Reproduce “Seeding for Success: Skill and Stochasticity in Tabletop Games” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
James Goodman, Diego Perez-Liebana, Simon Lucas
This research from Queen Mary University of London provides a quantitative framework for analyzing the impact of randomness in tabletop games and its interaction with player skill. The work demonstrates that the effect of random seeds on game outcomes generally increases with agent skill and offers a method using "mirror games" to reduce experimental variance in AI agent comparisons.
Key insights
- The impact of game randomness on outcomes generally increases with player skill, as more skilled agents are better able to exploit favorable random situations or mitigate unfavorable ones.
- The influence of specific randomness sources varies significantly across games; for instance, initial board setup in Catan had a greater impact than dice rolls, while initial character deals were key in Colt Express.
- The presence of "non-monotonic seeds" indicates complex interactions where increasing agent skill does not always monotonically improve performance for specific random setups, pointing to potential strategic complexities or search limitations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.