Quoting Jacques Basaldúa <[EMAIL PROTECTED]>:
Low variance is the clue for improvement.
Here I agree with you completely. I put it in my own words: As I see it MC-evaluation has a signal hidden by a lot of noise. This has some consequences. In order to evaluate a position correctly a lot of improvements have little to do with the actual shape of the stones in the position. For example everyone uses the rule to forbid filling in eyes to prevent every playout to have a arbitrary outcome when the maximumum number of moves has been reached. So first, tactically it is important that the playout does not make mistakes when it randomly invades and fill in liberties in the endgame. If such mistakes are avoided it reduces the noice in the opening and introduces a minimum of bias. Second, if we have a postion with stones played in strong shapes we want to conserve the strength of those shapes throughout the game. If you look at the patterns of the Mogo paper "Modifications of UTC with patterns...", these pattern defends the connectivity of stones and prevents stones from being surrounded. This reduces noise by making strong shapes more likely to survive the playouts, but this also introduces bias. Here I outlined two type of improvements. The first one should have an important impact on the endgame of playouts and the second one has an effect on the ability to play good strong shapes that are not overconcentrated by defending shapes correctly in the beginning of the playouts. But both types of improvements is certainly helpful at all stages of the playouts. -Magnus _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
