And then they died - Using Action Sequences for Data Driven, Context Aware Gameplay Analysis

Aug 28, 2026

Uploaded by Margaret

Many successful games depend on data analytics to understand player behavior and guide design. Existing approaches often use machine learning and statistical analysis of aggregated logs, which extract what actions happen but lose context about when and how those actions occur. Qualitative methods can explain motivations and goals yet are difficult to scale. This paper introduces a mixed-method, human-in-the-loop framework combining Interactive Behavior Analytics and sequence analysis. It uses behavioral labels and visualizations to support context-sensitive, scalable, and generalizable gameplay analysis, demonstrated through teamwork behavior patterns in DotA 2.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Data-driven gameplay analysis and its limits

Context-aware qualitative understanding

Proposed mixed-method methodology

Combining Interactive Behavior Analytics and sequence analysis

Human-in-the-loop labels and visualizations

Case study: DotA 2 teamwork patterns

"And then they died": Using Action Sequences for Data Driven, Context Aware Gameplay Analysis arXiv :2006 . 10823v1 [ cs .HC] 18 Jun 2020 Erica Kleinman Northeastern University Boston, Massachusetts [kleinman.e@husky.neu.edu](kleinman.e@husky.neu.edu) Andy Bryant Harvard University Cambridge, Massachusetts [A@andymbryant.com](A@andymbryant.com) Sabbir Ahmad Northeastern University Boston, Massachusetts [ahmad.sab@husky.neu.edu](ahmad.sab@husky.neu.edu) Truong-Huy D. Nguyen Google New York, New York [truonghuy@gmail.com](truonghuy@gmail.com) Zhaoqing Teng Northeastern University Boston, Massachusetts [teng.z@husky.neu.edu](teng.z@husky.neu.edu) Casper Harteveld Northeastern University Boston, Massachusetts [c.harteveld@northeastern.edu](c.harteveld@northeastern.edu) Magy SeifEl-Nasr Northeastern University Boston, Massachusetts [magy@northeastern.edu](magy@northeastern.edu) ABSTRACT Many successful games rely heavily on data analytics to understand players and inform design. Popular methodologies focus on machine learning and statistical analysis of aggregated data. While effective in extracting information regarding player action, much of the context regarding when and how those actions occurred is lost. Qualitative methods allow researchers to examine context and derive meaningful explanations about the goals and motivations behind player behavior, but are difficult to scale. In this paper, we build on previous work by combining two existing methodologies: Interactive Behavior Analytics (IBA) [2] and sequence analysis (SA), in order to create a novel, mixed methods, human-in-the-loop data analysis methodology that uses behavioral labels and visualizations to allow analysts to examine player behavior in a way that is context sensitive, scalable, and generalizable. We present the methodology along with a case study demonstrating how it can be used to analyze behavioral patterns of teamwork in the popular multiplayer game Defense of the Ancients 2 (DotA 2) . CCS CONCEPTS • Human-centered computing → Visualization. KEYWORDS Game Design, Game Data, Player Modeling, Data Analysis, Mixed Methods, Human-in-the-Loop ACM Reference Format: Erica Kleinman, Sabbir Ahmad, Zhaoqing Teng, Andy Bryant, Truong-Huy D. Nguyen, Casper Harteveld, and Magy SeifEl-Nasr. 2018. "And then they Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions [from permissions@acm.org](from permissions@acm.org). FDG’20, September 15–18, 2020, Bugibba, Malta © 2018 Association for Computing Machinery. ACM ISBN 978-1-4503-XXXX-X/18/06. . . $15.00 [https://doi.org/10.1145/1122445.1122456](https://doi.org/10.1145/1122445.1122456) died": Using Action Sequences for Data Driven, Context Aware Gameplay Analysis. In The Fifteenth International Conference on the Foundations of Digital Games (FDG’20), September 15–18, 2020, Bugibba, Malta. ACM, New York, NY, USA, 12 pages. [https://doi.org/10.1145/1122445.1122456](https://doi.org/10.1145/1122445.1122456) 1 INTRODUCTION Games have grown exponentially in popularity and accessibility in recent years. This is especially true of esports, an industry worth close to a billion dollars [47]. These online games are frequently updated and adjusted to keep them engaging and relevant. But with so much revenue on the line, it is critical that the developers making these design adjustments are able to do so in an informed manner. In order to make these informed decisions, many have turned to data dr

FAQs

为什么仅靠现有数据分析方法难以理解玩家行为的背景？

现有方法多基于聚合数据与统计/机器学习，能够高效提取动作信息，但难以保留动作发生的时间与情境细节，因此难以分析“情境化”的策略或问题解决过程。

文中如何将上下文概念化为可分析的信息？

上下文被界定为游戏内与玩家选择相关的因素，可能包括时间与位置，也包括角色、物品以及环境特征等更细粒度的要素。

该论文提出的分析框架如何兼顾可扩展性与情境敏感性？

作者提出混合方法的人在回路（human-in-the-loop）流程，将互动行为分析与序列分析结合，并通过行为标签与可视化让分析人员以情境敏感且可扩展的方式检视玩家行为。

Aug 28, 2026

Uploaded by Margaret