Optimizing numerical systems and mechanism design is crucial for enhancing player experience in Massively Multiplayer Online (MMO) games. Traditional optimization approaches rely on large-scale online experiments or parameter tuning over predefined statistical models, which are costly, time-consuming, and may disrupt player experience. Although simplified offline simulation systems are often adopted as alternatives, their limited fidelity prevents agents from accurately mimicking real player reasoning and reactions to interventions. To address these limitations, we propose a generative agent-based MMO simulation system empowered by Large Language Models (LLMs). By applying Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) on large-scale real player behavioral data, we adapt LLMs from general priors to game-specific domains, enabling realistic and interpretable player decision-making. In parallel, a data-driven environment model trained on real gameplay logs reconstructs dynamic in-game systems. Experiments demonstrate strong consistency with real-world player behaviors and plausible causal responses under interventions, providing a reliable, interpretable, and cost-efficient framework for data-driven numerical design optimization.
Figures & tables
Figure 1. (a) Simulated MMO game world with multiple generative agents acting as players; (b) Simulation monitor front-end for macroscopic and microscopic monitoring of player agents. Simulated MMO game
Figure 2. An overview of our agent-based game simulation system, which comprises five major components: the Simulation Server , Game Services , Data Services , Experiment Manager , and Real Game Data , jointly support the end-to-end workflow of large-scale, data-driven game simulations.
Figure 3. Player Agent Fine-tuning Pipeline. The three steps fine-tuning enables the player agent to fully adapt to the game domain and make reasonable decision like real players.
Figure 4. Data distribution and accuracy.
Figure 5. (a) five main clusters of real players’ profile; (b) and (c) show that the average income (RMSE= 0.0077) and winning rate (RMSE = 0.0016) of N-th match on 5 respective player clusters respectively align well with real game data (with a fixed shift due to data privacy)