Events

Prof. Zhikun Lu
Assistant Professor of Operations and Business Analytics
New York University Shanghai

The video game industry has experienced remarkable growth over the past few decades, with online gaming emerging as a key driver of its recent expansion. Fueled by an engagement-centric business model, game developers are increasingly relying on data-driven insights to optimize player engagement, particularly through incentive programs that reward players with virtual items. This paper explores the design of such reward systems, focusing on the trade-off between offering in-game currency (which provides player autonomy) and algorithmically assigned specific items (which represent system control). We examine this trade-off in the context of Honor of Kings, a highly popular mobile multiplayer online game developed by Tencent. Using real-time player data, we find that in-game currency outperforms specific virtual items in some cases, but not universally, highlighting the importance of aligning rewards with players’ understanding and experience. Our findings also reveal significant heterogeneity in player responses, suggesting that personalized reward strategies are crucial for maximizing engagement. To address this gap, we propose a data-driven framework for optimizing sequential treatment assignments in incentive programs under resource constraints, integrating dynamic causal machine learning with dynamic programming. It maximizes long-term user retention by dynamically adjusting reward allocations based on real-time player features. Our approach demonstrates substantial improvements in user retention and provides valuable insights that can extend beyond the gaming industry.The video game industry has experienced remarkable growth over the past few decades, with online gaming emerging as a key driver of its recent expansion. Fueled by an engagement-centric business model, game developers are increasingly relying on data-driven insights to optimize player engagement, particularly through incentive programs that reward players with virtual items. This paper explores the design of such reward systems, focusing on the trade-off between offering in-game currency (which provides player autonomy) and algorithmically assigned specific items (which represent system control). We examine this trade-off in the context of Honor of Kings, a highly popular mobile multiplayer online game developed by Tencent. Using real-time player data, we find that in-game currency outperforms specific virtual items in some cases, but not universally, highlighting the importance of aligning rewards with players’ understanding and experience. Our findings also reveal significant heterogeneity in player responses, suggesting that personalized reward strategies are crucial for maximizing engagement. To address this gap, we propose a data-driven framework for optimizing sequential treatment assignments in incentive programs under resource constraints, integrating dynamic causal machine learning with dynamic programming. It maximizes long-term user retention by dynamically adjusting reward allocations based on real-time player features. Our approach demonstrates substantial improvements in user retention and provides valuable insights that can extend beyond the gaming industry.