The Go Transformer: Natural Language Modeling for Game Play

TL;DR

This paper demonstrates that language models like GPT-2 can effectively generate plausible and strategic Go game moves by learning from historical game data, enabling new approaches to game analysis and creation.

Contribution

It introduces a novel application of GPT-2 for modeling and generating Go strategies from game records, bridging NLP techniques with game strategy synthesis.

Findings

GPT-2 captures game sequencing and strategic formations.

Fine-tuned GPT-2 favors corner openings over center and side.

Language modeling can be applied to other board games with textual move data.

Abstract

This work applies natural language modeling to generate plausible strategic moves in the ancient game of Go. We train the Generative Pretrained Transformer (GPT-2) to mimic the style of Go champions as archived in Smart Game Format (SGF), which offers a text description of move sequences. The trained model further generates valid but previously unseen strategies for Go. Because GPT-2 preserves punctuation and spacing, the raw output of the text generator provides inputs to game visualization and creative patterns, such as the Sabaki project's game engine using auto-replays. Results demonstrate that language modeling can capture both the sequencing format of championship Go games and their strategic formations. Compared to random game boards, the GPT-2 fine-tuning shows efficient opening move sequences favoring corner play over less advantageous center and side play. Game generation as a language modeling task offers novel approaches to more than 40 other board games where historical text annotation provides training data (e.g., Amazons & Connect 4/6).