Learning opening books in partially observable games: using random seeds in Phantom Go

TL;DR

This paper investigates the impact of random seed variability in Phantom Go AI and introduces seed selection algorithms to improve opening strategies, significantly boosting win rates in various game sizes.

Contribution

It demonstrates the importance of seed choice in randomized AIs and presents methods for learning effective opening books in Phantom Go.

Findings

Winning rate increased from 50% to 70% in 5x5 against the same AI.

Winning rate improved from ~0% to 40% in multiple sizes against a stronger opponent.

Seed selection algorithms effectively enhance AI performance in partially observable games.

Abstract

Many artificial intelligences (AIs) are randomized. One can be lucky or unlucky with the random seed; we quantify this effect and show that, maybe contrarily to intuition, this is far from being negligible. Then, we apply two different existing algorithms for selecting good seeds and good probability distributions over seeds. This mainly leads to learning an opening book. We apply this to Phantom Go, which, as all phantom games, is hard for opening book learning. We improve the winning rate from 50% to 70% in 5x5 against the same AI, and from approximately 0% to 40% in 5x5, 7x7 and 9x9 against a stronger (learning) opponent.