Game tree search algorithms, such as Monte Carlo Tree Search (MCTS), require access to a forward model (or "simulator") of the game at hand. However, in some games such forward model is not readily available. In this paper we address the problem of automatically learning forward models (more specifically, combats models) for two-player attrition games. We report experiments comparing several approaches to learn such combat model from replay data to models generated by hand. We use StarCraft, a Real-Time Strategy (RTS) game, as our application domain. Specifically, we use a large collection of already collected replays, and focus on learning a combat model for tactical combats.

In this paper we describe a hierarchical method for procedurally generating 2D game maps using multi-dimensional Markov chains (MdMCs). Our method takes a collection of 2D game maps, breaks them into small chunks and performs clustering to find a set of chunks that correspond to high-level structures (high-level tiles) in the training maps. This set of high-level tiles is then used to re-represent the training maps, and to fit two sets of MdMC models: a high-level model captures the distribution of high-level tiles in the map, and a set of low-level models capture the internal structure of each high-level tile. These two sets of models can then be used to hierarchically generate new maps. We test our approach using two classic games, Super Mario Bros. and Loderunner, and compare the results against other existing map generators.

Behavioral game analytics has predominantly been confined to work on single games, which means that the cross-game applicability of current knowledge remains largely unknown. Here four experiments are presented focusing on the relationship between game ownership, time invested in playing games, and the players themselves, across more than 3000 games distributed by the Steam platform and over 6 million players, covering a total playtime of over 5 billion hours. Experiments are targeted at uncovering high-level patterns in the behavior of players focusing on playtime, using frequent itemset mining on game ownership, cluster analysis to develop playtime-dependent player profiles, correlation between user game rankings and, review scores, playtime and game ownership, as well as cluster analysis on Steam games. Within the context of playtime, the analyses presented provide unique insights into the behavior of game players as they occur across games, for example in how players distribute their time across games.

Coordination with an unknown human teammate is a notable challenge for cooperative agents. Behavior of human players in games with cooperating AI agents is often sub-optimal and inconsistent leading to choreographed and limited cooperative scenarios in games. This paper considers the difficulty of cooperating with a teammate whose goal and corresponding behavior change periodically. Previous work uses Bayesian models for updating beliefs about cooperating agents based on observations. We describe belief models for on-line planning, discuss tuning in the presence of noisy observations, and demonstrate empirically its effectiveness in coordinating with inconsistent agents in a simple domain. Further work in this area promises to lead to techniques for more interesting cooperative AI in games.

We introduce Playspecs, an application of omega-regular expressions to specifying play traces (sequences of game states or events unfolding over time). This connects the automated analysis and model checking of games to the literature on formal software verification via Bu ̈chi automata. We show how to define desirable or undesirable sequences of game events with Playspecs and how associated algorithms can find examples (or prove the impossibility) of such sequences. Playspecs have two main benefits over existing techniques for specifying the behaviors of a game over time. First, they offer a scalable commitment to formal modeling: the same Playspecs can filter existing traces gathered by telemetry, search for satisfying traces using existing game code, or drive formal verification when paired with a logical model of a game. Second, Playspecs' syntax can be customized for the game engine or game in question so designers may write specifications using their game's native vocabulary. We define Playspecs' syntax and semantics (modulo gamespecific customizations) and outline algorithms for each of the applications mentioned above, providing examples from the social simulation game Prom Week and the puzzle game engine PuzzleScript.

Clustering is an essential game analysis tool for understanding There are many clustering procedures that could be used player strengths and preferences. For example, clustering to group players based upon their play styles, with k-means techniques have been used to identify player preferences clustering being the most common method. Our use of for using vehicles over direct combat (Drachen et al. 2012), a model-based semi-parametric Bayesian clustering procedure for taking time to solve puzzles over running through content has two important advantages. First, the number of (Drachen, Canossa, and Yannakakis 2009), for understanding clusters (unique player styles) does not have to be prespecified.

We present Heapcraft: an open-source suite of tools for monitoring and improving collaboration in Minecraft. At the core of our system is a data collection and analysis framework for recording gameplay. We collected over 3451 player-hours of game behavior from 908 different players, and performed a general study of online collaboration. To make our game analytics easily accessible, we developed interactive information visualization tools and an analysis framework for players, administrators, and researchers to explore graphs, maps and timelines of live server activity. As part of our research, we introduce the collaboration index, a metric which allows server administrators and researchers to quantify, predict, and improve collaboration on Minecraft servers. Our analysis reveals several possible predictors of collaboration which can be used to improve collaboration on Minecraft servers. Heapcraft is designed to be general, and has the potential to be used for other shared online virtual worlds.

Flow is a psychological state that is reported to improve people’s performance. Flow can emerge when the person’s skills and the challenges of their activity match. This paper applies this concept to artificial intelligence agents. We equip a decision-making agent with a metacontrol policy that guides the agent to activities where the agent’s skills match the activity difficulty. Consequently, we expect the agent’s performance to improve. We implement and evaluate this approach in the role-playing game of Angband.

Displaying believable emotional reactions in virtual characters is required in applications ranging from virtual-reality trainers to video games. Manual scripting is the most frequently used method and enables an arbitrarily high fidelity of the emotions displayed. However, scripting is labour intense and greatly reduces the scope of emotions displayed and emotionally affected behavior in virtual characters. As a result, only a few virtual characters can display believable emotions and only in pre-scripted encounters. In this paper we implement and evaluate a lightweight algorithm for procedurally controlling both emotionally affected behavior and emotional appearance of a virtual character. The algorithm is based on two psychological models of emotions: conservation of resources and appraisal. The former component controls emotionally affected behavior of a virtual character whereas the latter generates explicit numeric descriptors of the character's emotions which can be used to drive the character's appearance. We implement the algorithm in a simple testbed and compare it to two baseline approaches via a user study. Human participants judged the emotions displayed by the algorithm to be more believable than those of the baselines.

This paper describes computational processes which can simulate how human designers sketch and then iteratively refine their work. The paper uses the concept of a map sketch as an initial, low-resolution and low-fidelity prototype of a game level, and suggests how such map sketches can be refined computationally. Different case studies with map sketches of different genres showcase how refinement can be achieved via increasing the resolution of the game level, increasing the fidelity of the function which evaluates it, or a combination of the two. While these case studies use genetic algorithms to automatically generate levels at different degrees of refinement, the general method described in this paper can be used with most procedural generation methods, as well as for AI-assisted design alongside a human creator.