Computer beats amateur champion in 1 hour game

[Mark Schreiber]

Computer program Gekisashi beat Shimizugami. The time limit was 1 hour
with 1 minute byoyomi. Gekisashi was sente. Shimizugami Toru was the
Amateur Meijin. Gekisashi is world computer champion. The game was 149
moves. The game was played November 8, 2008 at the Game Programming
Workshop. The moves are at www.computer-shogi.org/kifu/gpw2008/vs_shimizugami.sjis.csa
. This is the first time an amateur champion was beaten by a computer
in a 1 hour game.

A second game was played at the Workshop. Kato beat Tanase Shogi in
121 moves. The moves are at www.computer-shogi.org/kifu/gpw2008/kato.kif
. I received this information from Takenobu Takizawa, Computer Shogi
Association Chairmen.

[Mark Schreiber]

There is more information. No amateur versus computer games at the
2009 annual world computer shogi championship (CSA WCSC 19).

I guess the computer is now too strong for short 30-minute games. And
the programmers are not interested in longer 2-hour games. Too bad, I
would like to see longer games at the championship. We need longer
games to get a computer rating. Then we would not have to speculate if
the computer reached professional level.

I heard there would be amateur versus computer games at the
Information Processing Society of Japan (IPSJ) annual conference in
March 2009.

[Mark Schreiber]

With this latest win can we guess how strong is Gekisashi? 2 years ago
the best program was about pro 2 dan. Now Gekisashi beat Shimizugami
in a 1 hour game. Is Shimizugami at about 6 dan pro level? Then is
Gekisashi at about 5 dan pro level?

> > Association Chairmen.- Hide quoted text -
>
> - Show quoted text -

[Gergely Buglyo]

Shimizugami is a top amateur, who has played numerous games against lower ranked professionals. Looking at his record, I would guess he's an even match for a C2 class professional. It's kind of hard to express in Dans, because a young 4 Dan professional may be stronger then an elderly 9 Dan, but C2 class consists of young professional players who just emerged from Shoreikai (holding a 4 Dan grade) and older ones who failed to get higher or fell back (though their Dan grade might be higher, they tend to lose more often to the rising "young titans").
 
However, it seems generally accepted that shogi software hasn't quite reached professional level yet. It's not quite clear to me what you meant by "pro 2 Dan" level - is it female pro (Joryu) level or Shoreikai 2 Dan level? I suppose that perhaps you meant Shoreikai, as programs now are much stronger than female pros. Some Japanese forums and blogs I've seen speculate that Tanase Shogi, Gekisashi and Bonanza are about Shoreikai 3 Dan level now, only one step away from professional level - with Tanase Shogi being the strongest of the three.
 

> Subject: Re: Computer beats amateur champion in 1 hour game
> From: msch...@gmail.com
> To: sho...@googlegroups.com

> More than mail–Windows Live™ goes way beyond your inbox. More than messages

[Mark Schreiber]

Hi Gergely Buglyo, Thank you for the information.
When I said pro 2 Dan I meant Shoreikai 2 Dan. So C2 class
professional is higher than Shoreikai 3 Dan level? Next is C1, B2, B1
and A is the highest level?

Why is it accepted software hasn't reached pro? I think we don’t know
because there are too few games against the computer to judge. We need
amateur champion versus computer matches at 2 hours.

Why do they say Tanase Shogi is better than Gekisashi? Gekisashi and
Tanase Shogi played 1 hour games with amateurs champions. Gekisashi
beat Shimizugami. Tanase Shogi lost to Kato. Gekisashi did better than
Tanase Shogi.

Is there an independent group running matches between programs and
rating programs? This would tell us which program is better.

On Jan 3, 12:29 pm, Gergely Buglyo <gbug...@hotmail.com> wrote:
> Shimizugami is a top amateur, who has played numerous games against lower ranked professionals. Looking at his record, I would guess he's an even match for a C2 class professional. It's kind of hard to express in Dans, because a young 4 Dan professional may be stronger then an elderly 9 Dan, but C2 class consists of young professional players who just emerged from Shoreikai (holding a 4 Dan grade) and older ones who failed to get higher or fell back (though their Dan grade might be higher, they tend to lose more often to the rising "young titans").
>
> However, it seems generally accepted that shogi software hasn't quite reached professional level yet. It's not quite clear to me what you meant by "pro 2 Dan" level - is it female pro (Joryu) level or Shoreikai 2 Dan level? I suppose that perhaps you meant Shoreikai, as programs now are much stronger than female pros. Some Japanese forums and blogs I've seen speculate that Tanase Shogi, Gekisashi and Bonanza are about Shoreikai 3 Dan level now, only one step away from professional level - with Tanase Shogi being the strongest of the three.
>

> > Subject: Re: Computer beats amateur champion in 1 hour game> From: mschr...@gmail.com> To: sho...@googlegroups.com> > > With this latest win can we guess how strong is Gekisashi? 2 years ago> the best program was about pro 2 dan. Now Gekisashi beat Shimizugami> in a 1 hour game. Is Shimizugami at about 6 dan pro level? Then is> Gekisashi at about 5 dan pro level?> > On Dec 27 2008, 9:15 pm, Mark Schreiber <mschr...@gmail.com> wrote:> > There is more information. No amateur versus computer games at the> > 2009 annual world computer shogi championship (CSA WCSC 19).> >> > I guess the computer is now too strong for short 30-minute games. And> > the programmers are not interested in longer 2-hour games. Too bad, I> > would like to see longer games at the championship. We need longer> > games to get a computer rating. Then we would not have to speculate if> > the computer reached professional level.> >> > I heard there would be amateur versus computer games at the> > Information Processing Society of Japan (IPSJ) annual conference in> > March 2009.> >> > On Dec 19, 3:44 pm, Mark Schreiber <mschr...@gmail.com> wrote:> >> >> >> > > Computer program Gekisashi beat Shimizugami. The time limit was 1 hour> > > with 1 minute byoyomi. Gekisashi was sente. Shimizugami Toru was the> > > Amateur Meijin. Gekisashi is world computer champion. The game was 149> > > moves. The game was played November 8, 2008 at the Game Programming> > > Workshop. The moves are atwww.computer-shogi.org/kifu/gpw2008/vs_shimizugami.sjis.csa> > > . This is the first time an amateur champion was beaten by a computer> > > in a 1 hour game.> >> > > A second game was played at the Workshop. Kato beat Tanase Shogi in> > > 121 moves. The moves are atwww.computer-shogi.org/kifu/gpw2008/kato.kif> > > . I received this information from Takenobu Takizawa, Computer Shogi> > > Association Chairmen.- Hide quoted text -> >> > - Show quoted text -> _________________________________________________________________
>
> More than messages–check out the rest of the Windows Live™.http://www.microsoft.com/windows/windowslive/

[Gergely Buglyo]

> Date: Mon, 5 Jan 2009 16:47:21 -0800

> Subject: Re: Computer beats amateur champion in 1 hour game

> From: msch...@gmail.com
> To: sho...@googlegroups.com

>
>
> Hi Gergely Buglyo, Thank you for the information.
> When I said pro 2 Dan I meant Shoreikai 2 Dan. So C2 class
> professional is higher than Shoreikai 3 Dan level? Next is C1, B2, B1
> and A is the highest level?
 

That's right, this is how professional players are classified.

> Why is it accepted software hasn't reached pro? I think we don’t know
> because there are too few games against the computer to judge. We need
> amateur champion versus computer matches at 2 hours.
 

You may be right about this, too. My personal guess is that if computers now are evenly matched with top amateurs in 1 hour games, then perhaps they could stand a chance against even a top professional in a Shogi Club 24-style short game (1 minute + 30 seconds byoyomi). I kind of remember a blog entry by Watanabe Ryuo a few years ago where he actually admitted having lost some short games against Bonanza, and shogi programs have become much stronger since then.
 
What needs to be understood, however, is that professional shogi players in Japan are looked upon in a very different way then chess professionals in the West. Their prestige is extremely high in the Japanese society, and they are considered the true masters of their art. Professionals losing games casually against programs would count as a severe prestige loss for the shogi world - imagine a renowned ikebana master whose work of art is outdone by a robot, and you'll have an idea of what kind of a reaction it would generate from the public. So, in my opinion, even if programs do reach professional level, this fact may not be generally accepted until programs beat top-ranking professionals in highly sponsored encounters with extensive media coverage like the Watanabe-Bonanza match in 2007. Such occasions will be rare, however, and the programs will have a harder time proving their strength against the real cream of the professional world than C2 class professionals.

> Why do they say Tanase Shogi is better than Gekisashi? Gekisashi and
> Tanase Shogi played 1 hour games with amateurs champions. Gekisashi
> beat Shimizugami. Tanase Shogi lost to Kato. Gekisashi did better than
> Tanase Shogi.
 

To be honest, I don't know. It is said that Tanase Shogi's style is more natural than Gekisashi's, because the latter program tends to make mistakes in the opening, for which it tries to make up later. However, I'm not in position to judge this, as I only own Tanase Shogi, and have never actually tried Gekisashi. I bought Tanase Shogi (sold under the name Shin Todai Shogi Muso) because I read it was stronger, and also because I prefer programs that play in a natural, human-like style.

> Is there an independent group running matches between programs and
> rating programs? This would tell us which program is better.
 

Some programs are playing on Shogi Club 24 and geting ratings from there. I don't know anything about a group that operates them, though.
 
 
I hope I could help somewhat in clearing up this subject, but I'm not an expert on computer shogi either. Maybe someone else can give better, more detailed answers regarding the programs. Best regards from Hungary:
 
Gergely

---

check out the rest of the Windows Live™. More than mail–Windows Live™ goes way beyond your inbox. More than messages

[Mark Schreiber]

Hi Gergely.
Short 1 minute games with a top pro are not serious. The top pro
will make mistakes. I would rather see the top pro play a Bishop
handicap game. But better than a handicap game would be an even game
with a c2 class pro.

I think the people who best understand what robots and computers
can do are the Japanese. There are more robots building things in
Japan than anywhere else in the world. The most sophisticated robots
are built in Japan. Look at the Honda Asimo robot. I think the
Japanese society know that computers will enter every area of life. It
is only the NSR who refuse to accept the inevitable.

In chess and every board game there were people who refused to let
computers play for different reasons. They said no computers because
computers do not really play the game or computers cheat. In shogi the
reason is prestige. I am not familiar with the ikebana art. But if
there is a big demand then someone will build a computer to do ikebana
art or any other task. The book Super Crunchers by Ian Ayres says the
computer does many things most people don’t know about or would never
have thought possible. The amazing thing is the computer does it
better than humans.

I think we could know when a computer passes the top pro without
playing the top pro. If the computer beats the amateur champion in a
match with an 8 piece handicap. At that point it will not be necessary
to play the top pro. The top pro would also lose. But this will take a
long time by 2030.

When you play Tanase Shogi do you use a handicap? At what handicap
can you beat Tanase Shogi?

I looked for computers at Shogi Club 24. I only found an old
version of YSS with a rating of 2744. Are there higher computer
ratings on Shogi Club 24? What are the time limits on Shogi Club 24
for these ratings? Are these 30 minute or 2 hour games?

I wonder do any pros use programs with handicaps for training?
Beating the computer with rook and left lance handicap maybe good
practice.
Mark


On Jan 6, 6:21 am, Gergely Buglyo <gbug...@hotmail.com> wrote:
> > Date: Mon, 5 Jan 2009 16:47:21 -0800> Subject: Re: Computer beats amateur champion in 1 hour game> From: mschr...@gmail.com> To: sho...@googlegroups.com> > > Hi Gergely Buglyo, Thank you for the information.> When I said pro 2 Dan I meant Shoreikai 2 Dan. So C2 class> professional is higher than Shoreikai 3 Dan level? Next is C1, B2, B1> and A is the highest level?

> _________________________________________________________________
> More than messages–check out the rest of the Windows Live™.http://www.microsoft.com/windows/windowslive/

[Gergely Buglyo]

> Date: Thu, 8 Jan 2009 19:35:27 -0800

> Subject: Re: Computer beats amateur champion in 1 hour game

> From: msch...@gmail.com
> To: sho...@googlegroups.com
>
>
> Hi Gergely.
 
Hi Mark, thank you for your reply.

> Short 1 minute games with a top pro are not serious. The top pro
> will make mistakes.
 

I absolutely agree - longer games have more value, they reflect the player's real strength.

 
> I would rather see the top pro play a Bishop
> handicap game. But better than a handicap game would be an even game
> with a c2 class pro.
>
> I think the people who best understand what robots and computers
> can do are the Japanese. There are more robots building things in
> Japan than anywhere else in the world. The most sophisticated robots
> are built in Japan. Look at the Honda Asimo robot. I think the
> Japanese society know that computers will enter every area of life. It
> is only the NSR who refuse to accept the inevitable.
>
> In chess and every board game there were people who refused to let
> computers play for different reasons. They said no computers because
> computers do not really play the game or computers cheat. In shogi the
> reason is prestige. I am not familiar with the ikebana art. But if
> there is a big demand then someone will build a computer to do ikebana
> art or any other task. The book Super Crunchers by Ian Ayres says the
> computer does many things most people don’t know about or would never
> have thought possible. The amazing thing is the computer does it
> better than humans.
 

The example of the ikebana master and the robot was only an analogy, of course I didn't mean it seriously. However, what needs to be understood is that people developing artificial intelligence and people who have mastered a traditional Japanese art almost belong to different worlds in Japan. Scientists working on robotics either do their research at prestigious universities like Todai or Waseda University, or work as engineers at a large electronic company or a powerful keiretsu. Most scientists haven't really experienced Japanese arts other than the form of school bukatsu (student clubs). On the other hand, masters of arts like ikebana, tea ceremony, calligraphy or Japanese musical instruments often spend some 40 years as a household apprentice (uchideshi) of an iemoto before they are really proclaimed a master. Even though shogi no longer uses the iemoto system, masters are still looked upon in a similar way. Of course, the day will come when they are surpassed by computers, but it will not be without some painful losses, actually. I also admire computers and their incredibly rapid development, but I have sympathy for the professional players who will fight and eventually fail in the name of human intellect. Have you read the novel Meijin by Yasunari Kawabata? It's about go, not shogi, but the concept is that an old master (symbolizing old Japan where artistic beauty was of utmost importance) is defeated by a youngster (who is standing as a symbol for a new, pragmatic way of thinking). At that time, after World War II, Japan has passed into a new age of modernism. Perhaps we're about to experience something similar now, when the human master is defeated by a computer, just like technology is taking the place of Japanese arts in everyday life as well. If that happens, only one question remains: will shogi itself, as an art or a game, survive at all? Already, the number of younger players is decreasing greatly - in most shogi clubs, some 90 percent of the players are over the age of 40. As computers and gaming consoles are spreading among young people, shogi as an activity is losing more and more territory. Perhaps the day will come when shogi is not played anymore, and fading out of everyday life, it becomes a matter of the past. I can only hope such a day will not come - in fact, that's one of the reasons why I'm doing my best to spread shogi in Hungary.

 
> I think we could know when a computer passes the top pro without
> playing the top pro. If the computer beats the amateur champion in a
> match with an 8 piece handicap. At that point it will not be necessary
> to play the top pro. The top pro would also lose. But this will take a
> long time by 2030.
 

Such a thing isn't going to happen, as 8 and even 6 piece handicaps are fairly simple because of the limited number of variations. So I think a 1 Dan amateur playing shita-te in a 6 piece handicap game could defeat a supercomputer that sees every possible move and position in shogi (like the "god of shogi"). However, from 4 piece handicap and on, shogi becomes much more difficult, so such a thing might actually happen in the distant future in 4 piece handicap games. However, there is also another reason why computers are bad at playing uwa-te (giving a handicap). Uwa-te often has to decide which move to make:
 
Move 1.: The "best" move. Because of the handicap, however, even the best move will result in a worse position eventually.
Move 2.: A "risky" move, inviting a mistake. If shita-te plays well and doesn't make the mistake, uwa-te will have a great disadvantage, and will have to resign soon - so by conventional judgement, this move is worse than the "best" move. However, even though this move is risky, uwa-te knows it will be difficult for shita-te to play well in this position. And if shita-te makes the mistake, the position will turn so now uwa-te has the advantage.
 
In such a situation, a computer chooses Move 1, because it is the best move, while humans prefer to play Move 2. This is the reason why humans do much better as uwa-te of handicap games then computers. Koji Tanigawa has written an excellent book (the title is "Taikyokukan ga Shobu wo Kimeru"), where he describes how one should choose the best move (a simple line) when the position is favorable, and a difficult line when one is at a disadvantage. However, I think this concept is beyond the level of the current state of artificial intelligence.

> When you play Tanase Shogi do you use a handicap? At what handicap
> can you beat Tanase Shogi?
 

No, I don't use a handicap. I once tried rook handicap, but it was too easy even for an amateur 2 Dan like me - programs strong in even games are still incredibly weak at giving handicaps, for the reasons I've described above. (And perhaps giving handicaps is not really the focus of software developers, either.) So I play only even games against it (mostly lower levels, as the top level is way too strong for me), and use the top level for analyzing my games.

> I looked for computers at Shogi Club 24. I only found an old
> version of YSS with a rating of 2744. Are there higher computer
> ratings on Shogi Club 24?
 

I don't know. I kind of remember reading on some Japanese forum some time ago that programs had surpassed 2800 points, but I have no idea which handle names the programs play under.

 
> What are the time limits on Shogi Club 24
> for these ratings? Are these 30 minute or 2 hour games?
 

I'm not sure, but I think they are from short games (1 minute + 30 seconds byoyomi). That's the only thing that could explain YSS being that high, as even I can beat the 2004 version of YSS in about 1 of 3 games. (For some reason, it seems to me YSS is doing better against computers than against humans.)

> I wonder do any pros use programs with handicaps for training?
> Beating the computer with rook and left lance handicap maybe good
> practice.
> Mark
 

Perhaps they do - computers are significantly better at receiving handicaps than giving them (in other words, they play better as shita-te).
 
Thank you again for your reply and the interesting questions you've raised. Best regards:
 
Gergely

[Mark Schreiber]

Hi Gergely.
The novel Meijin by Yasunari Kawabata sounds interesting. I will look
at it.
I think shogi will always be played. There will always be competition
to know who is the best human player. As there are world championships
still played in othello, draughts and chess. Computers can increase
the popularity by teaching shogi. Something similar to the Fritz
programs that teaches chess.

I think we can program the computer to play big handicaps. The
computer should try to create more complications if it sees it is
losing. The programmers will give higher weight to attacking pieces
and have more pieces under attack. Do this in more areas of the board
to give more opportunities for the human to pick the wrong move.

This is a short term solution. In the long run the computer will be so
much stronger that the human will need to play very precisely to win
even in a winning position. In western chess there are positions where
1 side has a forced win. But the win requires a precise sequence of
more than 200 moves. A human cannot beat the computer in this position
because the human cannot understand or memorize the wining moves.

If programs passed 2800 ratings on the Shogi Club 24 some time ago
then programs should be near 2850 today. They would rank within the
top 75 of the Shogi Club 24. The YSS program rated 2744 was an
experimental version running on a fast computer in 2007. I don’t think
you can compare your win to a 3 year older version on a slower
computer. Your version was weaker.
Mark


On Jan 9, 5:23 am, Gergely Buglyo <gbug...@hotmail.com> wrote:
> > Date: Thu, 8 Jan 2009 19:35:27 -0800> Subject: Re: Computer beats amateur champion in 1 hour game> From: mschr...@gmail.com> To: sho...@googlegroups.com> > > Hi Gergely.

>
> Hi Mark, thank you for your reply.
>
> > Short 1 minute games with a top pro are not serious. The top pro> will make mistakes.
>
> I absolutely agree - longer games have more value, they reflect the player's real strength.
>
> > I would rather see the top pro play a Bishop> handicap game. But better than a handicap game would be an even game> with a c2 class pro.> > I think the people who best understand what robots and computers> can do are the Japanese. There are more robots building things in> Japan than anywhere else in the world. The most sophisticated robots> are built in Japan. Look at the Honda Asimo robot. I think the> Japanese society know that computers will enter every area of life. It> is only the NSR who refuse to accept the inevitable.> > In chess and every board game there were people who refused to let> computers play for different reasons. They said no computers because> computers do not really play the game or computers cheat. In shogi the> reason is prestige. I am not familiar with the ikebana art. But if> there is a big demand then someone will build a computer to do ikebana> art or any other task. The book Super Crunchers by Ian Ayres says the> computer does many things most people don’t know about or would never> have thought possible. The amazing thing is the computer does it> better than humans.
>
> The example of the ikebana master and the robot was only an analogy, of course I didn't mean it seriously. However, what needs to be understood is that people developing artificial intelligence and people who have mastered a traditional Japanese art almost belong to different worlds in Japan. Scientists working on robotics either do their research at prestigious universities like Todai or Waseda University, or work as engineers at a large electronic company or a powerful keiretsu. Most scientists haven't really experienced Japanese arts other than the form of school bukatsu (student clubs). On the other hand, masters of arts like ikebana, tea ceremony, calligraphy or Japanese musical instruments often spend some 40 years as a household apprentice (uchideshi) of an iemoto before they are really proclaimed a master. Even though shogi no longer uses the iemoto system, masters are still looked upon in a similar way. Of course, the day will come when they are surpassed by computers, but it will not be without some painful losses, actually. I also admire computers and their incredibly rapid development, but I have sympathy for the professional players who will fight and eventually fail in the name of human intellect. Have you read the novel Meijin by Yasunari Kawabata? It's about go, not shogi, but the concept is that an old master (symbolizing old Japan where artistic beauty was of utmost importance) is defeated by a youngster (who is standing as a symbol for a new, pragmatic way of thinking). At that time, after World War II, Japan has passed into a new age of modernism. Perhaps we're about to experience something similar now, when the human master is defeated by a computer, just like technology is taking the place of Japanese arts in everyday life as well. If that happens, only one question remains: will shogi itself, as an art or a game, survive at all? Already, the number of younger players is decreasing greatly - in most shogi clubs, some 90 percent of the players are over the age of 40. As computers and gaming consoles are spreading among young people, shogi as an activity is losing more and more territory. Perhaps the day will come when shogi is not played anymore, and fading out of everyday life, it becomes a matter of the past. I can only hope such a day will not come - in fact, that's one of the reasons why I'm doing my best to spread shogi in Hungary. > I think we could know when a computer passes the top pro without> playing the top pro. If the computer beats the amateur champion in a> match with an 8 piece handicap. At that point it will not be necessary> to play the top pro. The top pro would also lose. But this will take a> long time by 2030.
>
> Such a thing isn't going to happen, as 8 and even 6 piece handicaps are fairly simple because of the limited number of variations. So I think a 1 Dan amateur playing shita-te in a 6 piece handicap game could defeat a supercomputer that sees every possible move and position in shogi (like the "god of shogi"). However, from 4 piece handicap and on, shogi becomes much more difficult, so such a thing might actually happen in the distant future in 4 piece handicap games. However, there is also another reason why computers are bad at playing uwa-te (giving a handicap). Uwa-te often has to decide which move to make:
>
> Move 1.: The "best" move. Because of the handicap, however, even the best move will result in a worse position eventually.
> Move 2.: A "risky" move, inviting a mistake. If shita-te plays well and doesn't make the mistake, uwa-te will have a great disadvantage, and will have to resign soon - so by conventional judgement, this move is worse than the "best" move. However, even though this move is risky, uwa-te knows it will be difficult for shita-te to play well in this position. And if shita-te makes the mistake, the position will turn so now uwa-te has the advantage.
>
> In such a situation, a computer chooses Move 1, because it is the best move, while humans prefer to play Move 2. This is the reason why humans do much better as uwa-te of handicap games then computers. Koji Tanigawa has written an excellent book (the title is "Taikyokukan ga Shobu wo Kimeru"), where he describes how one should choose the best move (a simple line) when the position is favorable, and a difficult line when one is at a disadvantage. However, I think this concept is beyond the level of the current state of artificial intelligence.
>
> > When you play Tanase Shogi do you use a handicap? At what handicap> can you beat Tanase Shogi?
>
> No, I don't use a handicap. I once tried rook handicap, but it was too easy even for an amateur 2 Dan like me - programs strong in even games are still incredibly weak at giving handicaps, for the reasons I've described above. (And perhaps giving handicaps is not really the focus of software developers, either.) So I play only even games against it (mostly lower levels, as the top level is way too strong for me), and use the top level for analyzing my games.
>
> > I looked for computers at Shogi Club 24. I only found an old> version of YSS with a rating of 2744. Are there higher computer> ratings on Shogi Club 24?
>
> I don't know. I kind of remember reading on some Japanese forum some time ago that programs had surpassed 2800 points, but I have no idea which handle names the programs play under.
>
> > What are the time limits on Shogi Club 24> for these ratings? Are these 30 minute or 2 hour games?
>
> I'm not sure, but I think they are from short games (1 minute + 30 seconds byoyomi). That's the only thing that could explain YSS being that high, as even I can beat the 2004 version of YSS in about 1 of 3 games. (For some reason, it seems to me YSS is doing better against computers than against humans.)
>
> > I wonder do any pros use programs with handicaps for training?> Beating the computer with rook and left lance handicap maybe good> practice.> Mark
>
> Perhaps they do - computers are significantly better at receiving handicaps than giving them (in other words, they play better as shita-te).
>
> Thank you again for your reply and the interesting questions you've raised. Best regards:
>
> Gergely
>
>
>
> > > > On Jan 6, 6:21 am, Gergely Buglyo <gbug...@hotmail.com> wrote:> > > Date: Mon, 5 Jan 2009 16:47:21 -0800> Subject: Re: Computer beats amateur champion in 1 hour game> From: mschr...@gmail.com> To: sho...@googlegroups.com> > > Hi Gergely Buglyo, Thank you for the information.> When I said pro 2 Dan I meant Shoreikai 2 Dan. So C2 class> professional is higher than Shoreikai 3 Dan level? Next is C1, B2, B1> and A is the highest level?> >> > That's right, this is how professional players are classified.> >> > > Why is it accepted software hasn't reached pro? I think we don’t know> because there are too few games against the computer to judge. We need> amateur champion versus computer matches at 2 hours.> >> > You may be right about this, too. My personal guess is that if computers now are evenly matched with top amateurs in 1 hour games, then perhaps they could stand a chance against even a top professional in a Shogi Club 24-style short game (1 minute + 30 seconds byoyomi). I kind of remember a blog entry by Watanabe Ryuo a few years ago where he actually admitted having lost some short games against Bonanza, and shogi programs have become much stronger since then.> >> > What needs to be understood, however, is that professional shogi players in Japan are looked upon in a very different way then chess professionals in the West. Their prestige is extremely high in the Japanese society, and they are considered the true masters of their art. Professionals losing games casually against programs would count as a severe prestige loss for the shogi world - imagine a renowned ikebana master whose work of art is outdone by a robot, and you'll have an idea of what kind of a reaction it would generate from the public. So, in my opinion, even if programs do reach professional level, this fact may not be generally accepted until programs beat top-ranking professionals in highly sponsored encounters with extensive media coverage like the Watanabe-Bonanza match in 2007. Such occasions will be rare, however, and the programs will have a harder time proving their strength against the real cream of the professional world than C2 class professionals.> Why do they say Tanase Shogi is better than Gekisashi? Gekisashi and> Tanase Shogi played 1 hour games with amateurs
>

> ...
>
> read more »- Hide quoted text -

[Gergely Buglyo]

> Date: Mon, 12 Jan 2009 11:22:43 -0800

> Subject: Re: Computer beats amateur champion in 1 hour game

> From: msch...@gmail.com
> To: sho...@googlegroups.com

>
>
> Hi Gergely.
> The novel Meijin by Yasunari Kawabata sounds interesting. I will look
> at it.
> I think shogi will always be played. There will always be competition
> to know who is the best human player. As there are world championships
> still played in othello, draughts and chess. Computers can increase
> the popularity by teaching shogi. Something similar to the Fritz
> programs that teaches chess.
 

Hi Mark,
 
I wish I could be so optimistic about it, too. But I think when people choose a hobby, a game to learn at a young age, they tend to choose the one that seems most appealing and interesting to them. So it's virtually like a competition, a rivalry between games, in which it is decided which game has more to offer than the others. This is actually how shogi as we know it evolved out of many other variants that just proved less interesting in the long run, so nobody really plays them anymore.
 
In past centuries in Japan, shogi's main rival was the game of go. In the rivalry shogi vs go, shogi was able to hold its own because it offered a very interesting alternative to go, with similar depth but different kinds of pieces that could be captured and reused, etc. However, in the 21st century, the rivalry will look something like: shogi vs more and more advanced Playstation consoles? Shogi vs online RPGs? If we're talking about the actual depth of the game, none of these rivals can even be compared to shogi, but on the surface, they might seem a lot more appealing. Looking at the kinds of hobbies young people pursue in Japan, it already seems shogi is losing to these new rivals.

> I think we can program the computer to play big handicaps. The
> computer should try to create more complications if it sees it is
> losing. The programmers will give higher weight to attacking pieces
> and have more pieces under attack. Do this in more areas of the board
> to give more opportunities for the human to pick the wrong move.
 

Yeah, I think you're right. I don't know too much about programming artificial intelligence, but I also tend to think it could be done.

> This is a short term solution. In the long run the computer will be so
> much stronger that the human will need to play very precisely to win
> even in a winning position. In western chess there are positions where
> 1 side has a forced win. But the win requires a precise sequence of
> more than 200 moves. A human cannot beat the computer in this position
> because the human cannot understand or memorize the wining moves.
>
> If programs passed 2800 ratings on the Shogi Club 24 some time ago
> then programs should be near 2850 today. They would rank within the
> top 75 of the Shogi Club 24. The YSS program rated 2744 was an
> experimental version running on a fast computer in 2007. I don’t think
> you can compare your win to a 3 year older version on a slower
> computer. Your version was weaker.
> Mark
 

It indeed seems so. Maybe YSS has improved a lot during those 3 years.

 
Best regards from Hungary:
 
Gergely

[Mark Schreiber]

I think consoles or RPG games and shogi do not compete. The people
who play these video games would more likely watch television and
movies if they did not have video games. Shogi has competition from
western chess. Western chess has a larger worldwide community. It is
easier to find a partner for western chess than for shogi. People who
would play Shogi may play western chess instead.

On Jan 13, 4:15 am, Gergely Buglyo <gbug...@hotmail.com> wrote:
> > Date: Mon, 12 Jan 2009 11:22:43 -0800> Subject: Re: Computer beats amateur champion in 1 hour game> From: mschr...@gmail.com> To: sho...@googlegroups.com> > > Hi Gergely.> The novel Meijin by Yasunari Kawabata sounds interesting. I will look> at it.> I think shogi will always be played. There will always be competition> to know who is the best human player. As there are world championships> still played in othello, draughts and chess. Computers can increase> the popularity by teaching shogi. Something similar to the Fritz> programs that teaches chess.

>
> Hi Mark,
>
> I wish I could be so optimistic about it, too. But I think when people choose a hobby, a game to learn at a young age, they tend to choose the one that seems most appealing and interesting to them. So it's virtually like a competition, a rivalry between games, in which it is decided which game has more to offer than the others. This is actually how shogi as we know it evolved out of many other variants that just proved less interesting in the long run, so nobody really plays them anymore.
>
> In past centuries in Japan, shogi's main rival was the game of go. In the rivalry shogi vs go, shogi was able to hold its own because it offered a very interesting alternative to go, with similar depth but different kinds of pieces that could be captured and reused, etc. However, in the 21st century, the rivalry will look something like: shogi vs more and more advanced Playstation consoles? Shogi vs online RPGs? If we're talking about the actual depth of the game, none of these rivals can even be compared to shogi, but on the surface, they might seem a lot more appealing. Looking at the kinds of hobbies young people pursue in Japan, it already seems shogi is losing to these new rivals.
>
> > I think we can program the computer to play big handicaps. The> computer should try to create more complications if it sees it is> losing. The programmers will give higher weight to attacking pieces> and have more pieces under attack. Do this in more areas of the board> to give more opportunities for the human to pick the wrong move.
>
> Yeah, I think you're right. I don't know too much about programming artificial intelligence, but I also tend to think it could be done.
>
> > This is a short term solution. In the long run the computer will be so> much stronger that the human will need to play very precisely to win> even in a winning position. In western chess there are positions where> 1 side has a forced win. But the win requires a precise sequence of> more than 200 moves. A human cannot beat the computer in this position> because the human cannot understand or memorize the wining moves.> > If programs passed 2800 ratings on the Shogi Club 24 some time ago> then programs should be near 2850 today. They would rank within the> top 75 of the Shogi Club 24. The YSS program rated 2744 was an> experimental version running on a fast computer in 2007. I don’t think> you can compare your win to a 3 year older version on a slower> computer. Your version was weaker.> Mark
>
> It indeed seems so. Maybe YSS has improved a lot during those 3 years.
>
> Best regards from Hungary:
>
> Gergely
>
>
>
> > On Jan 9, 5:23 am, Gergely Buglyo <gbug...@hotmail.com> wrote:> > > Date: Thu, 8 Jan 2009 19:35:27 -0800> Subject: Re: Computer beats amateur champion in 1 hour game> From: mschr...@gmail.com> To: sho...@googlegroups.com> > > Hi Gergely.> >> > Hi Mark, thank you for your reply.> >> > > Short 1 minute games with a top pro are not serious. The top pro> will make mistakes.> >> > I absolutely agree - longer games have more value, they reflect the player's real strength.> >> > > I would rather see the top pro play a Bishop> handicap game. But better than a handicap game would be an even game> with a c2 class pro.> > I think the people who best understand what robots and computers> can do are the Japanese. There are more robots building things in> Japan than anywhere else in the world. The most sophisticated robots> are built in Japan. Look at the Honda Asimo robot. I think the> Japanese society know that computers will enter every area of life. It> is only the NSR who refuse to accept the inevitable.> > In chess and every board game there were people who refused to let> computers play for different reasons. They said no computers because> computers do not really play the game or computers cheat. In shogi the> reason is prestige. I am not familiar with the ikebana art. But if> there is a big demand then someone will build a computer to do ikebana> art or any other task. The book Super Crunchers by Ian Ayres says the> computer does many things most people don’t know about or would never> have thought possible. The amazing thing is the computer does it> better than humans.> >> > The example of the ikebana master and the robot was only an analogy, of course I didn't mean it seriously. However, what needs to be understood is that people developing artificial intelligence and people who have mastered a traditional Japanese art almost belong to different worlds in Japan. Scientists working on robotics either do their research at prestigious universities like Todai or Waseda University, or work as engineers at a large electronic company or a powerful keiretsu. Most scientists haven't really experienced Japanese arts other than the form of school bukatsu (student clubs). On the other hand, masters of arts like ikebana, tea ceremony, calligraphy or Japanese musical instruments often spend some 40 years as a household apprentice (uchideshi) of an iemoto before they are really proclaimed a master. Even though shogi no longer uses the iemoto system, masters are still looked upon in a similar way. Of course, the day will come when they are surpassed by computers, but it will not be without some painful losses, actually. I also admire computers and their incredibly rapid development, but I have sympathy for the professional players who will fight and eventually fail in the name of human intellect. Have you read the novel Meijin by Yasunari Kawabata? It's about go, not shogi, but the concept is that an old master (symbolizing old Japan where artistic beauty was of utmost importance) is defeated by a youngster (who is standing as a symbol for a new, pragmatic way of thinking). At that time, after World War II, Japan has passed into a new age of modernism. Perhaps we're about to experience something similar now, when the human master is defeated by a computer, just like technology is taking the place of Japanese arts in everyday life as well. If that happens, only one question remains: will shogi itself, as an art or a game, survive at all? Already, the number of younger players is decreasing greatly - in most shogi clubs, some 90 percent of the players are over the age of 40. As computers and gaming consoles are spreading among young people, shogi as an activity is losing more and more territory. Perhaps the day will come when shogi is not played anymore, and fading out of everyday life, it becomes a matter of the past. I can only hope such a day will not come - in fact, that's one of the reasons why I'm doing my best to spread shogi in Hungary. > I think we could know when a computer passes the top pro without> playing the top pro. If the computer beats the amateur champion in a> match with an 8 piece handicap. At that point it will not be necessary> to play the top pro. The top pro would also lose. But this will take a> long time by 2030.> >> > Such a thing isn't going to happen, as 8 and even 6 piece handicaps are fairly simple because of the limited number of variations. So I think a 1 Dan amateur playing shita-te in a 6 piece handicap game could defeat a supercomputer that sees every possible move and position in shogi (like the "god of shogi"). However, from 4 piece handicap and on, shogi becomes much more difficult, so such a thing might actually happen in the distant future in 4 piece handicap games. However, there is also another reason why computers are bad at playing uwa-te (giving a handicap). Uwa-te often has to decide which move to make:> >> > Move 1.: The "best" move. Because of the handicap, however, even the best move will result in a worse position eventually.> > Move 2.: A "risky" move, inviting a mistake. If shita-te plays well and doesn't make the mistake, uwa-te will have a great disadvantage, and will have to resign soon - so by conventional judgement, this move is worse than the "best" move. However, even though this move is risky, uwa-te knows it will be difficult for shita-te to play well in this position. And if shita-te makes the mistake, the position will turn so now uwa-te has the advantage.> >> > In such a situation, a computer chooses Move 1, because it is the best move, while humans prefer to play Move 2. This is the reason why humans do much better as uwa-te of handicap games then computers. Koji Tanigawa has written an excellent book (the title is "Taikyokukan ga Shobu wo Kimeru"), where he describes how one should choose the best move (a simple line) when the position is favorable, and a difficult line when one is at a disadvantage. However, I think this concept is beyond the level of the current state of artificial intelligence.> >> > > When you play Tanase Shogi do you use a handicap? At what handicap> can you beat Tanase Shogi?> >> > No, I don't use a handicap. I once tried rook handicap, but it was too easy even for an amateur 2 Dan like me - programs strong in even games are still incredibly weak at giving handicaps, for the reasons I've described above. (And perhaps giving handicaps is not really the focus of software developers, either.) So I play only even games against it (mostly lower levels, as the top level is way too strong for me), and use the top level for analyzing my games.> >> > > I looked for computers at Shogi Club 24. I only found an old> version of YSS with a rating of 2744. Are there higher computer> ratings on Shogi Club 24?> >> > I don't know. I
>