A Dynamical Systems Approach for Static Evaluation in Go

In the paper arguments are given why the concept of static evaluation has the potential to be a useful extension to Monte Carlo tree search. A new concept of modeling static evaluation through a dynamical system is introduced and strengths and weaknesses are discussed. The general suitability of this approach is demonstrated.

💡 Research Summary

**
The paper addresses a fundamental limitation of modern Go programs that rely almost exclusively on Monte‑Carlo Tree Search (MCTS). While MCTS has driven the recent surge in playing strength, its pure statistical nature makes it increasingly expensive as board size grows, and it does not exploit the intrinsic locality and influence structure that human players use. The author therefore proposes a static evaluation (SE) function that is built on a discrete dynamical system (DS).

Core Idea
Each empty intersection i on the board is assigned two real‑valued probabilities, w_i (probability the point ends up belonging to Black) and b_i (probability it ends up belonging to White). Each block j receives a single probability s_j that the block survives to the end of the game. All variables lie in