evolutionary deckbuilding for the game Top Trumps that optimized different objectives with a goal of expressing fairness in the game. However, in Top Trumps all of the cards in a given pack are distributed to all of the players, whereas Hearthstone players build their own decks individually and with intention. ...
Quality diversity (QD) algorithms such as MAP-Elites have emerged as a powerful alternative to traditional single-objective optimization methods. They were initially applied to evolutionary robotics problems such as locomotion and maze navigation, but have yet to see widespread application. We argue that these algorithms are perfectly suited to the rich domain of video games, which contains many relevant problems with a multitude of successful strategies and often also multiple dimensions along which solutions can vary. This paper introduces a novel modification of the MAP-Elites algorithm called MAP-Elites with Sliding Boundaries (MESB) and applies it to the design and rebalancing of Hearthstone, a popular collectible card game chosen for its number of multidimensional behavior features relevant to particular styles of play. To avoid overpopulating cells with conflated behaviors, MESB slides the boundaries of cells based on the distribution of evolved individuals. Experiments in this paper demonstrate the performance of MESB in Hearthstone. Results suggest MESB finds diverse ways of playing the game well along the selected behavioral dimensions. Further analysis of the evolved strategies reveals common patterns that recur across behavioral dimensions and explores how MESB can help rebalance the game.
The balancing process for game levels in competitive two-player contexts involves a lot of manual work and testing, particularly for non-symmetrical game levels. In this work, we frame game balancing as a procedural content generation task and propose an architecture for automatically balancing of tile-based levels within the PCGRL framework (procedural content generation via reinforcement learning). Our architecture is divided into three parts: (1) a level generator, (2) a balancing agent, and (3) a reward modeling simulation. Through repeated simulations, the balancing agent receives rewards for adjusting the level towards a given balancing objective, such as equal win rates for all players. To this end, we propose new swap-based representations to improve the robustness of playability, thereby enabling agents to balance game levels more effectively and quickly compared to traditional PCGRL. By analyzing the agent's swapping behavior, we can infer which tile types have the most impact on the balance. We validate our approach in the Neural MMO (NMMO) environment in a competitive two-player scenario. In this extended conference paper, we present improved results, explore the applicability of the method to various forms of balancing beyond equal balancing, compare the performance to another search-based approach, and discuss the application of existing fairness metrics to game balancing.
Game balancing is a valuable asset to increase player enjoyment and engagement. In this study we look at the impact of balancing in a twin stick shooter game focusing on multiplayer balance, mechanics and player feedback. Psychological, behavioral and physiological metrics were used to analyze player experience. Data from the Game Experience Questionnaire went through statistical analysis, with the Shapiro-Wilk test confirming normality (p > 0.05). The findings underscore the importance of balancing for fair, engaging, and skill-driven gameplay, emphasizing player feedback's role in fostering satisfaction and inclusivity.
In miniature wargames, such as
Warhammer 40k
, players control asymmetrical armies, which include multiple units of different types and strengths. These games often use point costs to balance the armies. Each unit is assigned a point cost, and players have a budget they can spend on units. Calculating accurate point costs can be a tedious manual process, with iterative playtests required. If these point costs do not represent a units true power, the game can get unbalanced as overpowered units can have low point costs. In our previous paper, we proposed an automated way of estimating the point costs using a linear regression approach. We used a turn-based asymmetrical wargame called
Wizard Wars
to test our methods. Players were simulated using Monte Carlo tree search, using different heuristics to represent playstyles. We presented six variants of our method, and show that one method was able to reduce the unbalanced nature of the game by almost half. For this article, we introduce a framework called simple testing and evaluation of points, which allows for further and more granular analysis of point cost estimating methods, by providing a fast, simple, and configurable framework to test methods with. Finally, we compare how our methods do in
Wizard Wars
against expertly chosen point costs.
The balancing process for game levels in competitive two-player contexts involves a lot of manual work and testing, particularly for non-symmetrical game levels. In this work, we frame game balancing as a procedural content generation task and propose an architecture for automatically balancing of tile-based levels within the PCGRL framework (procedural content generation via reinforcement learning). Our architecture is divided into three parts: (1) a level generator, (2) a balancing agent, and (3) a reward modeling simulation. Through repeated simulations, the balancing agent receives rewards for adjusting the level towards a given balancing objective, such as equal win rates for all players. To this end, we propose new swap-based representations to improve the robustness of playability, thereby enabling agents to balance game levels more effectively and quickly compared to traditional PCGRL. By analyzing the agent's swapping behavior, we can infer which tile types have the most impact on the balance. We validate our approach in the Neural MMO (NMMO) environment in a competitive two-player scenario. In this extended conference paper, we present improved results, explore the applicability of the method to various forms of balancing beyond equal balancing, compare the performance to another search-based approach, and discuss the application of existing fairness metrics to game balancing.
Este artigo apresenta o balanceamento de um jogo educacional para ensinar desenvolvimento sustentável nas organizações. O sucesso de um jogo é um desafio que depende do balanço entre seus elementos. O balanceamento é um processo complexo, realizado em várias iterações, que começa na concepção e continua até em jogos de teste. Construímos modelos das mecânicas do jogo no Machinations e simulamos centenas de execuções e adequamos os ganhos e custos dos recursos para atingir os objetivos do jogo. Adotar o Machinations no processo de design do jogo contribuiu para identificar impasses, ações e feedbacks, e aprimorar o design sem a necessidade de construir um protótipo. Este estudo de caso contribuiu para demostrar que balancear o jogo o quanto antes no processo de desenvolvimento viabiliza o design e somou à gama de evidências de que simulações computacionais, como o Machinations, beneficiam o balanceamento de um jogo.
Evolutionary machine learning (EML) has been applied to games in multiple ways, and for multiple different purposes. Importantly, AI research in games is not only about playing games; it is also about generating game content, modeling players, and many other applications. Many of these applications pose interesting problems for EML. We will structure this chapter on EML for games based on whether evolution is used to augment machine learning (ML) or ML is used to augment evolution. For completeness, we also briefly discuss the usage of ML and evolution separately in games.
Metagame balance is a crucial task in game development, and automation of this process could assist game developers by vastly reducing time costs. We explore and evaluate a metagame balance model over the recently proposed VGC AI Competition Framework. We propose an adversarial model where team builder agents try to maximize their win rate by narrowing to the most optimal team configurations, resulting in a reduction of the diversity of
Pokémon
employed, while a balancing agent readapts the
Pokémon
inner attributes to incentivize the team builder agents to incorporate a greater variety of
Pokémon
into their teams increasing the metagame's overall diversity and balance. Furthermore, we develop multiple team builder agents divided into two groups: the first group assumes that individual
Pokémon
advantages are the primary factor to determine the outcome of game matches; the second group also exploits the implicit synergy between teammates. These agents make use of metagaming, linear optimization, and evolutionary search to find strong combinations against the current metagame. The strongest team builder is faced against the team metagame balance agent for its evaluation. Deep learning is also employed to predict the outcome of matches and recommend constructive elements of teams.
. This article is a continuation of the idea about developing a gaming prototype generator from natural language text. Game balance theme, which mentioned in previous authors' papers, fully opens up from theoretical and practice standpoints. Many expert opinions and definitions were analyzed to understand the identity of that problem. As a result, a full understanding of the problem was formed and challenges were listed, which exists in this direction. The purpose of the research is automation of game-designers routine in the prototyping stage. For this, the problem of generation of game balance from text documents is solved, so a number of scientific papers are considered, which offer algorithms, optimizing and automating approaches and computer games balance. The functionality of the dynamic balance editor Machinations is presented in detail, and the principle of operations is illustrated. To check the availability of approaches in the overall collection of the prototype generator work, a number of experiments were provided. They prove effective work with exhaustive diagrams and saves important development team resources. In addition, a particular problem of formalization and visualizing connection between gameplay and plot was solved, which is justified by the context dependence of the game balance. In conclusion, plans for the further development of a full-featured tool for the game scriptwriters and designers are given. As a conclusion, we state that automatic correction of the game balance is possible as well as its generation based on text.
The task of automating the routine work of computer game writers, narrative designers, set forth in earlier works, has been continued in the presented work. The issues of visualization of branching narrative structures of computer games are considered, the analysis of various approaches to visualization of the plot and other important components of a video game, such as, for example, automatic balancing of quantitative parameters, is carried out. The paper presents the chosen technological stack and gives specific solutions for storage in the form of a structured scenario, allowing the generation of continuing story branches and testing the narrative prototyping stage using the automatically generated text novel.
This study proposes a framework to automatically assess and bring balance in real-time strategy (RTS) games. A three-layered framework comprising intelligent bots, deep machine learning, explainable artificial intelligence (XAI), uncertainty quantification (UQ), and optimal learning is presented. For preliminary analysis, we conducted a study using the mi-croRTS game built specifically for advancing AI research. Data is generated through self play games between the intelligent bots, and game balance is measured through the predicted probability of each player winning a game. To demonstrate game re-balancing using this approach, a sample unbalanced game is shown along with proposed perturbations on important features identified using a popular XAI technique called SHapley Additive exPlanations (SHAP). Results indicate this framework enables efficient identification of game parameters causing imbalance and iterates over game parameters to restore balance. The three-layered framework is designed to be generic and applicable to more complicated RTS games, such as StarCraft II.
Real-world problems are often affected by uncertainties of different types and from multiple sources. Algorithms created for expensive optimisation, such as model-based optimisers, introduce additional errors. We argue that these uncertainties should be accounted for during the optimisation process. We thus introduce a benchmark as well as a new surrogate-assisted evolutionary algorithm to investigate this hypothesis further. The benchmark includes two function suites based on procedural content generation for games, which is a common problem observed in games research and also mirrors several types of uncertainties in the real-world. We find that observing and handling the uncertainty present in the problem can improve the optimiser, and also provides valuable insight into the function characteristics.
In procedural content generation, one is often interested in generating a large number of artifacts that are not only of high quality but also diverse, in terms of gameplay, visual impression or some other criterion. We investigate several search-based approaches to creating good and diverse game content, in particular approaches based on evolution strategies with or without diversity preservation mechanisms, novelty search and random search. The content domain is game levels, more precisely map sketches for strategy games, which are meant to be used as suggestions in the Sentient Sketchbook design tool. Several diversity metrics are possible for this type of content: we investigate tile-based, objective-based and visual impression distance. We find that evolution with diversity preservation mechanisms can produce both good and diverse content, but only when using appropriate distance measures. Reversely, we can draw conclusions about the suitability of these distance measures for the domain from the comparison of diversity preserving versus blind restart evolutionary algorithms.
We present Open Trumps, a version of the popular card game Top Trumps with decks that are procedurally generated based on open data. The game is played among multiple players through drawing cards and selecting the feature that is most likely to trump the same feature on the other players’ cards. Players can generate their own decks through choosing a suitable dataset and setting certain attributes; the generator then generates a balanced and playable deck using evolutionary computation. In the example dataset, each card represents a country and the features represent such entities as GDP per capita, mortality rate or tomato production, but in principle any dataset organised as instances with numerical features could be used. We also report the results of an evaluation intended to investigate both player experience and the hypothesis that players learn about the data underlying the deck they play with, since understanding the data is key to playing well. The results show that players enjoy playing the game, are enthusiastic about its potential and answer questions related to decks they have played significantly better than questions related to decks they have not played.
Dynamic balancing of game difficulty can help cater for different levels of ability in players. However, performance in some game tasks depends on not only the player's ability but also their desire to take risk. Taking or avoiding risk can offer players its own reward in a game situation. Furthermore, a game designer may want to adjust the mechanics differently for a risky, high ability player, as opposed to a risky, low ability player. In this work, we describe a novel modelling technique known as particle filtering which can be used to model various levels of player ability while also considering the player's risk profile. We demonstrate this technique by developing a game challenge where players are required to make a decision between a number of possible alternatives where only a single alternative is correct. Risky players respond faster but with more likelihood of failure. Cautious players wait longer for more evidence, increasing their likelihood of success, but at the expense of game time. By gathering empirical data for the player's response time and accuracy, we develop particle filter models. These models can then be used in real-time to categorise players into different ability and risk-taking levels.
This paper shows how multiobjective evolutionary algorithms can be used to procedurally generate complete and playable maps for real-time strategy (RTS) games. We devise heuristic objective functions that measure properties of maps that impact important aspects of gameplay experience. To show the generality of our approach, we design two different evolvable map representations, one for an imaginary generic strategy game based on heightmaps, and one for the classic RTS game StarCraft. The effect of combining tuples or triples of the objective functions are investigated in systematic experiments, in particular which of the objectives are partially conflicting. A selection of generated maps are visually evaluated by a population of skilled StarCraft players, confirming that most of our objectives correspond to perceived gameplay qualities. Our method could be used to completely automate in-game controlled map generation, enabling player-adaptive games, or as a design support tool for human designers.
We present the card game magic: the gathering as an interesting test bed for AI research. We believe that the complexity of the game offers new challenges in areas such as search in imperfect information domains and opponent modelling. Since there are a thousands of possible cards, and many cards change the rules to some extent, to successfully build AI for magic: the gathering ultimately requires a rather general form of game intelligence (although we only consider a small subset of these cards in this paper). We create a range of players based on stochastic, rule-based and Monte Carlo approaches and investigate Monte Carlo search with and without the use of a sophisticated rule-based approach to generate game rollouts. We also examine the effect of increasing numbers of Monte Carlo simulations on playing strength and investigate whether Monte Carlo simulations can enable an otherwise weak player to overcome a stronger rule-based player. Overall, we show that Monte Carlo search is a promising avenue for generating a strong AI player for magic: the gathering.
The hypervolume measure (or S metric) is a frequently applied quality measure for comparing the results of evolutionary multiobjective optimisation algorithms (EMOA). The new idea is to aim explicitly for the maximisation of the dominated hypervolume within the optimisation process. A steady-state EMOA is proposed that features a selection operator based on the hypervolume measure combined with the concept of non-dominated sorting. The algorithm’s population evolves to a well-distributed set of solutions, thereby focussing on interesting regions of the Pareto front. The performance of the devised Smetric selection EMOA (SMS-EMOA) is compared to state-of-the-art methods on two- and three-objective benchmark suites as well as on aeronautical real-world applications.
The attainment function has been proposed as a measure of the statistical performance of stochastic multiobjective optimisers
which encompasses both the quality of individual non-dominated solutions in objective space and their spread along the trade-off
surface. It has also been related to results from random closed-set theory, and cast as a mean-like, first-order moment measure
of the outcomes of multiobjective optimisers. In this work, the use of more informative, second-order moment measures for
the evaluation and comparison of multiobjective optimiser performance is explored experimentally, with emphasis on the interpretability
of the results.
Many new board games are designed each year, ranging from the unplayable to the truly exceptional. For each successful design there are untold numbers of failures; game design is something of an art. Players generally agree on some basic properties that indicate the quality and viability of a game, however these properties have remained subjective and open to interpretation. The aims of this thesis are to determine whether such quality criteria may be precisely defined and automatically measured through self-play in order to estimate the likelihood that a given game will be of interest to human players, and whether this information may be used to direct an automated search for new games of high quality. Combinatorial games provide an excellent test bed for this purpose as they are typically deep yet described by simple welldefined rule sets. To test these ideas, a game description language was devised to express such games and a general game system implemented to play, measure and explore them. Key features of the system include modules for measuring statistical aspects of self-play and synthesising new games through the evolution of existing rule sets. Experiments were conducted to determine whether automated game measurements correlate with rankings of games by human players, and whether such correlations could be used to inform the automated search for new high quality games. The results support both hypotheses and demonstrate the emergence of interesting new rule combinations.
Perfect information Monte Carlo (PIMC) search is the method of choice for constructing strong Al systems for trick-taking card games. PIMC search evaluates moves in imperfect information games by repeatedly sampling worlds based on state inference and estimating move values by solving the corresponding perfect information scenarios. PIMC search performs well in trick-taking card games despite the fact that it suffers from the strategy fusion problem, whereby the game's information set structure is ignored because moves are evaluated opportunistically in each world. In this paper we describe imperfect information Monte Carlo (IIMC) search, which aims at mitigating this problem by basing move evaluation on more realistic playout sequences rather than perfect information move values. We show that RecPIMC - a recursive IIMC search variant based on perfect information evaluation - performs considerably better than PIMC search in a large class of synthetic imperfect information games and the popular card game of Skat, for which PIMC search is the state-of-the-art cardplay algorithm.
In massively multiplayer online role-playing games (MMORPGs), each race holds some attributes and skills. Each skill contains several abilities such as physical damage and hit rate. All those attributes and abilities are functions of the character's level, which are called Ability-Increasing Functions (AIFs). A well-balanced MMORPG is characterized by having a set of well-balanced AIFs. In this paper, we propose a coevolutionary design method, including integration with the modified probabilistic incremental program evolution (PIPE) and the cooperative coevolutionary algorithm (CCEA), to solve the balance problem of MMORPGs. Moreover, we construct a simplest turn-based game model and perform a series of experiments based on it. The results indicate that the proposed method is able to obtain a set of well-balanced AIFs more efficiently, compared with the simple genetic algorithm (SGA), the simulated annealing algorithm (SAA) and the hybrid discrete particle swarm optimization (HDPSO) algorithm. The results also show that the performance of PIPE has been significantly improved through the modification works. (c) 2014 Elsevier B.V. All rights reserved.
Variations Forever is a novel game in which the player explores a vast design space of mini-games. In this paper, we present the procedural content generation research which makes the automatic generation of suitable game rulesets possible. Our generator, operating in the domain of code-like game content exploits answer-set programming as a means to declaratively represent a generative space as distinct from the domain-independent solvers which we use to enumerate it. Our generative spaces are powerfully sculptable using concise, declarative rules, allowing us to embed significant design knowledge into our ruleset generator as an important step towards a more serious automation of whole game design process.
Skat is Germany's national card game played by millions of players around the world. In this paper, we present the world's first computer skat player that plays at the level of human experts. This per- formance is achieved by improving state evalua- tions using game data produced by human players and by using these state evaluations to perform in- ference on the unobserved hands of opposing play- ers. Our results demonstrate the gains from adding inference to an imperfect information game player and show that training on data from average human players can result in expert-level playing strength.
Designing videogames involves weaving together systems of rules, called game mechanics, which support and str ucture com- pelling player experiences. Thus a significant port ion of game design involves reasoning about the effects of diff erent potential game mechanics on player experience. Unlike some design fields, such as architecture and mechanical design, that ha ve CAD tools to support designers in reasoning about and visuali zing designs, game designers have no tools for reasoning about an d visualizing systems of game mechanics. In this paper we perform a require- ments analysis for design-support tool for game des ign. We de- velop a proposal in two phases. First, we review th e design- support-system and game-design literatures to arriv e at a plausible system that helps designers reason about game mechanics and gameplay. We then refine these requirements in a st udy of three teams of game designers, investigating their curren t design prob- lems and gauging interest in our tool proposals and reactions to prototype tools. Our study finds that a game design assistant that is able to formally reason about abstract game mech anics would provide significant leverage to designers during mu ltiple stages of the design process.
In this paper, two approaches for estimating the generation in which a multi-objective evolutionary algorithm (MOEA) shows statistically significant signs of convergence are introduced. A set-based perspective is taken where convergence is measured by performance indicators. The proposed techniques fulfill the requirements of proper statistical assessment on the one hand and efficient optimisation for real-world problems on the other hand. The first approach accounts for the stochastic nature of the MOEA by repeating the optimisation runs for increasing generation numbers and analysing the performance indicators using statistical tools. This technique results in a very robust offline procedure. Moreover, an online convergence detection method is introduced as well. This method automatically stops the MOEA when either the variance of the performance indicators falls below a specified threshold or a stagnation of their overall trend is detected. Both methods are analysed and compared for two MOEA and on different classes of benchmark functions. It is shown that the methods successfully operate on all stated problems needing less function evaluations while preserving good approximation quality at the same time.
Discovering Unique Game Variants
- A Isaksen
- D Gopstein
- J Togelius
- A Nealen
A. Isaksen, D. Gopstein, J. Togelius, and
A. Nealen. Discovering Unique Game
Variants. In H. Toivonen et al., editors, Computational Creativity (ICCC 2015).
Sentient sketchbook: Computeraided game level authoring
- A Liapis
- G N Yannakakis
- J Togelius
A. Liapis, G. N. Yannakakis, and J. Togelius. Sentient sketchbook: Computeraided game level authoring. In Foundations of Digital Games (FDG 2013), pages
213-220. Society for the Advancement of
the Science of Digital Games, Santa Cruz,
CA, 2013.
Understanding Game Balance with Quantitative Methods
- A Jaffe
A. Jaffe. Understanding Game Balance with
Quantitative Methods. Phd thesis, University of Washington, 2013.
Zoran Zoran Popovic, Understanding game balance with quantitative methods
- Alexander Jaffe
- James R Lee
Quantifying over Play : Constraining Undesirable Solutions in Puzzle Design
- A M Smith
- E Butler
- Z Popović
A. M. Smith, E. Butler, and Z. Popović.
Quantifying over Play : Constraining Undesirable Solutions in Puzzle Design. In
Foundations of Digital Games (FDG 2013),
pages 221-228. Society for the Advancement of the Science of Digital Games,
Santa Cruz, CA, 2013.
Exploring the performance of stochastic multiobjective optimisers with the second-order attainment function
- C Fonseca
- V D Fonseca
- L Paquete
C. Fonseca, V. D. Fonseca, and L. Paquete. Exploring the performance of
stochastic multiobjective optimisers with
the second-order attainment function. In
C. A. C. Coello et al., editors, Evolutionary
Multi-Criterion Optimization (EMO 2005),
pages 250-264. Springer, Berlin, 2005. doi:
10.1007/b106458.