Reply to thread

Message

<blockquote data-quote="Celebrim" data-source="post: 6355598" data-attributes="member: 4937">I'm not talking about approaching Go like Chess.You don't have to explore the entire solution space if you have a strategy.&nbsp;&nbsp; For example, we could imagine a connection game played on an asymmetrical board such that the player going first had a simple algorithm (trivial strategy) based on the fact that for any move, he could either advance 1 closer or 'match' the opponent's last move.&nbsp; Playing the game on a 1000 x 1001 board would create a massive number of potential solutions, but a simple rule based state machine could still play the game perfectly.&nbsp; Plenty of games seem difficult to play until you are given a simple set of rules for deciding how to play.It's obviously true that human Go players are not doing a breadth first search of all possible moves when they play a game.&nbsp; It's obviously true that in fact, the human players are analyzing fewer moves than the computers.&nbsp; In fact, if it is anything like Chess, it's going to turn out to be true that the idea that players are looking ahead X number of moves is as much of a myth as anything else.&nbsp; (Did you watch Kasparov's matches with Deep Blue?)&nbsp;&nbsp; My guess is that the best players in the world are probably only analyzing about 20-25 moves and probably not following much of them more than 5 moves or so deep.&nbsp;&nbsp; What they understand but have yet to be able to explain is the secret language of parity that governs the game and lets them pick the moves worth exploring.Human intelligence isn't magic.&nbsp; There is a strategy for decomposing the game, such that it transforms into a simpler problem.&nbsp; If there wasn't, humans couldn't play the game well.</blockquote>

[QUOTE="Celebrim, post: 6355598, member: 4937"] I'm not talking about approaching Go like Chess. You don't have to explore the entire solution space if you have a strategy. For example, we could imagine a connection game played on an asymmetrical board such that the player going first had a simple algorithm (trivial strategy) based on the fact that for any move, he could either advance 1 closer or 'match' the opponent's last move. Playing the game on a 1000 x 1001 board would create a massive number of potential solutions, but a simple rule based state machine could still play the game perfectly. Plenty of games seem difficult to play until you are given a simple set of rules for deciding how to play. It's obviously true that human Go players are not doing a breadth first search of all possible moves when they play a game. It's obviously true that in fact, the human players are analyzing fewer moves than the computers. In fact, if it is anything like Chess, it's going to turn out to be true that the idea that players are looking ahead X number of moves is as much of a myth as anything else. (Did you watch Kasparov's matches with Deep Blue?) My guess is that the best players in the world are probably only analyzing about 20-25 moves and probably not following much of them more than 5 moves or so deep. What they understand but have yet to be able to explain is the secret language of parity that governs the game and lets them pick the moves worth exploring. Human intelligence isn't magic. There is a strategy for decomposing the game, such that it transforms into a simpler problem. If there wasn't, humans couldn't play the game well. [/QUOTE]

Verification