# Computer Go Musings/ Discussion

Sub-page of ComputerGoMusings

The page starts with:

"Computer chess works very simply. The computer looks at any of a dozen moves, and gives them a numerical rating. When the calculation time is done the move with the highest numerical rating gets played."

Sandy Harris Actually, that is utterly wrong. The program has a function that evaluates positions, not moves. It looks a certain distance ahead, evaluates the positions reached, then uses a minimax algorithm ( http://ai-depot.com/LogicGames/MiniMax.html) to choose a move.

The interesting question is how to evaluate positions. Obviously, things like territory, live groups, influence and connections come into it. But how do we quantify those? And what else needs evaluating?

Truc The original statement is correct. The minimax algorithm returns a number for each of the possible moves, and then plays accordingly. Simply put, it evaluates all of the followers recursively, and returns a number for each move.

Computer Go Musings/ Discussion last edited by Truc on July 28, 2005 - 07:24