Re: [computer-go] Stages of the game

[matth1@xxxxxxxxxxxxxx]

Professionals certainly do not treat the 'rest of the game' as routine! Most professional games are won by very small margins, this means that they spend enormous effort in extracting as much as they can from each position. Some professionals, such as Ye Chang Ho, are great experts at the endgame and have such a level of skill that they can reverse the result of a game by exquisitely accurate endgame play. Some players play a much more aggressive style and their games often end in resignation, the critical point for these games, being dominated by complex fighting, usually comes in the mid to late middlegame. I suggest you take a look at some professional games by the young Kato Masao from the 1970s, or by Rui Nawei in recent years.

Matthew Holton

>  from:    Robin Kramer <robin@xxxxxxxxxxxxxxxxx>
>  date:    Fri, 02 Jan 2004 16:18:36
>  to:      computer-go@xxxxxxxxxxxxxxxxx
>  subject: Re: [computer-go] Stages of the game
> 
> Hi,
> 
> I know code talks most on this list, so I am hoping not to annoy anyone
> with my plans, by the way does anyone know of sgf code out there?
> 
> 1)I wonder at what point in the game would go be feasible to solve by a
> depth first search with alpha beta pruning.  If assuming that there are
> between 200 and 300 moves per game, while in the opening the search space
> is worst case 361!, with each move this search space becomes significantly
> smaller given live groups and territory.  Is it not possible to know how
> many nodes it is possible to expand per unit time, and how many playable
> moves there are left.  I have been reading a book on Fuseki(the equivalent
> of full board openings) the statement in the book, is that Fuseki is the
> only part of Go which is only place where there is continuos development,
> and that Professional players spend most of their time on the first fifty
> moves the rest of the game is routine.
> 
> 2)Second, I have been contemplating, is it better to have a decsion tree
> or neural net for each move of the game, or is it better to have a
> decision tree or neural net for the entire game.
> 
> Sincerely,
> 
> Robin Kramer
> 
> 
> _______________________________________________
> computer-go mailing list
> computer-go@xxxxxxxxxxxxxxxxx
> http://computer-go.org/mailman/listinfo/computer-go

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go