|
RE: Watanabe's comments (2)
|
Manabu Terao
|
Mar 27, 2007 03:27 PST
|
Watanabe did not mention to concrete rating after the game.
On the other hand, they say Katsumata 5dan rated it 3 dan shoreikai during
the game. Furthermore, 3 crown Habu rated it 2 dan shoreikai in his article
in Nikkei Newspaper evening edition on March 26.
Manabu Terao
Mark Schreiber <mark_sc-@hotmail.com> wrote:
| |
Watanabe was surprised by some of the computer moves. Sometimes he
underestimated Bonanza. Sometimes he overestimated Bonanza. There should
be more computer vs pro games so we can get a better understanding how
well the computer plays.
Before the game Watanabe rated Bonanza about 2-3 kyu Shoreikai at 2
hour games. Did Watanabe rate Bonanza after this game?
Mark
Larry Kaufman wrote:
| |
----- Original Message -----
From: <jtis-@c2i.net>
To: <sho-@topica.com>
Sent: Thursday, March 22, 2007 2:23 AM
Subject: Re: Watanabe's comments (2)
Thanks very much for that, absolutely fascinating, and very open-minded.
Both the remark about computers finding strange moves that expand the
boundaries of the game, and the comment that their endgames are not yet
perfect are very interesting.
Looking back on how computer programs overtook the chess world it sounds
like they are closing the shogi gap remarkably quickly. It will be
interesting to see if there is a longer period where things are more
even
due to the differences in the games.
jt
I have a feeling that due to the differences in the games, the period
will be shorter, not longer. My point is that the draw margin in chess
made
it possible for top players and computers to be competitive for quite a
few
years. Once Shogi programs start to beat top pros in serious play (which
looks likely to happen in a year or two judging by this game), I think
it
will be only a couple more years before they can win pretty
consistently. We
may not be so far away from seeing shogi programs giving lance handicap
to
the Meijin or Ryu-Oh titleholders, just as we are now seeing chess
programs
giving pawn odds to grandmasters.
Larry Kaufman
----- Original Message ---------------
Subject: Re: Watanabe's comments (2)
From: Richard Sams <rj-@parkcity.ne.jp>
Date: Thu, 22 Mar 2007 16:00:15 +0900
To: sho-@topica.com
| |
Akira Watanabe is very impressive. After winning his game against
Bonanza under great pressure, I imagine he had to attend the inevitable
sponsor's party and answer a lot of questions from people not
necessarily familiar with shogi. According to his blog, he got home at
midnight. He must have been exhausted, but instead of going to bed he
started writing his report on the game for his readers. At around 2.15
a.m. I noticed that he had already posted the first part (Preparations
for the Game). Since it was quite short, I decided to translate it,
assuming he would write his report on the game after a good night's
sleep. But when I finished my translation at around 3.00 a.m. I noticed
that he had already posted his game report! I've just finished
translating that (see below). Now I notice that he's at it again,
answering questions from fans about the game. I'm afraid I can't keep
up with this pace .
Here's the URL for Watanabe's site:
http://blog.goo.ne.jp/kishi-akira/
Moves of the game:
Black: Bonanza
White: Akira Watanabe, Ryuo Titleholder
P-7f P-8d; P-6f P-3d; R-6h S-6b; K-4h P-5d; K-3h K-4b; K-2h K-3b; S-7b
G6a-5b; L-1h S-5c; K-1i B-3c; S-6c K-2b; S-2h G-3b; S-5f S-4d; P-4f
P-8e; B-7g G5b-4b; G-3i L-1b (Diagram); P-4e S-5c; S-4g K-1a; S-3f
P-2d; P-2f S-2b; G6i-5h S-2c; G5h-4h P-1d; S3f-2g G-2b; G4h-3h G4b-3b;
P-6e (Diagram) . Bx7g+; Nx7g G-4b; B*6f G4b-3b; B-7e B*4b; P-5f P-7d;
B-5g B-3c; B-4f R-8c; R-6g P-8f; Px8f Rx8f; Bx9a+ R-8i+; P-4d Sx4d;
R-4g +Rx9i (Diagram); P-6d S-3e; P-3f L*4f; R-3g Sx2f; Sx2f L-4i+;
Px6c+ +Lx3i; Gx3i P-2e; Sx2e P*2g; Rx2g P*2f; Rx2f G*1e (Diagram); P*2d
Gx2f; Px2c+ G3bx2c; P*2d P*2g; Px2c+ +Rx3i; +Px2b Bx2b; Sx3i G*2h; Sx2h
Px2h+; +Bx2h P*2g; +Bx2g Gx2g; S*3i S*3h; G*2h Gx2h; Sx2h P*2g Black
Resigns
From Akira Watanabe's Blog
March 22, 2007
Daiwa Securities Special Match Against Bonanza Part 2 - The Game
I got up at 8.00a.m. and left for the venue with an NHK employee who
had come to take photos of the event. We had the following conversation
in the car:
NHK photographer: "Bonanza has already been set up at the venue and
lost a game to Matsuo."
Me: "You're kidding! You mean that Matsuo?"
"Yes."
"Matsuo 6-dan lost to Bonanza?"
"Oh no, sorry, it wasn't Matsuo - it was Matsumoto-san (amateur 4 or 5
dan) who's in charge of the Internet live transmission."
"Ah, well that's not so surprising." (Here I am feeling really nervous
before the big game and he's trying to give me a heart attack!)
Matsumoto-san, who calls Bonanza "my friend," phoned me on the day
before the game: "Hello. Just to let you know that Bonanza is already
at the venue and can apparently analyze many more moves ahead than
before. (Sounding really pleased) Bonanza's rating is now 2800! Good
luck!"
Who is this guy supporting?
By the way, shogi ratings are provided on the Shogi Club 24 Internet
shogi dojo. The highest rating at the moment is 3084, so that means
Bonanza is only 300 points below the top. This is no joke - I'm
beginning to get scared.
We arrive at the venue. Judging from what people there say, it really
seems true that today Bonanza is functioning very well and fast too. I
was quite shocked. Returning to the back room, I try to relax but my
heart is beating faster and I'm already in slight panic mode. If I
think about the coming game I'll only have negative thoughts so I spend
the remaining time watching TV.
The game begins. As expected, Bonanza plays 4th file rook anaguma.
GAME DIAGRAM (after 30 . L-1b)
Up to now everything has gone as expected. I looked at various ways of
playing this position and came to the conclusion that this was the
safest. Here Bonanza usually plays P-6e S-5c; Bx3c+ G3bx3c; B*7a.
REFERENCE DIAGRAM
After . R-7b, Black has to exchange the bishop for the silver on 5c and
White already has a winning position. This is what I had been expecting
but today Bonanza didn't play this way. Afterwards I asked the
programmer, who said that Bonanza had looked at B*7a but stopped
thinking at that point!
Because it didn't play B*7a I felt "there's something different about
Bonanza today."
GAME DIAGRAM (after 47. P-6e)
P-6e seemed to be a surprise for many people, but having played Bonanza
many times, I was expecting it. This kind of move is generally
considered bad, but in practice it's not so easy to punish.
51. B*6f was a move I'd never seen before in such a position, but it
was actually quite good. It's the kind of move that shogi pros don't
consider, and you could even say that computers have contributed to
improving the technical level of shogi in this way. I think computers
will continue to show us new kinds of move like this and that this
could lead to the discovery of new ideas and tesuji.
GAME DAIGRAM (after 70 . *Rx9i)
This is a key moment. Here Bonanza played 71. P-6d. When I saw this
move, I said out loud "What's this (what's this weak move)?" P-6d cuts
off Black's promoted bishop on 9a and even if he makes a tokin with
Px6c+, it's a long way from my anaguma castle. Thinking "Thanks very
much!", I sank into deep thought.
Several minutes later I was thinking "Maybe P-6d is a good move after
all." If I play . Px6d, he takes my knight with *Bx8a and, compared to
taking the knight immediately in the diagram position, his attack is
one move faster because he can bring the promoted bishop back with
*Bx5d. So I can't take the pawn on 6d. If I can't take the pawn, I must
attack right away but I can't see any effective way of attacking. After
the move I played, . S-3e, I was expecting Px6c+, when I intended .
P-2e with a clear advantage. I had invited this knowing very well that
the computer program awards high points for making a tokin, but Bonanza
immediately played the best move, P-3f.
Bonanza had chosen P-6d based on the thinking "After P-6d the opponent
must attack. If so, the attack can be successfully repulsed." Human
players wouldn't consider P-6d because of their intuitive fear of
blocking the promoted bishop's diagonal.
Seeing these moves, I realized that I my opponent was much stronger
than the Bonanza I had encountered up to then.
GAME DIAGRAM (after 88 . G*1e)
This was the point where the outcome was decided. Just when I had
worked out that I would win after 89. P*2d, Bonanza actually played it.
If my opponent were human, I'd think "he's probably missed *Rx3i" but
since I trust Bonanza's endgame ability I was now thinking "*Rx3i ought
to win, but maybe there's an amazing reply ."
In the same position you think differently depending on whether the
opponent is human or a computer.
In fact, it seems that Bonanza underestimated . *Rx3i. Computers miss
things too. I thought the computer's endgame was perfect, so I felt a
little relieved!
After . G*5a, instead of P*2d, I had been afraid of the line L*2g Gx2f;
Lx2f P*2g; G*3h Px2h*; *Bx2h (Reference Diagram).
REFERENCE DIAGRAM (up to 95. *Bx2h)
To play this, the opponent has to think "I'm giving up a rook and my
silver will be taken by a pawn, but I'll be able to bring my promoted
bishop into the defense and my opponent has run out of pawns." I
figured that Bonanza wouldn't be capable of finding such a high-level
sequence.
So P*2d was the losing move. Not surprisingly, the intuitive "sense of
distance" from the king seems to have been hard for Bonanza.
Still, I'm glad I won. I'm really relieved too.
I had heard that computer programs will get stronger using
high-performance hardware, but I didn't expect it to be as strong as
this. With further advances in the hardware and improvements by the
developers, they are going to get even stronger. I thought they still
had quite a way to go, but now we have to recognize that they've
reached the point where they are getting to be a match for
professionals. I think we can no longer avoid matches between computers
and shogi professionals. And next time it will attract even greater
attention.
For the past month, I've been feeling quite a lot pressure as the big
day approached. Now at last I'm free from that pressure, so I'm going
to sleep very well .. zzzzz
|
|
|
|
|
 |
|
