Simplifications to Guide Monte Carlo Tree Search in Combinatorial Games

TL;DR

This paper introduces a heuristic approach to improve Monte Carlo Tree Search in complex combinatorial games by analyzing simplified strategies and game versions to predict algorithm performance and guide strategy development.

Contribution

It proposes a novel method of using simplified game analyses to inform and enhance MCTS strategies in complex combinatorial settings.

Findings

Simplified strategies can predict MCTS performance in complex games.

Ensemble strategies derived from simplified analyses improve decision-making.

The approach aids strategy development when direct simulation is infeasible.

Abstract

We examine a type of modified Monte Carlo Tree Search (MCTS) for strategising in combinatorial games. The modifications are derived by analysing simplified strategies and simplified versions of the underlying game and then using the results to construct an ensemble-type strategy. We present some instances where relative algorithm performance can be predicted from the results in the simplifications, making the approach useful as a heuristic for developing strategies in highly complex games, especially when simulation-type strategies and comparative analyses are largely intractable.