# Biasing MCTS with Features for General Games

## TL;DR

This paper introduces a linear feature-based approach to bias Monte Carlo tree search in general games, offering a more interpretable and resource-efficient alternative to deep neural networks, with demonstrated strength improvements.

## Contribution

It presents a novel method of using interpretable local pattern features with linear approximation to enhance MCTS in various general games, emphasizing efficiency and generality.

## Key findings

- Significant playing strength improvements in multiple games.
- Effective feature learning during self-play training.
- Resource-efficient alternative to neural network-based biasing.

## Abstract

This paper proposes using a linear function approximator, rather than a deep neural network (DNN), to bias a Monte Carlo tree search (MCTS) player for general games. This is unlikely to match the potential raw playing strength of DNNs, but has advantages in terms of generality, interpretability and resources (time and hardware) required for training. Features describing local patterns are used as inputs. The features are formulated in such a way that they are easily interpretable and applicable to a wide range of general games, and might encode simple local strategies. We gradually create new features during the same self-play training process used to learn feature weights. We evaluate the playing strength of an MCTS player biased by learnt features against a standard upper confidence bounds for trees (UCT) player in multiple different board games, and demonstrate significantly improved playing strength in the majority of them after a small number of self-play training games.