Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning to Play 7 Wonders Duel Without Human Supervision

Giovanni Paolini, Lorenzo Moreschini, Francesco Veneziano, Alessandro Iraci

TL;DR

This work introduces ZeusAI, an AlphaZero-inspired agent that learns to play 7 Wonders Duel through self-play using a Transformer encoder and Monte Carlo Tree Search, without human game data. It demonstrates near-top-human performance and offers insights into card and wonder strategies, while also proposing several modest rule variants to balance the game by reducing the first-player advantage. The study delivers a detailed analysis of training dynamics, strategic patterns, and balancing outcomes, highlighting the potential of AI to illuminate gameplay dynamics and inform community-driven balance changes. Overall, ZeusAI represents a significant step toward AI-assisted understanding and refinement of complex board games, with practical implications for game design and competitive play.

Abstract

This paper introduces ZeusAI, an artificial intelligence system developed to play the board game 7 Wonders Duel. Inspired by the AlphaZero reinforcement learning algorithm, ZeusAI relies on a combination of Monte Carlo Tree Search and a Transformer Neural Network to learn the game without human supervision. ZeusAI competes at the level of top human players, develops both known and novel strategies, and allows us to test rule variants to improve the game's balance. This work demonstrates how AI can help in understanding and enhancing board games.

Learning to Play 7 Wonders Duel Without Human Supervision

TL;DR

This work introduces ZeusAI, an AlphaZero-inspired agent that learns to play 7 Wonders Duel through self-play using a Transformer encoder and Monte Carlo Tree Search, without human game data. It demonstrates near-top-human performance and offers insights into card and wonder strategies, while also proposing several modest rule variants to balance the game by reducing the first-player advantage. The study delivers a detailed analysis of training dynamics, strategic patterns, and balancing outcomes, highlighting the potential of AI to illuminate gameplay dynamics and inform community-driven balance changes. Overall, ZeusAI represents a significant step toward AI-assisted understanding and refinement of complex board games, with practical implications for game design and competitive play.

Abstract

This paper introduces ZeusAI, an artificial intelligence system developed to play the board game 7 Wonders Duel. Inspired by the AlphaZero reinforcement learning algorithm, ZeusAI relies on a combination of Monte Carlo Tree Search and a Transformer Neural Network to learn the game without human supervision. ZeusAI competes at the level of top human players, develops both known and novel strategies, and allows us to test rule variants to improve the game's balance. This work demonstrates how AI can help in understanding and enhancing board games.
Paper Structure (12 sections, 6 figures, 1 table)

This paper contains 12 sections, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. The Game of 7 Wonders Duel
  3. ZeusAI
  4. The Model
  5. Action Selection
  6. Training
  7. Results
  8. Analysis of the Training Games
  9. Games Against Human Players
  10. Strategic Insights
  11. Balancing 7 Wonders Duel
  12. Conclusion and Future Work

Figures (6)

  • Figure 1: Structure of face-up cards (white) and face-down cards (colored) in ages I, II, and III (from left to right).
  • Figure 2: The 3 most frequently built cards by ZeusAI in ages I, II, III, and among the guilds. In every group, the most frequently built card is shown in front.
  • Figure 3: Distribution of victory types and winners (1 or 2 player) in self-play games throughout training.
  • Figure 4: Wonder preference in self-play games throughout training.
  • Figure 5: Wonder preference of ZeusAI (top) and human players (bottom).
  • ...and 1 more figures