Change is Hard: Consistent Player Behavior Across Games with Conflicting Incentives

Paper Title

Change is Hard: Consistent Player Behavior Across Games with Conflicting Incentives

Publication Info

• Topic area: Cross-game player behavior analysis in competitive gaming environments.
• Keywords: Player flexibility, specialization, competitive gaming, League of Legends, Teamfight Tactics, agency vs. structure, Structuration Theory, Self-Determination Theory, behavioral persistence, cross-platform design.

Background and Problem

• Problem / challenge: Existing research lacks empirical studies on whether user behavior is shaped more by system design constraints or by stable individual preferences, especially in cross-platform contexts.
• Significance: Understanding this dynamic is critical for designing systems that effectively influence user behavior, with applications in gaming, HCI, and cross-platform user modeling.
• Motivation and related work: Prior studies on player behavior focus on single games or platforms, often confounded by self-selection bias. This paper addresses this gap by analyzing the same players across two games with opposing structural incentives: League of Legends (specialization) and Teamfight Tactics (flexibility).

Solution

• Proposed approach: A cross-game behavioral analysis of 4,830 players who played at least 50 competitive games in both League of Legends and Teamfight Tactics, focusing on their flexibility and specialization tendencies.
• Novelty:
  1. Introduction of a cross-game analysis framework to reduce self-selection bias.
  2. Empirical testing of agency vs. structure dynamics using Giddens’ Structuration Theory.
  3. Operationalization of flexibility as a measurable behavioral trait across games.
  4. Insights into elite vs. non-elite player behavior and partial adaptation to structural incentives.
• Procedure and key techniques:
  1. Defined flexibility metrics for each game: play style diversity in League and composition diversity in TFT.
  2. Collected longitudinal data via Riot Games’ API, tracking players across both games.
  3. Tested hypotheses on flexibility, dispositional persistence, and partial adaptation using multivariate regression, kernel regression, and SHAP analysis.

Results

• Concrete findings:
  • League rewards specialization (negative correlation between flexibility and success), while TFT rewards flexibility (positive correlation).
  • Players exhibit consistent relative flexibility patterns across games, regardless of conflicting incentives.
  • Elite players show partial adaptation to game-specific incentives but retain baseline dispositional tendencies.
• Advantage over baselines:
  • Kernel regression with a Laplacian kernel outperformed other models, achieving R² values of 0.2235 (League flexibility) and 0.4551 (TFT flexibility).
  • Cross-game flexibility was a stronger predictor of behavior than competitive success in either game.
• Experiments / evaluation:
  • Dataset: 4,830 players with at least 50 games in each title, tracked over 14 months.
  • Metrics: Flexibility scores, competitive success (win rate), and engagement features.
  • Models: Multivariate regression, linear mixed-effects models, gradient boosting, kernel regression, and neural networks.
• Limitations and future work:
  • Limited to two games from Riot Games, potentially reducing generalizability.
  • Flexibility measured via end-game snapshots, missing in-game decision dynamics.
  • Dataset restricted to North American servers; cultural effects remain unexplored.
  • Oversampling of elite players may obscure variations within non-elite ranks.

Summary

This study investigates the interplay of agency and structure in shaping player behavior across two games with opposing incentives: League of Legends and Teamfight Tactics. Despite structural differences, players exhibit consistent behavioral patterns, with elite players showing partial adaptation to game-specific incentives. Using kernel regression and SHAP analysis, the study highlights the persistence of dispositional preferences and the limits of structural design in altering behavior. These findings have implications for game design, HCI, and cross-platform user modeling, emphasizing the need to balance structural incentives with individual agency. Future work should expand to diverse games, regions, and more granular in-game data.