Re: XieXie Master -- a recommendation

On Feb 27, 1:17 am, mikeh...@[EMAIL PROTECTED]
 wrote:
>
> this is the source of our disagreement.  u believe the draw window is
> sufficiently narrow enough in xiangqi such that the progress we've
> seen in xq programs will continue at the same pace for the next 20-yrs
> as we've seen in the previous 5.  what evidence do u have for this?

Progress in the last 25 years of computer xiangqi was because of
bigger opening book, bigger endgame database, smarter programs and
faster computers. There is no reason to believe these trends will not
continue for the next 20 years.

Look at the paper "computer Chinese chess". Find it at the World
Xiangqi Federation www.wxf.org/xq/in.htm. Click on Computer xiangqi.
Then click on Current state of computer Chinese chess (Xiang Qi).
There is a graph of computer progress from 1985 to 2003. Progress has
been steady even if we extend it 5 years to today 2008. The authors
also believe the computer will beat the world champion in 2010.

Progress may even speed up because of the internet and many games
played in automated tournaments between 2 programs. The draw nature of
the game comes into play when we reach perfect play. But we are very
far from perfect play.

> no, i'm not saying tinsley could actually *win* a game.  the draw
> window is so wide in checkers that tinsley could draw at will (w/ very
> rare losses on occassion).  so it's not the case at all that "tinsley
> could not even draw 1 game w/ chinook."  any checkers GM can draw
> (i.e. draw probability greater than say, 50%) a game w/ chinook,
> nevermind tinsley.

The draw window is wide only at perfect play and tinsey was not
perfect. The checkers programs and PCs of today are much better than
the chinook of 1996 that won the us champion****p by the largest margin
ever. So even today the best checkers PC programs would beat tinsey
every game nevermind today's chinook.

> furthermore, analysis using chinook of some of
> tinsley's games have shown that he in fact did play perfectly for
> several dozen moves in some of his games.  he was the closest thing to
> perfect that any player has ever achieved in any board game.  a large
> reason is because checkers is a much simpler game than the likes of
> chess, xiangqi, reversi, et al.

If tinsey played perfectly for several dozen moves and then made 1
mistake tinsey would lose the game. Which chinook analysis are you
referring to? Any analysis chinook did was with smaller endgame
databases. So it could not have known if tinsey made mistakes. I heard
Schaeffer will check tinsey's games when chinook completely solves
checkers. Only then can we know how many games tinsey would have lost
if he played today's chinook. My guess is tinsey was human so he would
have made at least 1 mistake each game and lost each game to today's
chinook.