Abstract
This note presents a simple counterexample to dispel the illusion that increasing the number of random playouts necessarily implies better estimates for simulation-based move planners. Incorporating a tree structure into the search helps redress the problem.
Get full access to this article
View all access options for this article.