Matchmaking Strategies for Maximizing Player Engagement in Video Games
研究了竞争性视频游戏中如何通过优化匹配策略来最大化玩家参与度，考虑了技能差异、失败厌恶、付费赢和AI机器人等因素，发现付费赢在低技能玩家占多数时能提升参与度，且优化匹配可减少所需机器人数量。
Managing player engagement is vital to the online gaming industry, given that many games generate revenue through subscription models and microtransactions. We scrutinize engagement management in the prevalent category of competitive video games, where players are frequently matched against one another, and matchmaking systems substantially impact engagement. We propose a dynamic model to analyze player dynamics and optimize matchmaking policies for maximum engagement. Our model takes into account two essential factors in competitive games: heterogeneous skill levels and players’ aversion to losing. Additionally, the model enables us to consider pay-to-win strategies and AI-powered bots, which are contentious methods of influencing player engagement and endogenously affect the optimal matchmaking policy. To provide sharp insights, we analyze a specific case where there are two skill levels, and players churn only after experiencing a losing streak. The optimal matchmaking policy considers both short-term rewards by matching players myopically and long-term rewards by adjusting skill distribution. The pay-to-win system can positively impact player engagement when the majority of players are low-skilled, because adopting pay-to-win also affects skill distribution. This result challenges the conventional wisdom that typically regards pay-to-win as trading player experience for revenue. When incorporating AI-powered bots, we demonstrate that optimizing the matchmaking policy can significantly reduce the number of required bots. We then extend our model and conduct a case study with real data from an online chess platform. The optimal policy can improve engagement by 4%–6% or reduce the percentage of bots by 3% in comparison with skill-based matchmaking. This paper was accepted by Jeannette Song, operations management. Funding: A. N. Elmachtoub is partially supported by NSF [Grant CMMI-1944428]. X. Lei is partially supported by the Hong Kong Research Grants Council [Early Career Scheme 27503123]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2023.02957 .