Improving Strategic Decisions in Sequential Games by Exploiting Positional Similarity

We study the strategic similarity of game positions in two-player extensive games perfect information by looking at structure their local trees, with aim improving performance game-playing agents detecting forcing continuations. present a range measures over induced trees and compare them against benchmark problems chess, observing promising level accuracy matching up trap states. Our results can be applied to chess-like interactions where moves play role, such as those arising bargaining negotiation.