Why did Kasparov blink?
Tom King
entitled "Why did Kasparov blink?". It refers to the position after move
19 in game 5 of the Kasparov-Deep Blue match 1996. (I don't have the PGN
for this position, sorry).
Anyhow, according to Berliner, B*Hitech (his program) found the key move
Bxh7+ after 3.1 hours of calculation. For fun, I put this position to my
program, Francesca, running on a Cyrix P233. To my delight, it found the
move Bxh7+ after 2 hours of calculating, although it was only scoring
+0.27 for the move, so it (she) didn't view Bxh7+ to be a winning move.
How do other programs do on this one? (I guess the commercials get it
very quickly? Does any program get this on a standard tournament level?
how about blitz?)
--
Tom King
Dann Corbit
create the EPD in a jiffy, but it would be helpful to many I am sure.
--
C-FAQ: http://www.eskimo.com/~scs/C-faq/top.html
"The C-FAQ Book" ISBN 0-201-84519-9
C.A.P. Newsgroup http://www.dejanews.com/~c_a_p
C.A.P. FAQ: ftp://38.168.214.175/pub/Chess%20Analysis%20Project%20FAQ.htm
Henri H. Arsenault
<t...@hatbulb.demon.co.uk> wrote:
> How do other programs do on this one? (I guess the commercials get it
> very quickly? Does any program get this on a standard tournament level?
> how about blitz?)
> --
Fritz 5.32 finds it in about a minute,if my memory is correct, and if I
remember correctly, Crafty also identifies it in a very short time. I
don't have Fritz here so I can't verify it, but I'll try to remember to
check it out tonight.
In my view, considering that much weaker programs than DB can identify
this move, as well as others in this game, it was not DB who cheated as
Kasparov claims, but KASPAROV who cheated by playing a series of weak
moves that could only come from a computer, and not from a human being...
;>)
Henri
Henri H. Arsenault
<dco...@solutionsiq.com> wrote:
> Why not post the EPD for the move? I could probably find the game and
> create the EPD in a jiffy, but it would be helpful to many I am sure.
> --
I posted the game on this forum a week or two ago in the course of a
discussion with KK on the Fritz infinite mode VS correspondence mode. The
title of the thread is Fritz 5.32 beats Kasparov and Deep Blue.
The message has the full game and the analysis by Friedel, as well as the
full score and the analysis by Fritz, who in the correspondence mode not
only found the move but gave the position as won for Deep Blue.
I can repost it if necessary.
Henri
Robert Hyatt
: In the June 1996 ICCA journal, there's a little article by Hans Berliner
: entitled "Why did Kasparov blink?". It refers to the position after move
: 19 in game 5 of the Kasparov-Deep Blue match 1996. (I don't have the PGN
: for this position, sorry).
: Anyhow, according to Berliner, B*Hitech (his program) found the key move
: Bxh7+ after 3.1 hours of calculation. For fun, I put this position to my
: program, Francesca, running on a Cyrix P233. To my delight, it found the
: move Bxh7+ after 2 hours of calculating, although it was only scoring
: +0.27 for the move, so it (she) didn't view Bxh7+ to be a winning move.
: How do other programs do on this one? (I guess the commercials get it
: very quickly? Does any program get this on a standard tournament level?
: how about blitz?)
: --
: Tom King
Several of us found this during the tournament. But according to analysis
from Deep Blue it saw a draw, if I recall correctly. I don't think
Kasparov wanted to take a chance on an unexpected move far into the combination
and lose when he thought his position was already 'quite ok.'
I don't remember who first suggested it, probably an IM/GM that was
commenting...
--
Robert Hyatt Computer and Information Sciences
hy...@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170
Robert Hyatt
: In article <804hRAAn...@hatbulb.demon.co.uk>, Tom King
: <t...@hatbulb.demon.co.uk> wrote:
:> How do other programs do on this one? (I guess the commercials get it
:> very quickly? Does any program get this on a standard tournament level?
:> how about blitz?)
:> --
: Fritz 5.32 finds it in about a minute,if my memory is correct, and if I
: remember correctly, Crafty also identifies it in a very short time. I
: don't have Fritz here so I can't verify it, but I'll try to remember to
: check it out tonight.
: In my view, considering that much weaker programs than DB can identify
: this move, as well as others in this game, it was not DB who cheated as
: Kasparov claims, but KASPAROV who cheated by playing a series of weak
: moves that could only come from a computer, and not from a human being...
: ;>)
: Henri
This was a couple of years ago, obviously, but I believe that Hsu said they
were expecting this and had a score of "draw". But I could be mistaken.
bruce moreland
Won for Deep Blue? The questions are as follows:
1) Does 1. Bxh7+ win?
2) Does 1. a3 Ba5 2. Bxh7+ win?
Even if the answer to both is "no", then 1. a3 Ba5 2. b4 Bc7 3. c5 is
a nice bind, as played in the game.
Perhaps you mean to say "won for Kasparov".
bruce
bruce moreland
wrote:
>Henri H. Arsenault <ars...@phy.ulaval.ca> wrote:
>: In article <804hRAAn...@hatbulb.demon.co.uk>, Tom King
>: <t...@hatbulb.demon.co.uk> wrote:
>
>:> How do other programs do on this one? (I guess the commercials get it
>:> very quickly? Does any program get this on a standard tournament level?
>:> how about blitz?)
>:> --
>: Fritz 5.32 finds it in about a minute,if my memory is correct, and if I
>: remember correctly, Crafty also identifies it in a very short time. I
>: don't have Fritz here so I can't verify it, but I'll try to remember to
>: check it out tonight.
>
>: In my view, considering that much weaker programs than DB can identify
>: this move, as well as others in this game, it was not DB who cheated as
>: Kasparov claims, but KASPAROV who cheated by playing a series of weak
>: moves that could only come from a computer, and not from a human being...
>: ;>)
>
>: Henri
>
>This was a couple of years ago, obviously, but I believe that Hsu said they
>were expecting this and had a score of "draw". But I could be mistaken.
Having little experience with DB, but a lot of experience with
computers saying "draw" a moment before they explode, I think that
this position is a good candidate for that situation.
A draw is when you are up material but your opponent is checking you
all over the place, but doesn't have enough draft to make the quiet
move that kills you.
bruce
bruce moreland
wrote:
>Tom King <t...@hatbulb.demon.co.uk> wrote:
>: In the June 1996 ICCA journal, there's a little article by Hans Berliner
>: entitled "Why did Kasparov blink?". It refers to the position after move
>: 19 in game 5 of the Kasparov-Deep Blue match 1996. (I don't have the PGN
>: for this position, sorry).
>
>: Anyhow, according to Berliner, B*Hitech (his program) found the key move
>: Bxh7+ after 3.1 hours of calculation. For fun, I put this position to my
>: program, Francesca, running on a Cyrix P233. To my delight, it found the
>: move Bxh7+ after 2 hours of calculating, although it was only scoring
>: +0.27 for the move, so it (she) didn't view Bxh7+ to be a winning move.
>
>: How do other programs do on this one? (I guess the commercials get it
>: very quickly? Does any program get this on a standard tournament level?
>: how about blitz?)
>: --
>: Tom King
>
>Several of us found this during the tournament. But according to analysis
>from Deep Blue it saw a draw, if I recall correctly. I don't think
>Kasparov wanted to take a chance on an unexpected move far into the combination
>and lose when he thought his position was already 'quite ok.'
He probably made the very deliberate decision to avoid complications
and simply sit on the thing until it died, since DB's real tactical
ability was (and seemingly still is) unknown.
This strategy worked like a champ in that game.
When a computer gives you a gift, it's either because the computer is
brilliant or because it is a moron. Kasparov gave it the benefit of
the doubt.
bruce
Henri H. Arsenault
moreland) wrote:
>
> Won for Deep Blue? The questions are as follows:
>
> 1) Does 1. Bxh7+ win?
> 2) Does 1. a3 Ba5 2. Bxh7+ win?
>
> Even if the answer to both is "no", then 1. a3 Ba5 2. b4 Bc7 3. c5 is
> a nice bind, as played in the game.
>
> Perhaps you mean to say "won for Kasparov".
I just realized that I may be speaking of the wrong move; I thought the
move was the Deep Blue move 29. Nxb7! against Kasparov, which was followed
by a win for Deep Blue, but I just noticed that some are speaking of Bxh7.
Was it Bxh7, and in what game and who won?
Anyway for the 29. Nxb7 move, which I thought was the move that Kasparov
claims that no computer would play, I made some tests last night.
a) In the infinite mode, Fritz 5.32 put the move on the top of it's list
in the 15th ply, after about two hours of analysis, during which the move
never dropped below third place in the ranking. The reason for the long
time is that the move Qd5 is almost as good and also wins. The move Rd1 is
also strong, but not as much.
b) in the correspondence mode, Fritz 5.32 identifies the move as a WINNING
move in less than five minutes, using a minimum setting of 3/2/2 branches
for moves 1/2/3, a ply depth of 9 and an analysis length of 8.
c) Crafty 16.5 identifies Nxb7 as a the best move (and a winning move) in
a couple of seconds, and keeps it in top spot all the way up to 15 plies,
where I stopped it.
In sum, for a computer of Deep Blue's speed, it was nothing at all to
identify this move as a winning move in a couple of minutes, since it is
fifty times faster than Fritz.
Kasparov lost the game because of a series of weak moves, and the analysis
of the move 29 Nxb7 shows that after Kasparov's 28...d5?, his game was
definitely lost, if not before.
Henri
Robert Hyatt
: moreland) wrote:
:>
:> Won for Deep Blue? The questions are as follows:
:>
:> 1) Does 1. Bxh7+ win?
:> 2) Does 1. a3 Ba5 2. Bxh7+ win?
:>
:> Even if the answer to both is "no", then 1. a3 Ba5 2. b4 Bc7 3. c5 is
:> a nice bind, as played in the game.
:>
:> Perhaps you mean to say "won for Kasparov".
: I just realized that I may be speaking of the wrong move; I thought the
: move was the Deep Blue move 29. Nxb7! against Kasparov, which was followed
: by a win for Deep Blue, but I just noticed that some are speaking of Bxh7.
: Was it Bxh7, and in what game and who won?
: Anyway for the 29. Nxb7 move, which I thought was the move that Kasparov
: claims that no computer would play, I made some tests last night.
: a) In the infinite mode, Fritz 5.32 put the move on the top of it's list
: in the 15th ply, after about two hours of analysis, during which the move
: never dropped below third place in the ranking. The reason for the long
: time is that the move Qd5 is almost as good and also wins. The move Rd1 is
: also strong, but not as much.
Everyone will slowly realize that when someone makes a statement like "No
computer will play this" they are doing so out of _gross_ ignorance. IE
recall the debacle here with the "Shirov" statements a while back. He
had posted several positions that "no computer could solve" and several
ran them against various programs (crafty included) and found that rather
than "no computer could solve them" it should have been "most computers can
solve them".
: b) in the correspondence mode, Fritz 5.32 identifies the move as a WINNING
: move in less than five minutes, using a minimum setting of 3/2/2 branches
: for moves 1/2/3, a ply depth of 9 and an analysis length of 8.
: c) Crafty 16.5 identifies Nxb7 as a the best move (and a winning move) in
: a couple of seconds, and keeps it in top spot all the way up to 15 plies,
: where I stopped it.
: In sum, for a computer of Deep Blue's speed, it was nothing at all to
: identify this move as a winning move in a couple of minutes, since it is
: fifty times faster than Fritz.
But Fritz couldn't find it quickly. And most of his 'analysis' and 'experience'
was with fritz. So if fritz can't do it, no computer can do it. :)
: Kasparov lost the game because of a series of weak moves, and the analysis
: of the move 29 Nxb7 shows that after Kasparov's 28...d5?, his game was
: definitely lost, if not before.
: Henri
Nah. It was the cheating. And nothing but the cheating. His mistakes had
nothing to do with it at all. :)
Adnan
wrote:
>Anyway for the 29. Nxb7 move, which I thought was the move that Kasparov
>claims that no computer would play, I made some tests last night.
29. Nxb7 is a typical computer move. Hiarcs 6 "finds" (find is in quotes, of
course), it in 0.01 sec. Fritz for some reason considers Qd5 stronger. I did
not bother to wait more than 5 sec on Fritz because there is no need to. 29.
Nxb7 is a computer type move. What is special about it? Even Fritz considers
it as one of the best three.
Perhaps Kasparov made his cheating accusation on some other move? He can't be
that idiot. 29. Nxb7 is not only easy for computers, but it *is* typical for
computers to play moves like this. As I said, Hiarcs 6 "finds" it in 0.01
sec.
The original question was about the game six of 1996 where Deep Blue allowed
20. Bxh7 by playing 19. ... Nge7. Now that's more difficult for computers, and
perhaps Deep Blue simply made a tactical blunder. Kasparov did not play 20.
Bxh7 (even when he saw it). He played 20. a3 and still won.
Tom King
<hy...@crafty.cis.uab.edu> writes>:> How do other programs do on this one? (I guess the commercials get it
>:> very quickly? Does any program get this on a standard tournament level?
>:> how about blitz?)
>:> --
>: remember correctly, Crafty also identifies it in a very short time. I
>: don't have Fritz here so I can't verify it, but I'll try to remember to
>: check it out tonight.
>
>: In my view, considering that much weaker programs than DB can identify
>: this move, as well as others in this game, it was not DB who cheated as
>: Kasparov claims, but KASPAROV who cheated by playing a series of weak
>: moves that could only come from a computer, and not from a human being...
>: ;>)
>
>: Henri
>
>This was a couple of years ago, obviously, but I believe that Hsu said they
>were expecting this and had a score of "draw". But I could be mistaken.
>
Were they mistaken, then? I mean does Bxh7+ lead to a win (or a
substantial advantage). Could you let Crafty think on this one for a
while, Bob? I'd be interested to see what it finds..
--
Tom King
Tom King
Arsenault <ars...@phy.ulaval.ca> writes
>In article <3721cfba....@news.seanet.com>, bru...@seanet.com (bruce
>moreland) wrote:
>>
>> Won for Deep Blue? The questions are as follows:
>>
>> 1) Does 1. Bxh7+ win?
>> 2) Does 1. a3 Ba5 2. Bxh7+ win?
>>
>> Even if the answer to both is "no", then 1. a3 Ba5 2. b4 Bc7 3. c5 is
>> a nice bind, as played in the game.
>>
>> Perhaps you mean to say "won for Kasparov".
>
>I just realized that I may be speaking of the wrong move; I thought the
>move was the Deep Blue move 29. Nxb7! against Kasparov, which was followed
>by a win for Deep Blue, but I just noticed that some are speaking of Bxh7.
>Was it Bxh7, and in what game and who won?
>
This is deinitely a different game. The game we're taking about is Deep
Blue vs Kasparov, game 5 *1996*.
>Anyway for the 29. Nxb7 move, which I thought was the move that Kasparov
>claims that no computer would play, I made some tests last night.
>
>a) In the infinite mode, Fritz 5.32 put the move on the top of it's list
>in the 15th ply, after about two hours of analysis, during which the move
>never dropped below third place in the ranking. The reason for the long
>time is that the move Qd5 is almost as good and also wins. The move Rd1 is
>also strong, but not as much.
>
>b) in the correspondence mode, Fritz 5.32 identifies the move as a WINNING
>move in less than five minutes, using a minimum setting of 3/2/2 branches
>for moves 1/2/3, a ply depth of 9 and an analysis length of 8.
>
>c) Crafty 16.5 identifies Nxb7 as a the best move (and a winning move) in
>a couple of seconds, and keeps it in top spot all the way up to 15 plies,
>where I stopped it.
>
>In sum, for a computer of Deep Blue's speed, it was nothing at all to
>identify this move as a winning move in a couple of minutes, since it is
>fifty times faster than Fritz.
>
>Kasparov lost the game because of a series of weak moves, and the analysis
>of the move 29 Nxb7 shows that after Kasparov's 28...d5?, his game was
>definitely lost, if not before.
>
>Henri
--
Tom King
Robert Hyatt
: In article <7d6g5d$f5p$4...@juniper.cis.uab.edu>, Robert Hyatt
: --
: Tom King
I let it run a good while on the quad, and I always thought black was
slightly better, ie the score was about +.3, after playing Bxh7+ and
then letting it run. So maybe Hsu was right. DB either correctly thought
this led to a draw, or the score was close enough to zero that the search
thought that 0.00 was the best that could happen.
I'll try it again on the quad P6/200 and just let it grind, once some tests
I am running over there have finished...
Adnan
wrote:
>Anyway for the 29. Nxb7 move, which I thought was the move that Kasparov
>claims that no computer would play, I made some tests last night.
The other problem with Nxb7 is that it is not the only winning move in this
position. There is nothing beautiful about Nxb7; it is computer like move of
grabbing pawn and hoping not to get mated. Other than that, what makes Nxb7 so
"beautiful" that one would excuse Deep Blue's team of cheating? Nxb7 is not
even the only winning move in this position! I doubt that Kasparov can be that
idiot! (i.e. accusing Deep Blue of cheating because of moves like Nxb7).
That's silly. He must be referring to some other game.
Adnan
>Were they mistaken, then? I mean does Bxh7+ lead to a win (or a
>substantial advantage). Could you let Crafty think on this one for a
>while, Bob? I'd be interested to see what it finds..
Without bothering to do any calculation I would say that Bxh7 would win. And I
think most humans would play it without much consideration. This type of
speculative sacrifice is difficult for computers but for humans it is easy to
find. Kasparov did not play Bxh7, but he certainly would have if he was
playing against a human opponent.
Adnan
>I let it run a good while on the quad, and I always thought black was
>slightly better, ie the score was about +.3, after playing Bxh7+ and
>then letting it run. So maybe Hsu was right. DB either correctly thought
>this led to a draw, or the score was close enough to zero that the search
>thought that 0.00 was the best that could happen.
I doubt that. Play Bxh7 and let Crafty play vs Crafty in this position. White
is better. Moves such as Bxh7 are above computers. But even a low 1600 rated
human would find it wth no problem, I think.
Adnan
Better still, start your analysis after 20. Bxh7+ Kxh7. 21. Ng5+ Kg8. 22. Qh5.
Start from the position after Qh5 on crafty (to analyze it) because these
three moves are almost forced if Black wants to refute 20. Bxh7+
Adnan
>
>Better still, start your analysis after 20. Bxh7+ Kxh7. 21. Ng5+ Kg8. 22. Qh5.
Or start after 20. Bxh7 Kxh7. 21. Ng5+ Kg6. 22. Qg4
Adnan
>This is deinitely a different game. The game we're taking about is Deep
>Blue vs Kasparov, game 5 *1996*.
I thought it was game six
Robert Hyatt
: On 24 Mar 1999 00:14:57 GMT, Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
:>I let it run a good while on the quad, and I always thought black was
:>slightly better, ie the score was about +.3, after playing Bxh7+ and
:>then letting it run. So maybe Hsu was right. DB either correctly thought
:>this led to a draw, or the score was close enough to zero that the search
:>thought that 0.00 was the best that could happen.
: I doubt that. Play Bxh7 and let Crafty play vs Crafty in this position. White
: is better. Moves such as Bxh7 are above computers. But even a low 1600 rated
: human would find it wth no problem, I think.
that is no good. It produces an invalid or imaginary game. Because
each search will be based on a shallow tree, rather than one huge tree.
Playing bxh7 is ok, but you have to let the program search until the
score drops out the bottom. As I reported before, DB had evaluated this
to be a draw and was expecting it, if I am not mistaken. I will let it run
a while when the machine is free...
If you just play and change sides, if one side makes a mistake due to a
search that was too shallow, the game ends improperly.
Robert Hyatt
: Better still, start your analysis after 20. Bxh7+ Kxh7. 21. Ng5+ Kg8. 22. Qh5.
: Start from the position after Qh5 on crafty (to analyze it) because these
: three moves are almost forced if Black wants to refute 20. Bxh7+
I tried this thru 12 plies. Black is still positive, it seems to like
Nf5 best, score = +.37 (for black) after the moves above.
It has also liked other moves here. Like Qe8. I'm not sure how white
is supposed to win this.
Adnan
>Adnan <kar...@softhome.net> wrote:
>
>: Better still, start your analysis after 20. Bxh7+ Kxh7. 21. Ng5+ Kg8. 22. Qh5.
>
>
>: Start from the position after Qh5 on crafty (to analyze it) because these
>: three moves are almost forced if Black wants to refute 20. Bxh7+
>
>
>I tried this thru 12 plies. Black is still positive, it seems to like
>Nf5 best, score = +.37 (for black) after the moves above.
After Nf5 play cxd5 and let crafty think after cxd5. I think analyzing this
position is the right thing, because this is the critical position in Bxh7
line. These moves are almost forced.
White looks very dangerous to me here.
Robert Hyatt
Sorry... I screwed up above... It changed from Nf5 at depth=11. I somehow hit
the 'scroll tab' which backed the stuff up and hid the bottom part of the analysis.
It actually wants to play Qe8 here...
So we are now at Bxh7+ Kxh7 Ng5+ Kg8 Qh5 Qe8.
next move?
Blaise
Could someone please send in the position FEN it愀 very annoying to see
moves discussed without a a starting position.
Adnan
>Sorry... I screwed up above... It changed from Nf5 at depth=11. I somehow hit
>the 'scroll tab' which backed the stuff up and hid the bottom part of the analysis.
>
>It actually wants to play Qe8 here...
>
>So we are now at Bxh7+ Kxh7 Ng5+ Kg8 Qh5 Qe8.
>
>next move?
Well, 23. Qh7+ Kf8 24. Qh8+ Ng8 are obvious moves.
Let Crafty evaluate white position after Ng8. One obvious line white can play
after Ng8 is 25. Nh7+ followed by Qxg7.
What does Crafty think about the position after Qxg7?
bruce moreland
wrote:
>Adnan <kar...@softhome.net> wrote:
>: On 24 Mar 1999 03:56:28 GMT, Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
>
>:>Adnan <kar...@softhome.net> wrote:
>:>
>:>: Better still, start your analysis after 20. Bxh7+ Kxh7. 21. Ng5+ Kg8. 22. Qh5.
>:>
>:>
>:>: Start from the position after Qh5 on crafty (to analyze it) because these
>:>: three moves are almost forced if Black wants to refute 20. Bxh7+
>:>
>:>
>:>I tried this thru 12 plies. Black is still positive, it seems to like
>:>Nf5 best, score = +.37 (for black) after the moves above.
>
>: After Nf5 play cxd5 and let crafty think after cxd5. I think analyzing this
>: position is the right thing, because this is the critical position in Bxh7
>: line. These moves are almost forced.
>
>: White looks very dangerous to me here.
>
>
>Sorry... I screwed up above... It changed from Nf5 at depth=11. I somehow hit
>the 'scroll tab' which backed the stuff up and hid the bottom part of the analysis.
>
>It actually wants to play Qe8 here...
>
>So we are now at Bxh7+ Kxh7 Ng5+ Kg8 Qh5 Qe8.
>
>next move?
Berliner claims that 22. ... Nf5 is forced, and gives 23. cxd5!! Nxe3
24. dxe6! fxe6 25. fxe3 and Berliner claims that this is easily won.
He also analyzes 23 ... Re8 24. dxe6! Rxe6! 25. Qh7+ Kf8 26. Nxe6+
fxe6 27. d5! Nxe3 28. Qh8+ Ke7 29. Qxg7+ Kd8 30. Qxd7+ Kxd7 31. dxc6+
"and wins".
Berliner briefly mentions that the sacrifice is better after the game
continuation 20. a3 Ba5, now 21. Bxh7+ is stronger than before because
"several of the key defenses mentioned above are no longer playable
because the B at a5 will be en prise."
My program wants to play 22. ... Qe8, but I don't think it understand
this in the amount of time I gave it to think. I don't know what the
obvious win is after 22. ... Qe8.
bruce
bruce moreland
>On 24 Mar 1999 00:14:57 GMT, Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
>
>
>>I let it run a good while on the quad, and I always thought black was
>>slightly better, ie the score was about +.3, after playing Bxh7+ and
>>then letting it run. So maybe Hsu was right. DB either correctly thought
>>this led to a draw, or the score was close enough to zero that the search
>>thought that 0.00 was the best that could happen.
>
>I doubt that. Play Bxh7 and let Crafty play vs Crafty in this position. White
>is better. Moves such as Bxh7 are above computers. But even a low 1600 rated
>human would find it wth no problem, I think.
Anyone who has seen the thematic Bxh7+ in the French Defense will see
Bxh7+ here.
If this position is a tactical test, Bxh7+ is the obvious answer.
But it's easier to see than to prove that it works.
bruce
Adnan
>Berliner briefly mentions that the sacrifice is better after the game
>continuation 20. a3 Ba5, now 21. Bxh7+ is stronger than before because
>"several of the key defenses mentioned above are no longer playable
>because the B at a5 will be en prise."
I don't like a3 here. It blocks the possibility of Ba3 in some variations.
>My program wants to play 22. ... Qe8, but I don't think it understand
>this in the amount of time I gave it to think. I don't know what the
>obvious win is after 22. ... Qe8.
As I asked Hyatt, what does your program says after 23. Qh8+ Ng8 24. Nh7+ Ke7
25. Qxg7
White already has two pawns for his piece. White's h pawn is passed, and
Black's King is stuck in the middle. What can be better than that?
Adnan
>As I asked Hyatt, what does your program says after 23. Qh8+ Ng8 24. Nh7+ Ke7
>25. Qxg7
Typo. I meant (after Black's Qe8) 23. Qh7+ Kf8. 24. Qh8+ Ng8 24. Nh7+ Ke7 25.
Qxg7
Why isn't white winning here? He has two pawns for his piece. He has a passed
h pawn, and black king is stuck in the center.
It's obvious that Deep Blue's Nge7 allowing Bxh7+ was a tactical blunder, but
it is a blunder that one would expect a computer to make.
Adnan
>If this position is a tactical test, Bxh7+ is the obvious answer.
>
>But it's easier to see than to prove that it works.
And why exactly it does not work? It looks to me that it works.
At any rate, it's your turn to move after. 20. Bxh7+ Kxh7 21. Ng5+ Kg8. 22.
Qh5 Qe8 22. Qh7+ Kf8. 23. Qh8+ Ng8. 24. Nh7+ Ke7 25. Qxg7.
Adnan
>
>Could someone please send in the position FEN it愀 very annoying to see
>moves discussed without a a starting position.
I don't have FEN but it is game 6 of 96 Kasparov Deep Blue match. Here is the
game from my CB7.
Kasparov,G - Deep Blue [D30]
Philadelphia, USA (6), 1996
[Adnan]
1.Ng1-f3 d7-d5 2.d2-d4 c7-c6 3.c2-c4 e7-e6 4.Nb1-d2 Ng8-f6 5.e2-e3 c6-c5
6.b2-b3 Nb8-c6 7.Bc1-b2 c5xd4 8.e3xd4 Bf8-e7 9.Ra1-c1 0-0 10.Bf1-d3 Bc8-d7
11.0-0 Nf6-h5 12.Rf1-e1 Nh5-f4 13.Bd3-b1 Be7-d6 14.g2-g3 Nf4-g6 15.Nf3-e5
Ra8-c8 16.Ne5xd7 Qd8xd7 17.Nd2-f3 Bd6-b4 18.Re1-e3 Rf8-d8 19.h2-h4 Ng6-e7
20.a2-a3 [20.Bb1xh7+ Kg8xh7 21.Nf3-g5+ Kh7-g8 22.Qd1-h5 Qd7-e8 (22...Ne7-f5
23.c4xd5+-) 23.Qh5-h7+ Kg8-f8 24.Qh7-h8+ Ne7-g8 25.Ng5-h7+ Kf8-e7 26.Qh8xg7]
20...Bb4-a5 21.b3-b4 Ba5-c7 22.c4-c5 Rd8-e8 23.Qd1-d3 g7-g6 24.Re3-e2 Ne7-f5
25.Bb2-c3 h7-h5 26.b4-b5 Nc6-e7 27.Bc3-d2 Kg8-g7 28.a3-a4 Rc8-a8 29.a4-a5
a7-a6 30.b5-b6 Bc7-b8 31.Bb1-c2 Ne7-c6 32.Bc2-a4 Re8-e7 33.Bd2-c3 Nc6-e5
34.d4xe5 Qd7xa4 35.Nf3-d4 Nf5xd4 36.Qd3xd4 Qa4-d7 37.Bc3-d2 Re7-e8 38.Bd2-g5
Re8-c8 39.Bg5-f6+ Kg7-h7 40.c5-c6 b7xc6 41.Qd4-c5 Kh7-h6 42.Re2-b2 Qd7-b7
43.Rb2-b4 1-0
Adnan
>On Wed, 24 Mar 1999 09:33:48 +0100, "Blaise" <cal...@tninet.se> wrote:
>>Could someone please send in the position FEN it愀 very annoying to see
>>moves discussed without a a starting position.
>I don't have FEN but it is game 6 of 96 Kasparov Deep Blue match. Here is the
>game from my CB7.
Well, I just looked around in CB7, and found FEN. Here it is:
2rr2k1/pp1qnppp/2n1p3/3p4/1bPP3P/1P2RNP1/PB3P2/1BRQ2K1 w - - 0 20
Black (Deep Blue) just played Nge7, allowing Bxh7+
Robert Hyatt
: On Wed, 24 Mar 1999 09:28:19 GMT, bru...@seanet.com (bruce moreland) wrote:
:>If this position is a tactical test, Bxh7+ is the obvious answer.
:>
:>But it's easier to see than to prove that it works.
: And why exactly it does not work? It looks to me that it works.
What he means is that so far no one has shown a variation where black
is mated or hopelessly lost. Two pawns for a piece is _not_ winning
in most positions. And notice how 'deep' we are now search-depth wise.
In a game with a million dollars on the line, would you _really_ play
Bxh7+ against the strongest tactician in the world, knowing that one
loss might cost you over 1/4 million dollars? I think that's why
Kasparov didn't play this. He couldn't convince himself that it was
winning for certain. And he didn't want to 'roll the dice'.
: At any rate, it's your turn to move after. 20. Bxh7+ Kxh7 21. Ng5+ Kg8. 22.
: Qh5 Qe8 22. Qh7+ Kf8. 23. Qh8+ Ng8. 24. Nh7+ Ke7 25. Qxg7.
I will try this in a bit. My quad xeon is playing someone on ICC at
present, and the quad P6 is running a parallel test that will run for
several more hours...
Robert Hyatt
: Could someone please send in the position FEN it愀 very annoying to see
: moves discussed without a a starting position.
We are starting here:
2rr2k1/pp1qnppp/2n1p3/3p4/1bPP3P/1P2RNP1/PB3P2/1BRQ2K1 w
with white to move and play Bxh7+
Bob
Blaise
Adnan
>What he means is that so far no one has shown a variation where black
>is mated or hopelessly lost. Two pawns for a piece is _not_ winning
>in most positions. And notice how 'deep' we are now search-depth wise.
>In a game with a million dollars on the line, would you _really_ play
>Bxh7+ against the strongest tactician in the world, knowing that one
>loss might cost you over 1/4 million dollars? I think that's why
>Kasparov didn't play this. He couldn't convince himself that it was
>winning for certain. And he didn't want to 'roll the dice'.
Actually white follow up moves are pretty straight forward. Certainly not
difficult for a players of Kasparov caliber. But at any rate, the question was
whether or not Deep Blue made a tactical mistake by allowing Bxh7.
Henri H. Arsenault
<hy...@crafty.cis.uab.edu> wrote:
> I will try this in a bit. My quad xeon is playing someone on ICC at
> present, and the quad P6 is running a parallel test that will run for
> several more hours...
>
>
The Fritz 5.32 correspondence mode analyzing the position (2/2/2 9 plies
deep with analysis length of 8) after Qh5+ thinks that it is a draw
(advantage of +/= 0.37 for White), and agrees that ... Qe8 is Black's best
move and that Nf5 loses, but I will carry out a deeper analysis tonight
and post the result tomorrow.
Henri
Adnan
On Mon, 22 Mar 1999 20:58:15 +0000, Tom King <t...@hatbulb.demon.co.uk> wrote:
So here is the variation now:
20. Bxh7 Kxh7. 21 Ng5 Kg8. 22 Qh5 Qe8. 23. Qh7+ Kf8. 24. Qh8+ Ng8 24. Nh7+ Ke7
25. Qxg7
If Crafty plays 25. ...Bd2, then 26. Ba3+ Rd6. 26 cxd5 Bxe3. 27. fxe3 exd5 28.
Rf1 +-
Adnan
wrote:
>The Fritz 5.32 correspondence mode analyzing the position (2/2/2 9 plies
>deep with analysis length of 8) after Qh5+ thinks that it is a draw
>(advantage of +/= 0.37 for White), and agrees that ... Qe8 is Black's best
>move and that Nf5 loses, but I will carry out a deeper analysis tonight
>and post the result tomorrow.
Well, as I said in another post. Let Fritz improve on this variation:
20. Bxh7 Kxh7. 21 Ng5 Kg8. 22 Qh5 Qe8. 23. Qh7+ Kf8. 24. Qh8+ Ng8 24. Nh7+ Ke7
25. Qxg7
If Crafty plays 25. ...Bd2, then 26. Ba3+ Rd6. 26 cxd5 Bxe3. 27. fxe3 exd5 28.
Rf1 +-
Why does not this win?
Jeremiah Penery
>
> On Wed, 24 Mar 1999 09:28:19 GMT, bru...@seanet.com (bruce moreland) wrote:
>
> >If this position is a tactical test, Bxh7+ is the obvious answer.
> >
> >But it's easier to see than to prove that it works.
>
> And why exactly it does not work? It looks to me that it works.
>
Perhaps: 26. ... Bd2 27. Ba3+ Rd6 28. Ng5 Bxe3 29. Bxd6+ Kxd6 30. Nxf7+
Kc7
31. fxe3 Nge7 32. cxd5 exd5 33. Ne5 Kb6 34. Qf6 Qg8
After these moves, white's advantage, if any, is still quite small. It
might continue with: 35. Kh2 a6 36. Rc5 Ka7 37. Nxc6+ Nxc6 38. Qf5 Rd8
39. h5 Rd6
40. g4 b6 41. Qf4
Jeremiah
Adnan
Adnan
>Perhaps: 26. ... Bd2 27. Ba3+ Rd6 28. Ng5 Bxe3 29. Bxd6+ Kxd6 30. Nxf7+
>Kc7
>31. fxe3 Nge7 32. cxd5 exd5 33. Ne5 Kb6 34. Qf6 Qg8
Wrong.
My variation was:
20. Bxh7 Kxh7. 21 Ng5 Kg8. 22 Qh5 Qe8. 23. Qh7+ Kf8. 24. Qh8+ Ng8 25. Nh7+ Ke7
26. Qxg7.Bd2 27. Ba3+ Rd6. 28 cxd5 Bxe3. 29. fxe3 exd5 28 Re1.
Your Ng5 was not in my variation. Play cxd5 instead of Ng5
Adnan
>My variation was:
>
>20. Bxh7 Kxh7. 21 Ng5 Kg8. 22 Qh5 Qe8. 23. Qh7+ Kf8. 24. Qh8+ Ng8 25. Nh7+ Ke7
>26. Qxg7.Bd2 27. Ba3+ Rd6. 28 cxd5 Bxe3. 29. fxe3 exd5 28 Re1.
Opps. Last move is Rf1, not Re1
Blaise
Tom King
<hy...@crafty.cis.uab.edu> writes
>Tom King <t...@hatbulb.demon.co.uk> wrote:
>: In article <7d6g5d$f5p$4...@juniper.cis.uab.edu>, Robert Hyatt
>: <hy...@crafty.cis.uab.edu> writes
>:>Henri H. Arsenault <ars...@phy.ulaval.ca> wrote:
>:>: In article <804hRAAn...@hatbulb.demon.co.uk>, Tom King
>:>: <t...@hatbulb.demon.co.uk> wrote:
>:>
>:>:> How do other programs do on this one? (I guess the commercials get it
>:>:> very quickly? Does any program get this on a standard tournament level?
>:>:> how about blitz?)
>:>:> --
>:>: Fritz 5.32 finds it in about a minute,if my memory is correct, and if I
>:>: remember correctly, Crafty also identifies it in a very short time. I
>:>: don't have Fritz here so I can't verify it, but I'll try to remember to
>:>: check it out tonight.
>:>
>:>: In my view, considering that much weaker programs than DB can identify
>:>: this move, as well as others in this game, it was not DB who cheated as
>:>: Kasparov claims, but KASPAROV who cheated by playing a series of weak
>:>: moves that could only come from a computer, and not from a human being...
>:>: ;>)
>:>
>:>: Henri
>:>
>:>This was a couple of years ago, obviously, but I believe that Hsu said they
>:>were expecting this and had a score of "draw". But I could be mistaken.
>:>
>
>: Were they mistaken, then? I mean does Bxh7+ lead to a win (or a
>: substantial advantage). Could you let Crafty think on this one for a
>: while, Bob? I'd be interested to see what it finds..
>
>: --
>: Tom King
>
>I let it run a good while on the quad, and I always thought black was
>slightly better, ie the score was about +.3, after playing Bxh7+ and
>then letting it run. So maybe Hsu was right. DB either correctly thought
>this led to a draw, or the score was close enough to zero that the search
>thought that 0.00 was the best that could happen.
>
>I am running over there have finished...
>
>
ok, ok. But does any program actually want to play Bxh7+? If so, with
what score? Maybe someone could let Fritz or another tactical wizard
crunch on this overnight on a fast PC? How about ferret, Bruce? Does it
play Bxh7+?
Rgds,
--
Tom King
Henri H. Arsenault
>
> Well, as I said in another post. Let Fritz improve on this variation:
>
> 20. Bxh7 Kxh7. 21 Ng5 Kg8. 22 Qh5 Qe8. 23. Qh7+ Kf8. 24. Qh8+ Ng8 24. Nh7+ Ke7
> 25. Qxg7
>
> If Crafty plays 25. ...Bd2, then 26. Ba3+ Rd6. 26 cxd5 Bxe3. 27. fxe3 exd5 28.
> Rf1 +-
>
> Why does not this win?
Because 26...Re7! draws, a line that seems to have been missed by everyone
(but found thanks to the help of Fritz 5.32, an overnight search, and some
interactive playing around)...
26...Re7; 27 Ng5 Qf8 28 Rxe6+ Kd8; 29 Qh8 Be7;
30 Qh5 Bf6; 31 cxd5 Bxg5 =
Other moves for either side are inferior. This line shows that Bxh7 does
NOT lead to a forced win, but to equality with best play by both sides,
but in fact that the game still remains to be played out at that point,
since there are many occasions to slip up, which could tip the game either
way. It is hard to say if DB could have fouynd the move over the board,
but it is also hard to say if Kasparov could have picked his way without
tripping through the many pitfalls for White.
Henri
Adnan
wrote:
>Because 26...Re7! draws, a line that seems to have been missed by everyone
>(but found thanks to the help of Fritz 5.32, an overnight search, and some
>interactive playing around)...
>
>26...Re7; 27 Ng5 Qf8 28 Rxe6+ Kd8; 29 Qh8 Be7;
>30 Qh5 Bf6; 31 cxd5 Bxg5 =
What the hell are you talking about? 26. ... Re7 is illegal. If you mean
26... Rd7, see the analysis I posted in another post.
Better yet, post the whole variation starting from 20. Bxh7+, so we can follow
what you are saying.
Adnan
wrote:
>t is hard to say if DB could have fouynd the move over the board,
>but it is also hard to say if Kasparov could have picked his way without
>tripping through the many pitfalls for White.
There are no pitfalls for white. White moves are pretty straightforward.
Adnan
wrote:
>26...Re7; 27 Ng5 Qf8 28 Rxe6+ Kd8; 29 Qh8 Be7;
>30 Qh5 Bf6; 31 cxd5 Bxg5 =
Even in your variation (assuming you meant 26.. Rd7) 30. Nh7 (instead of 30.
Qh5) Qe8. Re2 white keeps strong pressure.
Adnan
Since you obviously did mean Rd7 intead of Re7, so let Fritz improve on this
variation. Most of the moves are forced, but some variations will baffle all
computers :-)
Let me know if any computer can find the beautiful pawn sac in this
variation:
26...Rd6(or d7) 27.cxd5 Rd6xd5 28.Ng5 Qf8 29.Qh7 Ke8 30.Rf3 Rxg5 31.h4xg5 Nge7
And now comes the computer buster :-)
32.d5!! (Can Fritz find d5?)
32... exd5 33.g6! Nxg6 34.Bg7 Qd6 35.Rxf7 +-
Ha ha ha ha!
Henri H. Arsenault
Yes, it WAS Rd7.
Yes, but there is no forced win for White. I don't have the analysis for
all the lines, but the line I gave (above) is the best move for each side
(according to the Fritz analysis). But I'll see if I can check out Nh7
with Fritz.
Henri
Henri H. Arsenault
> On Thu, 25 Mar 1999 12:54:23 GMT, ars...@phy.ulaval.ca (Henri H. Arsenault)
> wrote:
>
> >Because 26...Re7! draws, a line that seems to have been missed by everyone
> >(but found thanks to the help of Fritz 5.32, an overnight search, and some
> >interactive playing around)...
> >
> >26...Re7; 27 Ng5 Qf8 28 Rxe6+ Kd8; 29 Qh8 Be7;
> >30 Qh5 Bf6; 31 cxd5 Bxg5 =
>
> What the hell are you talking about? 26. ... Re7 is illegal. If you mean
> 26... Rd7, see the analysis I posted in another post.
>
> Better yet, post the whole variation starting from 20. Bxh7+, so we can follow
> what you are saying.
Sorry, I should have written 26...Rd7.
Here is the full variation with what Fritz 5.32 considers the best moves
for both sides; there are many places to go wrong for both sides,
including lines that give the other side a winning advantage.
20.Bxh7+ Kxh7 21.Ng5+ Kg8 22.Qh5 Qe8 23.Qh7+ Kf8 24.Qh8+ Ng8 25.Nh7+ Ke7
26.Qxg7 Rd7 27. Ng5 Qf8 28. Rxe6+ Kd8 29. Qh8 Be7 30 Qh5 Bf6 31 cxd5 Bxg5=
In a previous message, Adnan says that 26...Rd7 (which Fritz 5.32 found
after an all-night full-width analysis 14 plies deep) transposes to the
26...Rd6 variation, but Fritz considers this variation as inferior for
Black so he must be able to avoid it. I don't have the analysis for
26...Rd6, but I'll see if I can have Fritz analyze it again and post the
analysis here.
Henri
Adnan
wrote:
>20.Bxh7+ Kxh7 21.Ng5+ Kg8 22.Qh5 Qe8 23.Qh7+ Kf8 24.Qh8+ Ng8 25.Nh7+ Ke7
>26.Qxg7 Rd7 27. Ng5 Qf8 28. Rxe6+ Kd8 29. Qh8 Be7 30 Qh5 Bf6 31 cxd5 Bxg5=
>
>In a previous message, Adnan says that 26...Rd7 (which Fritz 5.32 found
>after an all-night full-width analysis 14 plies deep) transposes to the
>26...Rd6 variation, but Fritz considers this variation as inferior for
>Black so he must be able to avoid it. I don't have the analysis for
>26...Rd6, but I'll see if I can have Fritz analyze it again and post the
>analysis here.
What Fritz found was busted as I already posted in another post.
As for transposition, white can force Rd7 or Rd6 to my line, so I was
justified to say it transposes to variation above. It makes no difference (at
least as far as my variation is concerned) whether black plays 26. .. Rd7 or
26... Rd6. Because after cxd5 (which you do not even consider above) you are
back to my line.
Henri H. Arsenault
> What Fritz found was busted as I already posted in another post.
> As for transposition, white can force Rd7 or Rd6 to my line, so I was
> justified to say it transposes to variation above. It makes no difference (at
> least as far as my variation is concerned) whether black plays 26. .. Rd7 or
> 26... Rd6. Because after cxd5 (which you do not even consider above) you are
> back to my line.
not according to Fritz; here is the analysis by Fritz following 27. cxd5 after
20.Bxh7+ Kxh7 21.Ng5+ Kg8 22.Qh5 Qe8 23.Qh7+ Kf8 24.Qh8+ Ng8 25.Nh7+ Ke7
26.Qxg7 Rd7
27 cxd5 Rxd5 28. Ng3 Qf8 29. Qh7 Ke8 30. Nxe6 fxe6 31. Qxb7 Nge7 =/+
with a slight advantage to black.
Henri
Henri H. Arsenault
> > wrote:
> >
> >
> > >26...Re7; 27 Ng5 Qf8 28 Rxe6+ Kd8; 29 Qh8 Be7;
> > >30 Qh5 Bf6; 31 cxd5 Bxg5 =
> >
> > Even in your variation (assuming you meant 26.. Rd7) 30. Nh7 (instead of 30.
> > Qh5) Qe8. Re2 white keeps strong pressure.
>
the Black king eventually makes it to b8, and White's slight advantage
evaporates and he is left with a piece down for two pawns. I can post the
analysis if requested.
By the way, I forgot to mention in the other posting that 26...Rd6 is not
advantageous to White as you say, According to Fritz 5.32, your variation
26...Rd6 27. cxd5 Rxd5 28. Ng5 Qf8 Qh7 =/+ (-0.34)
yields a slight BLACK advantage and not a White one; where did you get the
+/- evaluation from?
In sum, so far, all the analysis seems to prove that not only is there not
a forced win for White after Bxh7, but many of the lines lead to an
advantage for Black; and in every case where Black plays the best lines
(which Deep Blue could have found in the time alloted), at best White is
left with equality having to fight an ending with a piece down and one of
his extra pawns isolated...
Kasparov was right not to try the sacrifice, he might have lost the game.
Henri
Jeremiah Penery
>
> Since you obviously did mean Rd7 intead of Re7, so let Fritz improve on this
> variation. Most of the moves are forced, but some variations will baffle all
> computers :-)
>
> Let me know if any computer can find the beautiful pawn sac in this
> variation:
>
> 26...Rd6(or d7) 27.cxd5 Rd6xd5 28.Ng5 Qf8 29.Qh7 Ke8 30.Rf3 Rxg5 31.h4xg5 Nge7
How about 28 ... Kd7 29. Nxf7 Qe7 30. Re5 Nf6 31. Rxd5+ Nxd5 32. h5 Rf8
33. Ne5+ Nxe5 34. Qxe5 Qf6 35. Qxf6 Rxf6
It seems quite equal for black.
Jeremiah Penery
Or 28. ... Kd7 29. Nxe6 Nce7 30. Nf4 Rxc1+ 31. Bxc1 Ra5 32. h5 Rxa2
33. h6 Qc8 34. Nd3 Qc2 35. Nxb4 Qxc1+
Adnan
wrote:
>> > Even in your variation (assuming you meant 26.. Rd7) 30. Nh7 (instead of 30.
>> > Qh5) Qe8. Re2 white keeps strong pressure.
>>
>I made a quick check with Fritz for 30. Nh7. This leads to equality, since
>the Black king eventually makes it to b8, and White's slight advantage
>evaporates and he is left with a piece down for two pawns. I can post the
>analysis if requested.
>
>By the way, I forgot to mention in the other posting that 26...Rd6 is not
>advantageous to White as you say, According to Fritz 5.32, your variation
>
>26...Rd6 27. cxd5 Rxd5 28. Ng5 Qf8 Qh7 =/+ (-0.34)
>
>yields a slight BLACK advantage and not a White one; where did you get the
>+/- evaluation from?
>
>In sum, so far, all the analysis seems to prove that not only is there not
>a forced win for White after Bxh7,
What ? What? What? What is your rating??? Am I just wasting my breath
here?????
Why don't you follow my analysis after Qh7 instead of just blindly trusting
Fritz evaluation??
If there is no forced win show us by at least addressing my variation,
instead of just simply swallowing Fritz evaluation.
(By the way, Fritz evaluation in positions like this is pathetic)
Adnan
wrote:
>Kasparov was right not to try the sacrifice, he might have lost the game.
Please!! Address my moves instead of posting your own moves for white and
proclaiming the sacrifice is wrong.
Secondly, we are not talking about whether or not Kasparov made the right
choice. We are discussing whether or not Deep Blue made a blunder by allowing
Bxh7. So far you have not shown anything wrong with my analysis.
Adnan
>How about 28 ... Kd7 29. Nxf7 Qe7 30. Re5 Nf6 31. Rxd5+ Nxd5 32. h5 Rf8
>33. Ne5+ Nxe5 34. Qxe5 Qf6 35. Qxf6 Rxf6
I have to go to work. Would be back around 9 p.m. or somewher.. A quick answer
to you is that 29. Nxe6 followed Nf4 or Nc5 looks much better instead of 29.
Nxf7
Adnan
I posted the last post before looking at this one.
>Or 28. ... Kd7 29. Nxe6 Nce7 30. Nf4 Rxc1+ 31. Bxc1 Ra5 32. h5 Rxa2
>33. h6 Qc8 34. Nd3 Qc2 35. Nxb4 Qxc1+
35. h7 instead of 35. Nxb4 wins for white in your analysis.
Like Fritz the problem is with your evaluation (I am not talking about
overlooking 35. h7 but over all position starting from 28 Kd7.
I would be back latter but what is your response to 35. h7?
White wins easily.
But at any rate, I would be back latter
Jeremiah Penery
>
> On Thu, 25 Mar 1999 17:06:11 -0500, Jeremiah Penery <fl...@lords.com> wrote:
>
> I posted the last post before looking at this one.
>
> >Or 28. ... Kd7 29. Nxe6 Nce7 30. Nf4 Rxc1+ 31. Bxc1 Ra5 32. h5 Rxa2
> >33. h6 Qc8 34. Nd3 Qc2 35. Nxb4 Qxc1+
>
> 35. h7 instead of 35. Nxb4 wins for white in your analysis.
What would you do after 31. ... Rf5?
Adnan
28... Kd7 looks like black's best defense, but I think white keeps strong
advantage. Black's has three problems. (1) His king is unsafe (2) His pieces
are badly placed, and (3) White's h pawn is very dangerous. In some variations
white can play g4 followed by h5.
One possibility is: 31. ...Rf5 32. Nd3 Bd6. 33. Ne5+ Bxe5. 34. dxe5 Nd5
(34...Qc8 35. Rc3) 35. e6+ Kc6. 36. Re4 so white is much better at least
after, 32... Bd6 followed by 33... Bxe5, black moves played by my comp.
Henri H. Arsenault
wrote:
> Secondly, we are not talking about whether or not Kasparov made the right
> choice. We are discussing whether or not Deep Blue made a blunder by allowing
> Bxh7. So far you have not shown anything wrong with my analysis.
OK, I looked at it last night and it seems that inded your move 32. d5!
does lead to a forced win for White, but the evaluation only shows it as a
win past move 37.
So assuming that the moves from move 19 are played all the way to 32. d5,
White would have won the game. Now from move 20 to 32 is 24 plies, and
from to 37 is 34 plies, and it is impossible for a machine or human to see
nearly that far.
So although I must admit that with the extensive but limited analysis that
I have done, Bxh7 DOES eventually lead to a win for White, the fact that
thre are many pitfalls for both sides along the way and that the winning
position only appears 34 plies later, it does not seem reasonable to claim
that it was a blunder by Deep Blue to allow Bxh7. Assuming that Deep Blue
could see 25 plies ahead (and no computer can see NEARLY that far), it
STILL would not have seen the move 32 d5 and its consequences.
The remaining question is whether Kasparov could have seen the move
himself; clearly not, since he did not ply the move.
So it seems irrefutable to say that to allow Bxh7 was NOT a blunder,
especially given the fact that it is not clear whether or not other lines
are better. At most, one can say that DB's move may not have been the
strongest one, but given the status of computer chess knowledge, there is
no such thing as a "best" move beyond what it is possible to analyze,
since so far evidence suggests that there are always new best moves for
every new ply.
Henri
Henri H. Arsenault
wrote:
> What ? What? What? What is your rating??? Am I just wasting my breath
> here?????
>
> Why don't you follow my analysis after Qh7 instead of just blindly trusting
> Fritz evaluation??
>
My rating is immaterial; we are discussing whether or not a computer made
a blunder, so computer analysis with the best engines available are in
order.
The response to your analysis is in anothr posting (you are right about d4).
Henri