[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [computer-go] Modern brute force search
In whatever game we play.
The 361 stone go database will beat of course every player very easily,
just like a 32 stone database in chess will beat the strongest players very
easily.
The reality is of course that by the time we have a 32 stone database and
that there will be a 361 stone database for go, that players have looked
into it and already know the best line by head.
So they can win with white in chess and with black in go very easily when
using the optimum line.
In far endgames of chess we already see that the advantage of optimal play
is that you can really win based upon the utmost smallest margin in a game.
Kasparov by the way is not the best player in chess and definitely not in
endgames. Kasparov is the best the first 30 moves of the game.
After that i know another 150 who are better.
New training methods have last years been developed and are producing non
stop real strong youth players. Young players now are at playing levels
which in 1970 would have been number 3 in the world. I assume that is in go
also the case. Nevertheless a world champion which dominates all other top
players too, we haven't had yet.
Of course we had a winner of the FIDE knockout tournament a few years ago
(Ponomariov).
The exceptions proof the rule :)
At this moment Kasparov has managed, by playing nearly no games a year, to
still have the highest rating. Second is Anand.
A problem of the world top is that they have so much a higher rating than
the players below, that currently there is huge deflation in the 2600-2750
regions. So players there are a lot stronger now than players at the same
rating a few years ago.
The 'old generation' players need to do more and more effort against the
very strong new generations that are trying to get into the world top.
A result of this, is that it is very unclear to say who has what chance
against another within the rating system.
Where 750 points difference means at amateur level a near to 0% chance to
win, at the professional levels this is more like 250 rating points. This
doesn't tell the draw chance however. Not so long ago i also drew number
#46 of the world. He had at that moment 334 rating points more than i had.
However he played a few bad moves in opening. He did this on purpose to
avoid showing others his tournament preparings, which was the reason i
could draw him as i played a few good moves. for the same reason you must
take all human-computer games with a grain of salt. the top players aren't
showing their tournament preparings at all against them. they are 'too
valuable' in their eyes, though the computer games pay very well when
organized. Best example is kasparov there.
Additional i imagine go in west europe and america's has the same problem
like japanese/chinese/korean players have with chess.
That's that if you go play go in a pub it is unlikely you play players who
can explain to you a lot of positional things, and the number of go books
is very limited explaining this here. I have a book or 10 here and they all
suck. Explaining 2 or 3 things basically, the rest is all similar to the
other 9 books.
If you walk into a chess store here, there are 3000 books, and more
importantly in every town there is chess trainers.
In eastern europe it is even more drastically. Every young chess player has
his own trainer there.
That's of course the BIG difference in go. You make more moves which shows
the difference clearer between players and there is no such thing as 30
moves opening theory i would guess, and even if there were, then still 30
moves don't say much yet as there is another 100 to go.
So rating system will not work in go, as every game the difference is shown.
But you can be sure that worlds strongest player has a hard 0% chance
against a perfect computer in either game.
At 19:06 10-11-2004 -0500, Don Dailey wrote:
>
>The 95% number was completely arbitrary. A 95% percent score in chess
>might mean sccraping together 1 draw every 10 games. I don't really
>have a clue how good a world champion player is compared to an
>omnicient player, and I used 95% as an arbitrary starting figure. I
>actually believe the number to be more like 99% or higher.
>
>I do believe that the very best players might occasionally draw,
>probably more from dumb luck than anything (by managing not to make a
>blunder.) Please note that you can play random moves and have a
>statistical chance of drawing against perfect play. The odds would be
>astronomically against such an event occuring, but it is possible. It
>may be that the very best players can avoid playing a losing move on
>any given turn with a fairly high probability, they would just have to
>put a large number of these moves together in the same game.
>
>So assuming chess to be a draw, an omnicient player would never lose a
>game, but an incredibly good player might manage an occasional draw.
>
>I think at this level, style is important. There are only 3 game
>theoretic values in chess, win, loss or draw. An omnicient player can
>play ANY move that maintains the best theoretic value possible. If
>chess is a draw, it may be the case that most of the opening moves
>result in a drawn position. If an omnicient player always chooses
>it's moves randomly among equal moves (in the game theoretic sense), I
>believe a good human might draw relatively often, since many moves do
>not present any real challenges. But if an omnicient players "knew"
>something about making it tough on it's fallible human opponent, there
>might be very little chance of gatting that occasional draw.
>
>Good chess players know how this work. I am not a great chess player,
>but I once avoided an exchange or rooks, in what was almost certainly
>a drawn position against a much weaker opponent. The exchange of
>rooks would have made the draw simple, and I needed a win.
>
>In 8x8 checkers, the top players would often manage draws against an
>omnicient player, but the omnicient player would never lose. Chinook
>is probably quite close to omnicient with the big endgame databases.
>
>- Don
>
>
>
> X-Original-To: computer-go@xxxxxxxxxxxxxxxxx
> From: Chris Fant <chrisfant@xxxxxxxxxxxxxxxxx>
> Reply-To: Chris Fant <chrisfant@xxxxxxxxxxxxxxxxx>,
> computer-go <computer-go@xxxxxxxxxxxxxxxxx>
> Sender: computer-go-bounces@xxxxxxxxxxxxxxxxx
> X-Scanned-By: MIMEDefang 2.42
>
> > drd@xxxxxxxxxxxxxxxxx wrote:
> > > I don't know how to compare this gap with GO. If an omniscient chess
> > > player could win 95% of it's games against a Bobby Fischer, how would
> > > that compare to "number of stones" in Go?
>
> Where the heck did 95% come from? Wasn't the whole idea of an
> omniscient player one that knew perfect play but would try to confuse
> the opponent in hopes of something better than a draw? But at worst,
> it would draw. Are you saying that Bobby Fischer was capable of
> perfect play 5% of the time (leading to a draw)? Surely you are not.
> _______________________________________________
> computer-go mailing list
> computer-go@xxxxxxxxxxxxxxxxx
> http://www.computer-go.org/mailman/listinfo/computer-go/
>
>_______________________________________________
>computer-go mailing list
>computer-go@xxxxxxxxxxxxxxxxx
>http://www.computer-go.org/mailman/listinfo/computer-go/
>
>
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/