Abstract
Simulation-based algorithms are particularly suited for high-uncertainty environments such as adversarial board games characterized by significant randomness and hidden information. Various Monte Carlo Tree Search (MCTS) variants are extensively used in these domains. In this paper, we propose a series of enhancements for Ensemble Determinization MCTS, introducing two axes for dynamic resource allocation.
First, Dynamic Number of Determinizations adjusts the count of currently utilized determinization trees based on the observed search behavior. Second, Dynamic Simulation Allocation unevenly distributes the simulation budget across determinization trees, leveraging simulation-to-simulation decisions to select the tree with the potential for the highest knowledge gain.
For benchmarking, we employed three popular tabletop games: Jaipur, Lost Cities, and Splendor. Testing our enhancements in iteration- and time-based settings revealed that specific configurations yield a statistically significant increase in the algorithm's performance.
Blogger's Review: This paper effectively enhances MCTS performance in complex environments through dynamic resource allocation strategies, showcasing the flexibility and adaptability of simulation algorithms in addressing uncertainty, warranting further exploration in various domains.