On Sep 10, 2008, at 2:12 PM, Don Dailey wrote:
The rules are exactly the same for 9x9 as for 19x19. The boardsize is
different and that changes the game some.
I would suggest that if a top go player plays a game of chess
immediately after first learning the rules, he would lose very badly
to even mediocre players or even advanced beginners.
Usually i hesitate commenting just upon 1 statement out of an entire
story. Additionally you have the advantage of the native English skill;
kind of Obama type statement that you can still define an 'advanced
beginner'
as being a titled player who doesn't make money with it.
But we must correct you here in case you no longer see yourself as a
beginner
or as an advanced beginner.
Directly after learning the games of
chess, a strong go player will be able to win from you.
Strategically and tactically they're that much above your level that
they will
completely annihilate you.
Regarding that there are big differences between 9x9 and 19x19 i agree.
9x9 is a very simple game compared to 19x19.
From computer algorithms viewpoint seen that hard forward pruning is
tougher to
do there; explaining why the current random searching methods do so
bad in 9x9 and a lot better
at 19x19, relatively spoken. In absolute sense the will fail at both
games of course.
The only proof the current random searching methods give in go is
that searching deeply in random manner
is better than searching short lines in super-dubious manner. IMHO
this is logical.
The fact that there is zero fulltime salary prospects to get clever
guys interested in putting in years of fulltime effort into 9x9,
is the reason why the programs play this still so weak, as just a
single one of them would raise the game quality a lot.
It is of course a combination of evaluation and search.
What you typically see is that players with little domain knowldge in
chess nor go,
are busy with it now, doing a brute force attempt.
Note that computer-go has one big advantage over computer-chess;
because there is little sales possible to
achieve, there is little money at stake, that gives the advantage
that the programmers at least communicate
with each other in a forum like this and at tournaments. In
computerchess it is very difficult to find talkative persons.
The main progress that happens in computerchess is simply by
debugging other persons code.
It goes that far that a Russian author has even simply automatically
converted the program Rybka 1.0 back
from assembler into C code, it is called strelka.
This communicative skills problem is why progress in computerchess
goes relative slow. Still many brilliant guys
get lured into it, because chess gets played in 105+ nations. You see
clearly now that they do not do much effort
for it nowadays, as just like computer-go there is nearly no money to
make with an engine (in contradiction to GUI).
In doing that, it is selfexplaining that those who have a parallel go-
program,
still didn't figure out what computerchess already knew in the 80s,
namely that
to run parallel you need to avoid central locking, and need to search
with
a global hashtable (Feldmann, somewhere in the 80s)
and a decentralized form of doing the search. That sure involves
locking in case you want a good speedup, something Leierson&co with
CILK did not
manage nor Feldmann, but that is all very well solvable with a big
effort.
Yet those big guys who were busy with chess in the past years, who
already knew back in the 80s a lot
about decentralizing the parallel search, in computer-go they seem
absent.
Vincent
I really doubt this would be the case with 9x9 go. I don't think you
can really make a strong argument that 9x9 isn't go or that it's
not the
same game. You CAN argue that the characteristics of the game are
different and different aspects of the game are emphasized.
Some really strong players may not be specialists in 9x9 and may
lose to
players who specialize in 9x9 but are otherwise a few stones weaker at
19x19, but that's not remarkable. In chess you can also be weakened
significantly and be "thrown off your game" by a surprise opening - or
we could imagine a game where your opening is decided for you and it
would make you very uncomfortable.
My guess (and it's only a guess) is that strong players playing on the
9x9 board are simply very uncomfortable but probably do not play as
weak
as they imagine. In chess I heard that someone once did a study to
find out if playing speed chess weakened the performance of some
players
more than others and despite the fact that many players imagine
that it
does, it turned out that there was a remarkable correlation,
although no
doubt some players who specialize at different time controls would
have
an edge.
- Don
On Wed, 2008-09-10 at 11:27 +0900, Hideki Kato wrote:
Christoph Birk: <Pine.LNX.
[EMAIL PROTECTED]>:
On Tue, 9 Sep 2008, Olivier Teytaud wrote:
testbed for
parallelization because it's more difficult) and as "real"
targets (as there
are players
for both).
Sorry, but there are (almost) no players for 9x9. To repeat
D.Fotland's earlier comment: 9x9 is just for beginner's practice.
It's not go.
Other than the match CS vs. Kaori Aoba 4p, which Rémi reported
recently, there was a 9x9 match CS vs. Meien O 9p with no komi at
FIT2008. CS (B) won by 3 points.
I'd like to emphasize "9x9 is Go."
Hideki
Christoph
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
--
[EMAIL PROTECTED] (Kato)
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
