Main
Date: 15 Oct 2008 11:20:32
From: chessplayer
Subject: In game 2 Kramnik had a slight advantage upto move 25.
Kramnik made a few slight mistakes in game 2 where he did have a
slight advantage. In move 26 he should have simply taken the pawn on
c4. Even later on move 29 he could have played Rh6 instead of Nd3.
Black could have pulled off a victory had he played more aggresively
like he started earlier on.

Also, when they decided to call it a draw it is actually a win for
white (but a very difficult one). If perfect moves are calculated
white could actually pull off a very difficult victory from the point
of calling it a draw on move 32. However, I still feel Kramnik with
his earlier aggresive moves could have pulled off a victory. But since
neither Kramnik pushed for a victory and later since neither Anand did
so I guess a draw in that sense was a fair result. At least this game
was more exciting than the boring first one.




 
Date: 17 Oct 2008 21:44:51
From: chessplayer
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
On Oct 17, 6:16=A0pm, David Richerby <[email protected] >
wrote:
> chessplayer <[email protected]> wrote:
> > David Richerby <[email protected]> wrote:
> >> chessplayer <[email protected]> wrote:
> >>> David Richerby <[email protected]> wrote:
> >>>> In the absence of 19-man tablebases, any comment about what
> >>>> `perfect moves' can achieve from that position is pure speculation.
> >>>> We suspect that perfect chess looks a lot like the chess that
> >>>> strong computers and grandmasters play but there's a lot of
> >>>> conditioning and reinforcement going on there.
>
> >>> Agreed. There may really be no such thing as perfect chess.
>
> >> There definitely is such a thing because chess is a finite two-player
> >> game of perfect information.
>
> > I was merely being open to the suggestion but I too believe that there
> > is perfect chess.
>
> Being `open to the suggestion' that there might be no such thing as
> perfect chess is like being `open to the suggestion' that eight is an
> odd number. =A0Well, except that the proof of determinacy for finite
> two-player games of perfect information is rather more complex than
> the proof that eight is even, so there's more scope for it containing
> a mistake.
>
> > As I stated in my earlier post, the minute advantages of white
> > cannot compensate for the immense factors at play in going for a
> > draw. So, if both players are playing the perfect responses the only
> > way one player can really win is if the other makes a mistake.
>
> There are no mistakes in perfect chess. =A0That's why it's called
> perfect. ;-)
>
> > Can computers calculate all the perfect moves just by knowing all
> > the moves of the chess pieces and the neccessary knowledges of the
> > rules of the game.
>
> In principle, yes. =A0All you need to do is program a computer with the
> rules of the game and have it search through the possibilities. =A0You
> don't need an evaluation function, except to guide the search into
> areas of the tree that look most promising. =A0Because, ultimately,
> there are only three possible evaluations of a position:
>
> 1. whatever your opponent does, you have a way to checkmate him;
> 2. whatever you do, your opponent has a way to checkmate you;
> 3. as long as nobody makes a mistake, it's a draw.
>
> In perfect chess, there's no such thing as a `slight advantage'.
>
> There are two problems with this approach. =A0The first is that it would
> take an unimaginably long time to come up with an answer. =A0The second
> is that, to the best of my knowledge, it's the only one we have.
> (Aside from optimizations such as tablebases.)
>
> > For instance I honestly believe that Kasparov's last championship
> > against a computer (in the game which he won) was really not that
> > great of a program. I would really like Kasparov to just play a game
> > against Fritz or Chessmaster on a simple home computer running with
> > a pentium 4 and a 3 GHz processor. If Kasparov can prove to me just
> > on something so simple how he can beat the computer I would be
> > convinced of his superiority over current programs and computers.
>
> Well, Kasparov retired three years ago, now. =A0He played a six-game
> match against Deep Junior and a four-game match against (deep?) Fritz,
> both in 2003. =A0Kramnik played an eight-game match against Deep Fritz
> in 2002. =A0All of these matches were drawn.
>

Didn't Kasparov win one of those matches.

> Dave.
>
> --
> David Richerby =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =
=A0Gigantic Lead Bulb (TM): it's likewww.chiark.greenend.org.uk/~davidr/=A0=
=A0 =A0 =A0 a light bulb that weighs a ton but
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =
=A0 =A0 =A0 =A0 it's huge!- Hide quoted text -
>
> - Show quoted text -



  
Date: 18 Oct 2008 19:32:32
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
chessplayer <[email protected] > wrote:
> David Richerby <[email protected]> wrote:
>> chessplayer <[email protected]> wrote:
>>> For instance I honestly believe that Kasparov's last championship
>>> against a computer (in the game which he won) was really not that
>>> great of a program. I would really like Kasparov to just play a
>>> game against Fritz or Chessmaster on a simple home computer
>>> running with a pentium 4 and a 3 GHz processor.
>>
>> Well, Kasparov retired three years ago, now. He played a six-game
>> match against Deep Junior and a four-game match against (deep?)
>> Fritz, both in 2003. Kramnik played an eight-game match against
>> Deep Fritz in 2002. All of these matches were drawn.
>
> Didn't Kasparov win one of those matches.

No. He won the 1996 match against IBM's Deep Blue but that ran on
custom hardware. I didn't mention that match or the 1997 match
against Deeper Blue (which Kasparov lost) as I was addressing your
comment that Kasparov should play a match against a publicly available
program running on commodity hardware.

I don't recall any later matches against computers and Wikipedia
doesn't mention any.


Dave.

--
David Richerby Swiss Apple (TM): it's like a tasty
www.chiark.greenend.org.uk/~davidr/ fruit but it's made in Switzerland!


 
Date: 16 Oct 2008 11:48:00
From: chessplayer
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
On Oct 16, 9:31=A0pm, David Richerby <[email protected] >
wrote:
> chessplayer <[email protected]> wrote:
> > David Richerby <[email protected]> wrote:
> >> chessplayer <[email protected]> wrote:
> >>> Also, when they decided to call it a draw it is actually a win for
> >>> white (but a very difficult one). If perfect moves are calculated
> >>> white could actually pull off a very difficult victory from the
> >>> point of calling it a draw on move 32.
>
> >> In the absence of 19-man tablebases, any comment about what
> >> `perfect moves' can achieve from that position is pure speculation.
> >> We suspect that perfect chess looks a lot like the chess that
> >> strong computers and grandmasters play but there's a lot of
> >> conditioning and reinforcement going on there.
>
> > Agreed. There may really be no such thing as perfect chess.
>
> There definitely is such a thing because chess is a finite two-player
> game of perfect information.

I was merely being open to the suggestion but I too believe that there
is perfect chess. Its just that if perfect chess does not exist today
even on computers its only because we haven't discovered the way to
make the perfect program which can attribute all the qualities in a
mathematical way into the program.


=A0As such, in any legal position, one or
> both of the following is true
>
> 1. white has a strategy that guarantees at least a draw, regardless
> =A0 =A0of black's responses;
> 2. black has a strategy that guarantees at least a draw, regardless
> =A0 =A0of white's responses.
>
> If only 1. holds, the position is a forced win for white (he can get
> at least a draw but black guarantee a draw, so white must be able to
> win against any defence); if only 2. holds, the position is a forced
> win for black; if both hold, the position is drawn with best play.

Well here it gets interesting. White may play for a draw and black in
trying to go for a win might make a mistake by which white may end up
winning and vice versa. Its going for a win which makes players take
the extra chances which give opportunities to the other players. If
there are moves by which a player can go for a win WITHOUT TAKING
EXTRA CHANCES I disagree with that theory. As I stated in my earlier
post, the minute advantages of white cannot compensate for the immense
factors at play in going for a draw. So, if both players are playing
the perfect responses the only way one player can really win is if the
other makes a mistake. Depending upon the quality of the player and
the mistake this can be exploited. Strategy as it is called so
eloquently is nothing but a move which is overlooked by another player
and therefore that player fails to counter that strategy and the time
he or she should. Later on that strategic move proves deadly. So, the
mistake was made by the player in not seeing where that earlier
strategic move was made which may have looked innocuous at the time.

I honestly believe that once all the parameters are put into a
sophisticated chess program and we can calculate all the mathematical
variations the game that will be played by the computers which have
that chess program will be a perfect one. The question that remains to
be asked is that do we know all the mathematical equations that go
into making the perfect chess program. Can computers calculate all the
perfect moves just by knowing all the moves of the chess pieces and
the neccessary knowledges of the rules of the game. Over here, the
pawn movements are critical. But giving each pawn at each level the
pawn moves up a significantly higher value and coordinating that with
the other pieces around is critical. If the chess program can take all
these factors into consideration and calculate all the possible
combinations to give the best possible move for every particular move
we would certainly have a draw each time.
>
> These strategies define exactly what `perfect chess' is. =A0Now, if what
> you actually mean is that we might never know what the strategies are,
> I agree with you. =A0But the mathematics guarantees that they do exist,
> even if we can never know what they are.

Yes, the ordinary mortal could certainly not calculate all the
possibilites but a sophisticated chess program on a supercomputer
certainly could. Since the mathematics guarantees that the perfect
moves exist the question becomes can we at least feed the mathematical
equations into the program even though human beings cannot calculate
all the possible combinations to find the perfect one certainly the
computer can. Over here the question becomes do we know how to program
ALL the mathematical equations into the computer program.
>
> > However, given the parameters of chess theory today I guess we could
> > extrapolate what would be the perfect moves given our combined level
> > of knowledge of chess theory. Of course that could be challenged
> > tomorrow.
>
> No. =A0You can't extrapolate from `very good' to `perfect'. =A0Current
> theory tells us that white should probably open the game with c4, d4,
> e4 or Nf3 (in alphabetical order to avoid arguments about which is
> best). =A0But, for all we know, black might have a forced win via 1.e4
> f6. =A0That's the problem with minimax -- the re-evaluation of a single
> at great depth in the tree can change the evaluation of the root and
> cause something that looks `close to perfect' today to look like
> complete bunk tomorrow.
>
> >> It's still technically possible that chess is a forced win for
> >> white but that the only winning opening move is a3.
>
> > True. But I believe that if "perfect" chess is played by both sides
> > it will always result in a draw.
>
> Many other people believe that, too. =A0Some other people (though not,
> it seems, so many) believe that chess is a forced win for white and a
> few even believe it's a forced win for black (arguing that black,
> moving second, always has more information about the position when he
> moves).

As I stated I feel that the tempo advantage of white is negated by the
rules of chess which have many factors in it favoring a draw. However,
the point of Black actually having an advantage is interesting. If not
advantage at the very least black could have a neutral position in
that the one tempo advantage of white gets negated by the fact that
black always knows what white is doing and can therefore counter it.
>
> So, I'm not trying to discredit your opinion that perfect chess is a
> draw. =A0Indeed, I think it probably is, too. =A0But this is just the
> majority opinion and that doesn't make it true.

Your points are certainly very interesting.
>
> > In fact whatever minute advantage white has by playing first is
> > completely negated by the aspect of a draw.
>
> That is not a `fact' -- it is just your unproven belief.
>
> > However, I guess if theoritical "perfect" chess is played from any
> > given point of a particular game (which in this case would be after
> > move 18 of white in game 2) one could extrapolate who would be the
> > most likely winner.
>
> Well, there's a subtlety here, about just how perfect we want our
> perfect chess to be. :-) =A0The definition of `perfect' I gave earlier
> is actually very strong -- it requires a strategy for perfect play
> from every possible position, including such things as how to
> checkmate with three knights against a lone king (there is a forced
> mate, here).
>
> But it might be that, starting from the initial position, neither side
> ever has to checkmate with three knights -- maybe that pawn always
> gets promoted to a queen. =A0So, arguably, it's `perfect enough' to be
> able to guarantee at least a draw from the initial position, and not
> worry about positions that can never come up when the strategy is
> followed. =A0This is a bit like saying that, if 1.e4 gives a forced win
> for white, you don't need to know what to do after 1.d4 is played
> because nobody would ever do that.
>
> So, perhaps the position in question is one of those that would never
> come up in `perfect enough' chess so we wouldn't know how to play on
> from there.
>
> In game theory terms, finding the way to play `perfect' chess is
> usually called `strongly solving chess' and finding the way to play
> `perfect enough' chess is called `weakly solving chess'. =A0There's
> another category called `ultra-weakly solving' -- to ultra-weakly
> solve chess, you'd just prove somehow that, for example, chess is a
> draw with perfect play, but without actually demonstrating how to
> achieve the draw.

The interesting point to note here is that what we call perfect may
just be the standard of sophistication we have achieved. Here the
point to note would be just how much has chess theory advanced after
the introduction of supercomputers and extremely powerful chess
programs. Is it continuously evolving or are we seeing more and more
draws being played out.

For instance I honestly believe that Kasparov's last championship
against a computer (in the game which he won) was really not that
great of a program. I would really like Kasparov to just play a game
against Fritz or Chessmaster on a simple home computer running with a
pentium 4 and a 3 GHz processor. If Kasparov can prove to me just on
something so simple how he can beat the computer I would be convinced
of his superiority over current programs and computers.

But I digress. The issue here is of trying to achieve perfection
(since we both agree it exists even though we may never achieve it).
>
> > But its the mistakes which allow for someone to win. =A0In supergrand-
> > masters and world champions the minutest of errors could result in a
> > victory for the opponent.
>
> Actually, I would say the opposite. =A0The drawish nature of top-level
> chess suggests that chess is not very sensitive to tiny mistakes. =A0If
> you make a tiny mistake, there's a pretty good chance that all you've
> done is make it slightly harder to get the draw. =A0And, in practical
> terms, your opponent will probably make a tiny mistake back at you
> before too long.

I guess its the quantum of the mistake. As we saw in game 2 of the
current championship. The mistakes went unnoticed and therefore were
not exploited by these two world class players. But a subtle mistake
which may go unnoticed by us might be exploited by them. Of course we
see amongst club level players that leave alone mistakes, even
blunders sometimes go unnoticed by some club level players and you
have a situation where both players commit blunders and neither
exploits them. This goes on till someone commits another blunder which
sometimes gets noticed and the opponent exploits it.

This has been a very interesting discussion.
>
> Dave.
>
> --
> David Richerby =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =
=A0 Confusing Shack (TM): it's like awww.chiark.greenend.org.uk/~davidr/=A0=
=A0 =A0 =A0 =A0house in the woods but you can't
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =
=A0 =A0 =A0 =A0 =A0understand it!



  
Date: 17 Oct 2008 14:16:05
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
chessplayer <[email protected] > wrote:
> David Richerby <[email protected]> wrote:
>> chessplayer <[email protected]> wrote:
>>> David Richerby <[email protected]> wrote:
>>>> In the absence of 19-man tablebases, any comment about what
>>>> `perfect moves' can achieve from that position is pure speculation.
>>>> We suspect that perfect chess looks a lot like the chess that
>>>> strong computers and grandmasters play but there's a lot of
>>>> conditioning and reinforcement going on there.
>>>
>>> Agreed. There may really be no such thing as perfect chess.
>>
>> There definitely is such a thing because chess is a finite two-player
>> game of perfect information.
>
> I was merely being open to the suggestion but I too believe that there
> is perfect chess.

Being `open to the suggestion' that there might be no such thing as
perfect chess is like being `open to the suggestion' that eight is an
odd number. Well, except that the proof of determinacy for finite
two-player games of perfect information is rather more complex than
the proof that eight is even, so there's more scope for it containing
a mistake.

> As I stated in my earlier post, the minute advantages of white
> cannot compensate for the immense factors at play in going for a
> draw. So, if both players are playing the perfect responses the only
> way one player can really win is if the other makes a mistake.

There are no mistakes in perfect chess. That's why it's called
perfect. ;-)

> Can computers calculate all the perfect moves just by knowing all
> the moves of the chess pieces and the neccessary knowledges of the
> rules of the game.

In principle, yes. All you need to do is program a computer with the
rules of the game and have it search through the possibilities. You
don't need an evaluation function, except to guide the search into
areas of the tree that look most promising. Because, ultimately,
there are only three possible evaluations of a position:

1. whatever your opponent does, you have a way to checkmate him;
2. whatever you do, your opponent has a way to checkmate you;
3. as long as nobody makes a mistake, it's a draw.

In perfect chess, there's no such thing as a `slight advantage'.

There are two problems with this approach. The first is that it would
take an unimaginably long time to come up with an answer. The second
is that, to the best of my knowledge, it's the only one we have.
(Aside from optimizations such as tablebases.)

> For instance I honestly believe that Kasparov's last championship
> against a computer (in the game which he won) was really not that
> great of a program. I would really like Kasparov to just play a game
> against Fritz or Chessmaster on a simple home computer running with
> a pentium 4 and a 3 GHz processor. If Kasparov can prove to me just
> on something so simple how he can beat the computer I would be
> convinced of his superiority over current programs and computers.

Well, Kasparov retired three years ago, now. He played a six-game
match against Deep Junior and a four-game match against (deep?) Fritz,
both in 2003. Kramnik played an eight-game match against Deep Fritz
in 2002. All of these matches were drawn.


Dave.



--
David Richerby Gigantic Lead Bulb (TM): it's like
www.chiark.greenend.org.uk/~davidr/ a light bulb that weighs a ton but
it's huge!


 
Date: 16 Oct 2008 06:37:05
From: chessplayer
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
On Oct 16, 4:39=A0pm, David Richerby <[email protected] >
wrote:
> [Crosspost trimmed -- this has nothing to do with computer chess.
> Anything that belongs in a more specific group is off-topic in .misc.]
>
> chessplayer <[email protected]> wrote:
> > Also, when they decided to call it a draw it is actually a win for
> > white (but a very difficult one). If perfect moves are calculated
> > white could actually pull off a very difficult victory from the
> > point of calling it a draw on move 32.
>
> In the absence of 19-man tablebases, any comment about what `perfect
> moves' can achieve from that position is pure speculation. =A0We suspect
> that perfect chess looks a lot like the chess that strong computers
> and grandmasters play but there's a lot of conditioning and
> reinforcement going on there.

Agreed. There may really be no such thing as perfect chess. However,
given the parameters of chess theory today I guess we could
extrapolate what would be the perfect moves given our combined level
of knowledge of chess theory. Of course that could be challenged
tomorrow.
>
> It's still technically possible that chess is a forced win for white
> but that the only winning opening move is a3.

True. But I believe that if "perfect" chess is played by both sides it
will always result in a draw. In fact whatever minute advantage white
has by playing first is completely negated by the aspect of a draw.
Since we have the 1) 50 moves draw rule. 2) Stalemate. 3) Cannot mate
with just one bishop (and practically impossible although
theoritically possible with 2 knights) black can always come up with a
draw if black decides to do so from the very beginning as the rules
allow for a draw due to certain conditions and even a material
advantage is meaningless.

However, I guess if theoritical "perfect" chess is played from any
given point of a particular game (which in this case would be after
move 18 of white in game 2) one could extrapolate who would be the
most likely winner.

But its the mistakes which allow for someone to win. In
supergrandmasters and world champions the minutest of errors could
result in a victory for the opponent. In lesser ranked players and
club level players the big mistakes or blunders could result in a
victory. (Provided of course the opponent can exploit those mistakes
or blunders).
>
> Dave.
>
> --
> David Richerby =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Sum=
erian Umbrella (TM): it's like anwww.chiark.greenend.org.uk/~davidr/=A0 =A0=
=A0 umbrella that's really old!



  
Date: 16 Oct 2008 17:31:04
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
chessplayer <[email protected] > wrote:
> David Richerby <[email protected]> wrote:
>> chessplayer <[email protected]> wrote:
>>> Also, when they decided to call it a draw it is actually a win for
>>> white (but a very difficult one). If perfect moves are calculated
>>> white could actually pull off a very difficult victory from the
>>> point of calling it a draw on move 32.
>>
>> In the absence of 19-man tablebases, any comment about what
>> `perfect moves' can achieve from that position is pure speculation.
>> We suspect that perfect chess looks a lot like the chess that
>> strong computers and grandmasters play but there's a lot of
>> conditioning and reinforcement going on there.
>
> Agreed. There may really be no such thing as perfect chess.

There definitely is such a thing because chess is a finite two-player
game of perfect information. As such, in any legal position, one or
both of the following is true

1. white has a strategy that guarantees at least a draw, regardless
of black's responses;
2. black has a strategy that guarantees at least a draw, regardless
of white's responses.

If only 1. holds, the position is a forced win for white (he can get
at least a draw but black guarantee a draw, so white must be able to
win against any defence); if only 2. holds, the position is a forced
win for black; if both hold, the position is drawn with best play.

These strategies define exactly what `perfect chess' is. Now, if what
you actually mean is that we might never know what the strategies are,
I agree with you. But the mathematics guarantees that they do exist,
even if we can never know what they are.

> However, given the parameters of chess theory today I guess we could
> extrapolate what would be the perfect moves given our combined level
> of knowledge of chess theory. Of course that could be challenged
> tomorrow.

No. You can't extrapolate from `very good' to `perfect'. Current
theory tells us that white should probably open the game with c4, d4,
e4 or Nf3 (in alphabetical order to avoid arguments about which is
best). But, for all we know, black might have a forced win via 1.e4
f6. That's the problem with minimax -- the re-evaluation of a single
at great depth in the tree can change the evaluation of the root and
cause something that looks `close to perfect' today to look like
complete bunk tomorrow.

>> It's still technically possible that chess is a forced win for
>> white but that the only winning opening move is a3.
>
> True. But I believe that if "perfect" chess is played by both sides
> it will always result in a draw.

Many other people believe that, too. Some other people (though not,
it seems, so many) believe that chess is a forced win for white and a
few even believe it's a forced win for black (arguing that black,
moving second, always has more information about the position when he
moves).

So, I'm not trying to discredit your opinion that perfect chess is a
draw. Indeed, I think it probably is, too. But this is just the
majority opinion and that doesn't make it true.

> In fact whatever minute advantage white has by playing first is
> completely negated by the aspect of a draw.

That is not a `fact' -- it is just your unproven belief.

> However, I guess if theoritical "perfect" chess is played from any
> given point of a particular game (which in this case would be after
> move 18 of white in game 2) one could extrapolate who would be the
> most likely winner.

Well, there's a subtlety here, about just how perfect we want our
perfect chess to be. :-) The definition of `perfect' I gave earlier
is actually very strong -- it requires a strategy for perfect play
from every possible position, including such things as how to
checkmate with three knights against a lone king (there is a forced
mate, here).

But it might be that, starting from the initial position, neither side
ever has to checkmate with three knights -- maybe that pawn always
gets promoted to a queen. So, arguably, it's `perfect enough' to be
able to guarantee at least a draw from the initial position, and not
worry about positions that can never come up when the strategy is
followed. This is a bit like saying that, if 1.e4 gives a forced win
for white, you don't need to know what to do after 1.d4 is played
because nobody would ever do that.

So, perhaps the position in question is one of those that would never
come up in `perfect enough' chess so we wouldn't know how to play on
from there.

In game theory terms, finding the way to play `perfect' chess is
usually called `strongly solving chess' and finding the way to play
`perfect enough' chess is called `weakly solving chess'. There's
another category called `ultra-weakly solving' -- to ultra-weakly
solve chess, you'd just prove somehow that, for example, chess is a
draw with perfect play, but without actually demonstrating how to
achieve the draw.


> But its the mistakes which allow for someone to win. In supergrand-
> masters and world champions the minutest of errors could result in a
> victory for the opponent.

Actually, I would say the opposite. The drawish nature of top-level
chess suggests that chess is not very sensitive to tiny mistakes. If
you make a tiny mistake, there's a pretty good chance that all you've
done is make it slightly harder to get the draw. And, in practical
terms, your opponent will probably make a tiny mistake back at you
before too long.


Dave.

--
David Richerby Confusing Shack (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ house in the woods but you can't
understand it!


   
Date: 17 Oct 2008 01:37:47
From: David Kane
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.

"David Richerby" <[email protected] > wrote in message
news:Wuu*[email protected]...
> chessplayer <[email protected]> wrote:
>> David Richerby <[email protected]> wrote:
>>> chessplayer <[email protected]> wrote:
>>>> Also, when they decided to call it a draw it is actually a win for
>>>> white (but a very difficult one). If perfect moves are calculated
>>>> white could actually pull off a very difficult victory from the
>>>> point of calling it a draw on move 32.
>>>
>>> In the absence of 19-man tablebases, any comment about what
>>> `perfect moves' can achieve from that position is pure speculation.
>>> We suspect that perfect chess looks a lot like the chess that
>>> strong computers and grandmasters play but there's a lot of
>>> conditioning and reinforcement going on there.
>>
>> Agreed. There may really be no such thing as perfect chess.
>
> There definitely is such a thing because chess is a finite two-player
> game of perfect information. As such, in any legal position, one or
> both of the following is true
>
> 1. white has a strategy that guarantees at least a draw, regardless
> of black's responses;
> 2. black has a strategy that guarantees at least a draw, regardless
> of white's responses.
>
> If only 1. holds, the position is a forced win for white (he can get
> at least a draw but black guarantee a draw, so white must be able to
> win against any defence); if only 2. holds, the position is a forced
> win for black; if both hold, the position is drawn with best play.
>
> These strategies define exactly what `perfect chess' is. Now, if what
> you actually mean is that we might never know what the strategies are,
> I agree with you. But the mathematics guarantees that they do exist,
> even if we can never know what they are.
>
>> However, given the parameters of chess theory today I guess we could
>> extrapolate what would be the perfect moves given our combined level
>> of knowledge of chess theory. Of course that could be challenged
>> tomorrow.
>
> No. You can't extrapolate from `very good' to `perfect'. Current
> theory tells us that white should probably open the game with c4, d4,
> e4 or Nf3 (in alphabetical order to avoid arguments about which is
> best). But, for all we know, black might have a forced win via 1.e4
> f6. That's the problem with minimax -- the re-evaluation of a single
> at great depth in the tree can change the evaluation of the root and
> cause something that looks `close to perfect' today to look like
> complete bunk tomorrow.
>
>>> It's still technically possible that chess is a forced win for
>>> white but that the only winning opening move is a3.
>>
>> True. But I believe that if "perfect" chess is played by both sides
>> it will always result in a draw.
>
> Many other people believe that, too. Some other people (though not,
> it seems, so many) believe that chess is a forced win for white and a
> few even believe it's a forced win for black (arguing that black,
> moving second, always has more information about the position when he
> moves).
>
> So, I'm not trying to discredit your opinion that perfect chess is a
> draw. Indeed, I think it probably is, too. But this is just the
> majority opinion and that doesn't make it true.
>
>> In fact whatever minute advantage white has by playing first is
>> completely negated by the aspect of a draw.
>
> That is not a `fact' -- it is just your unproven belief.
>
>> However, I guess if theoritical "perfect" chess is played from any
>> given point of a particular game (which in this case would be after
>> move 18 of white in game 2) one could extrapolate who would be the
>> most likely winner.
>
> Well, there's a subtlety here, about just how perfect we want our
> perfect chess to be. :-) The definition of `perfect' I gave earlier
> is actually very strong -- it requires a strategy for perfect play
> from every possible position, including such things as how to
> checkmate with three knights against a lone king (there is a forced
> mate, here).
>
> But it might be that, starting from the initial position, neither side
> ever has to checkmate with three knights -- maybe that pawn always
> gets promoted to a queen. So, arguably, it's `perfect enough' to be
> able to guarantee at least a draw from the initial position, and not
> worry about positions that can never come up when the strategy is
> followed. This is a bit like saying that, if 1.e4 gives a forced win
> for white, you don't need to know what to do after 1.d4 is played
> because nobody would ever do that.
>
> So, perhaps the position in question is one of those that would never
> come up in `perfect enough' chess so we wouldn't know how to play on
> from there.
>
> In game theory terms, finding the way to play `perfect' chess is
> usually called `strongly solving chess' and finding the way to play
> `perfect enough' chess is called `weakly solving chess'. There's
> another category called `ultra-weakly solving' -- to ultra-weakly
> solve chess, you'd just prove somehow that, for example, chess is a
> draw with perfect play, but without actually demonstrating how to
> achieve the draw.


While everything you say is true enough, it doesn't address the real
question. Namely, can you "know" the result without proving it?

People have played chess a long time, and people have reached
conclusions about the game. It is not valid to claim that an absence
of proof means that no conclusions can be reached. An absence
of proof is simply not evidence of anything. For example, it is *not*
evidence in support of the speculation that 1. e4 f6 is a forced win for
Black.

To illustrate, imagine a simpler version of chess. 4 symmetrical pawns
each on the same side of the board, a couple of minor pieces each,
no immediate tactics- the type of position that everyone will immediately
recognize as being a draw with best play. Yet even such an obvious draw could be
hard to prove as there are still 14 pieces on the board.

Do you think Queen odds is a forced win? OK, prove it.

Or take checkers. Known to be drawn for ages,
the proof has recently been completed. It was an exercize that taught
humans absolutely nothing.

These proof arguments are akin to saying that we can never know
the excited states of a hydrogen atom until we perform a physical
test on every such atom in the universe. While such an argument might
keep some philosophers busy, it comes across as just daft to
the scientist.

>> But its the mistakes which allow for someone to win. In supergrand-
>> masters and world champions the minutest of errors could result in a
>> victory for the opponent.
>
> Actually, I would say the opposite. The drawish nature of top-level
> chess suggests that chess is not very sensitive to tiny mistakes. If
> you make a tiny mistake, there's a pretty good chance that all you've
> done is make it slightly harder to get the draw. And, in practical
> terms, your opponent will probably make a tiny mistake back at you
> before too long.
>
>
> Dave.
>
> --
> David Richerby Confusing Shack (TM): it's like a
> www.chiark.greenend.org.uk/~davidr/ house in the woods but you can't
> understand it!



    
Date: 17 Oct 2008 13:38:34
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
David Kane <[email protected] > wrote:
> "David Richerby" <[email protected]> wrote:
>> [...]
>> No. You can't extrapolate from `very good' to `perfect'. Current
>> theory tells us that white should probably open the game with c4,
>> d4, e4 or Nf3 (in alphabetical order to avoid arguments about which
>> is best). But, for all we know, black might have a forced win via
>> 1.e4 f6. That's the problem with minimax -- the re-evaluation of a
>> single at great depth in the tree can change the evaluation of the
>> root and cause something that looks `close to perfect' today to
>> look like complete bunk tomorrow.
>> [...]
>
> While everything you say is true enough, it doesn't address the real
> question. Namely, can you "know" the result without proving it?
>
> People have played chess a long time, and people have reached
> conclusions about the game. It is not valid to claim that an absence
> of proof means that no conclusions can be reached.

The question of whether chess is a forced win or draw is *exactly* a
question of proof. Without proof there is no forced win or forced
draw.

I *believe* that, with best play, chess is a draw. But that's not
nearly the same thing as knowing it. I also believe that, even if
chess is proven to be a forced win for either side,

> An absence of proof is simply not evidence of anything. For example,
> it is *not* evidence in support of the speculation that 1. e4 f6 is
> a forced win for Black.

I never took the absence of a solution to chess as evidence that 1.e4
f6 is a forced win for black! The text is quoted above.

My point is that we are getting very good at a certain kind of chess.
Anyone who studies chess gets taught to play in a certain way. The
grandmasters all play in essentially the same way. The computers are
taught to play like grandmasters; the grandmasters teach themselves to
play like computers. We'd like to think that this kind of chess that
we're playing is the right kind but we don't actually know that. Back
in the 19th Century, people got very good at playing a certain kind of
attacking, sacrificial, devil-may-care chess. Maybe they thought they
were well on the way to playing perfect chess and that it was a win
for white. And then Steinitz came along and said, `This is all
rubbish! You should play like this.'

> Do you think Queen odds is a forced win?

Yes. (For the side with the queen.)

> OK, prove it.

I can't. Hence, I don't claim to *know* that queen odds is a forced
win.

> Or take checkers. Known to be drawn for ages, the proof has recently
> been completed. It was an exercize that taught humans absolutely
> nothing.

Not at all. It taught us that our intuition about the drawishness of
checkers was correct.

> These proof arguments are akin to saying that we can never know
> the excited states of a hydrogen atom until we perform a physical
> test on every such atom in the universe.

Huh?


Dave.

--
David Richerby Slimy Smokes (TM): it's like a pack
www.chiark.greenend.org.uk/~davidr/ of cigarettes but it's covered in goo!


     
Date: 17 Oct 2008 07:54:18
From: David Kane
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.

"David Richerby" <[email protected] > wrote in message
news:YJr*[email protected]...
> David Kane <[email protected]> wrote:
>> "David Richerby" <[email protected]> wrote:
>>> [...]
>>> No. You can't extrapolate from `very good' to `perfect'. Current
>>> theory tells us that white should probably open the game with c4,
>>> d4, e4 or Nf3 (in alphabetical order to avoid arguments about which
>>> is best). But, for all we know, black might have a forced win via
>>> 1.e4 f6. That's the problem with minimax -- the re-evaluation of a
>>> single at great depth in the tree can change the evaluation of the
>>> root and cause something that looks `close to perfect' today to
>>> look like complete bunk tomorrow.
>>> [...]
>>
>> While everything you say is true enough, it doesn't address the real
>> question. Namely, can you "know" the result without proving it?
>>
>> People have played chess a long time, and people have reached
>> conclusions about the game. It is not valid to claim that an absence
>> of proof means that no conclusions can be reached.
>
> The question of whether chess is a forced win or draw is *exactly* a
> question of proof. Without proof there is no forced win or forced
> draw.
>

Not so. Chinook, e.g. played perfectly long before the checkers proof was
found.

Give me Q +R +R+ B+ N odds, and I can force a win without proving it.


> I *believe* that, with best play, chess is a draw. But that's not
> nearly the same thing as knowing it. I also believe that, even if
> chess is proven to be a forced win for either side,
>
>> An absence of proof is simply not evidence of anything. For example,
>> it is *not* evidence in support of the speculation that 1. e4 f6 is
>> a forced win for Black.
>
> I never took the absence of a solution to chess as evidence that 1.e4
> f6 is a forced win for black! The text is quoted above.

But the point is that your speculation that 1.e4 f6 might be a Black win was
not supported by any evidence. On the other hand, those arguing the contrary
position can provide all sorts of evidence. (generating move trees that give
White a good position)

Is there a colony of genius monkey-like creatures living in the interior of
Mercury
who spend their existence playing a game much more complex than chess?
I can't prove (in your sense) there isn't, but I can provide lots of biological
and planetary
science evidence arguing against the possibility.

>My point is that we are getting very good at a certain kind of chess.
> Anyone who studies chess gets taught to play in a certain way. The
> grandmasters all play in essentially the same way. The computers are
> taught to play like grandmasters; the grandmasters teach themselves to
> play like computers. We'd like to think that this kind of chess that
> we're playing is the right kind but we don't actually know that. Back
> in the 19th Century, people got very good at playing a certain kind of
> attacking, sacrificial, devil-may-care chess. Maybe they thought they
> were well on the way to playing perfect chess and that it was a win
> for white. And then Steinitz came along and said, `This is all
> rubbish! You should play like this.'
>
>> Do you think Queen odds is a forced win?
>
> Yes. (For the side with the queen.)
>
>> OK, prove it.
>
> I can't. Hence, I don't claim to *know* that queen odds is a forced
> win.
>
>> Or take checkers. Known to be drawn for ages, the proof has recently
>> been completed. It was an exercize that taught humans absolutely
>> nothing.
>
> Not at all. It taught us that our intuition about the drawishness of
> checkers was correct.


The knowledge that checkers was a draw was not intuition. It was based
on centuries of empirical evidence based on careful examination of the
game. Is it rational to discount such evidence for no reason but that it is not
mathematical proof. Of course not.


>
>> These proof arguments are akin to saying that we can never know
>> the excited states of a hydrogen atom until we perform a physical
>> test on every such atom in the universe.
>
> Huh?
>
>
> Dave.
>
> --
> David Richerby Slimy Smokes (TM): it's like a pack
> www.chiark.greenend.org.uk/~davidr/ of cigarettes but it's covered in goo!



      
Date: 18 Oct 2008 19:59:45
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
David Kane <[email protected] > wrote:
> "David Richerby" <[email protected]> wrote:
>> David Kane <[email protected]> wrote:
>>> People have played chess a long time, and people have reached
>>> conclusions about the game. It is not valid to claim that an
>>> absence of proof means that no conclusions can be reached.
>>
>> The question of whether chess is a forced win or draw is *exactly*
>> a question of proof. Without proof there is no forced win or
>> forced draw.
>
> Not so. Chinook, e.g. played perfectly long before the checkers
> proof was found.

No. Chinook played extremely well long before the checkers proof was
found. Chinook did not `force' wins until it got close to its endgame
tablebases. Earlier in the game, it merely found ways to continue
that looked very much like they were winning.

> Give me Q +R +R+ B+ N odds, and I can force a win without proving
> it.

No. You can obtain a win. You haven't forced the win until you can
demonstrate a win against every possible defensive strategy.

>>> An absence of proof is simply not evidence of anything. For
>>> example, it is *not* evidence in support of the speculation that
>>> 1. e4 f6 is a forced win for Black.
>>
>> I never took the absence of a solution to chess as evidence that
>> 1.e4 f6 is a forced win for black! The text is quoted above.
>
> But the point is that your speculation that 1.e4 f6 might be a Black
> win was not supported by any evidence. On the other hand, those
> arguing the contrary position can provide all sorts of
> evidence. (generating move trees that give White a good position)

Yes, there is lots of really good evidence that 1.e4 f6 is bad for
black. The positions you that are reached look as if they give white
the advantage. After a reasonable-looking amount of analysis, nobody
can find sufficient compensation for black. It *probably* isn't there
but people don't (and can't) analyze everything out to the bitter end.
There comes a point where one's time is better spent on lines that
*look* more promising and that are much more likely to come up in
practice. (In practical terms, who cares if 1.e4 f6 is a forced win
for black if, for the first forty moves, he has to play an intricate
series of only moves and loses if he gets any of them wrong? Nobody
would manage to play that right.)

> Is there a colony of genius monkey-like creatures living in the
> interior of Mercury who spend their existence playing a game much
> more complex than chess? I can't prove (in your sense) there isn't,
> but I can provide lots of biological and planetary science evidence
> arguing against the possibility.

This is irrelevant. Nobody is talking about Mercurian genius monkeys,
the existence of which is not a mathematical statement and, therefore,
not amenable to proof. In contrast, the statement that chess is a
draw with perfect play is precisely a statement of mathematics and,
therefore, mathematical proof is the only way of *knowing* it to be
true.

We can happily go around *believing* chess to be a draw with perfect
play on the evidence that exists. We can also believe that the
current evolution of grandmaster chess is leading towards a drawn
game.

>>> Or take checkers. Known to be drawn for ages, the proof has
>>> recently been completed. It was an exercize that taught humans
>>> absolutely nothing.
>>
>> Not at all. It taught us that our intuition about the drawishness
>> of checkers was correct.
>
> The knowledge that checkers was a draw was not intuition.

OK. It taught us that our empirical belief in the drawishness of
checkers was correct. There was no elusive counter-intuitive winning
line, after all. It was unlikely that there would be but, until the
proof, we couldn't be certain.

> It was based on centuries of empirical evidence based on careful
> examination of the game.

So was the belief that the endgame of KRRN vs KRR was drawn, even
discarding the fifty-move rule. Even many winning KQP vs KQ endgames
were thought to be drawn until tablebases came along. If centuries of
empirical evidence can get a `simple' five-man endgame wrong, why do
we believe we're right about the 32-man endgame?

> Is it rational to discount such evidence for no reason but that it
> is not mathematical proof. Of course not.

I am not discounting the empirical evidence. I am merely pointing out
that no amount of empirical evidence (except for a pruned game tree)
is sufficient to justify *knowledge* that chess is drawn, rather than
just *belief*. In the case of mathematics, it is precisely proof that
separates knowledge from belief.


Dave.

--
David Richerby Crystal Addictive Cat (TM): it's like
www.chiark.greenend.org.uk/~davidr/ a cat but you can never put it down
and it's completely transparent!


       
Date: 19 Oct 2008 11:20:55
From: David Kane
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.

"David Richerby" <[email protected] > wrote in message
news:PAp*[email protected]...
> David Kane <[email protected]> wrote:
>> "David Richerby" <[email protected]> wrote:
>>> David Kane <[email protected]> wrote:
>>>> People have played chess a long time, and people have reached
>>>> conclusions about the game. It is not valid to claim that an
>>>> absence of proof means that no conclusions can be reached.
>>>
>>> The question of whether chess is a forced win or draw is *exactly*
>>> a question of proof. Without proof there is no forced win or
>>> forced draw.
>>
>> Not so. Chinook, e.g. played perfectly long before the checkers
>> proof was found.
>
> No. Chinook played extremely well long before the checkers proof was
> found. Chinook did not `force' wins until it got close to its endgame
> tablebases. Earlier in the game, it merely found ways to continue
> that looked very much like they were winning.
>
>> Give me Q +R +R+ B+ N odds, and I can force a win without proving
>> it.
>
> No. You can obtain a win. You haven't forced the win until you can
> demonstrate a win against every possible defensive strategy.
>
>>>> An absence of proof is simply not evidence of anything. For
>>>> example, it is *not* evidence in support of the speculation that
>>>> 1. e4 f6 is a forced win for Black.
>>>
>>> I never took the absence of a solution to chess as evidence that
>>> 1.e4 f6 is a forced win for black! The text is quoted above.
>>
>> But the point is that your speculation that 1.e4 f6 might be a Black
>> win was not supported by any evidence. On the other hand, those
>> arguing the contrary position can provide all sorts of
>> evidence. (generating move trees that give White a good position)
>
> Yes, there is lots of really good evidence that 1.e4 f6 is bad for
> black. The positions you that are reached look as if they give white
> the advantage. After a reasonable-looking amount of analysis, nobody
> can find sufficient compensation for black. It *probably* isn't there
> but people don't (and can't) analyze everything out to the bitter end.
> There comes a point where one's time is better spent on lines that
> *look* more promising and that are much more likely to come up in
> practice. (In practical terms, who cares if 1.e4 f6 is a forced win
> for black if, for the first forty moves, he has to play an intricate
> series of only moves and loses if he gets any of them wrong? Nobody
> would manage to play that right.)
>
>> Is there a colony of genius monkey-like creatures living in the
>> interior of Mercury who spend their existence playing a game much
>> more complex than chess? I can't prove (in your sense) there isn't,
>> but I can provide lots of biological and planetary science evidence
>> arguing against the possibility.
>
> This is irrelevant. Nobody is talking about Mercurian genius monkeys,
> the existence of which is not a mathematical statement and, therefore,
> not amenable to proof. In contrast, the statement that chess is a
> draw with perfect play is precisely a statement of mathematics and,
> therefore, mathematical proof is the only way of *knowing* it to be
> true.
>
> We can happily go around *believing* chess to be a draw with perfect
> play on the evidence that exists. We can also believe that the
> current evolution of grandmaster chess is leading towards a drawn
> game.



>
>>>> Or take checkers. Known to be drawn for ages, the proof has
>>>> recently been completed. It was an exercize that taught humans
>>>> absolutely nothing.
>>>
>>> Not at all. It taught us that our intuition about the drawishness
>>> of checkers was correct.
>>
>> The knowledge that checkers was a draw was not intuition.
>
> OK. It taught us that our empirical belief in the drawishness of
> checkers was correct. There was no elusive counter-intuitive winning
> line, after all. It was unlikely that there would be but, until the
> proof, we couldn't be certain.
>
>> It was based on centuries of empirical evidence based on careful
>> examination of the game.
>
> So was the belief that the endgame of KRRN vs KRR was drawn, even
> discarding the fifty-move rule. Even many winning KQP vs KQ endgames
> were thought to be drawn until tablebases came along. If centuries of
> empirical evidence can get a `simple' five-man endgame wrong, why do
> we believe we're right about the 32-man endgame?

On the whole, computer examination has confirmed human
understanding of the game. Sure there might be a handful of
interesting findings in contrived endgames
that never occur in reality and which humans had no need to
analyze, but chess understanding has not been transformed.

>
>> Is it rational to discount such evidence for no reason but that it
>> is not mathematical proof. Of course not.
>
> I am not discounting the empirical evidence. I am merely pointing out
> that no amount of empirical evidence (except for a pruned game tree)
> is sufficient to justify *knowledge* that chess is drawn, rather than
> just *belief*. In the case of mathematics, it is precisely proof that
> separates knowledge from belief.
>

Do people "believe" that the earth revolves around the
sun, or do they "know" it?

If you want to get literal, then you should realize that
it is impossible in principle to "prove" that chess is a draw because
the game of chess (played by people moving plastic pieces etc.)
is not the same as its mathematical approximation to which you
refer. The strongly solved 32-man tablebase won't help you obtain
the theoretical result if you die in the middle of the game, e.g..

I don't see anything in your argument that justifies what seems to be
your position that mathematical "proof" is the only driver of
knowledge. My point was that in the current discussion, mathematical
proof is little more than a red herring. Issues should be resolved
by examination of the evidence. Period.







        
Date: 20 Oct 2008 16:26:26
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
David Kane <[email protected] > wrote:
> David Richerby <[email protected]> wrote:
>> So was the belief that the endgame of KRRN vs KRR was drawn, even
>> discarding the fifty-move rule. Even many winning KQP vs KQ
>> endgames were thought to be drawn until tablebases came along.
>
> On the whole, computer examination has confirmed human understanding
> of the game.

On the whole, yes. But I've already made the point that computers and
grandmasters are a mutually reinforcing system that could, in theory,
be barking up the wrong tree. The fact that a computer that's been
programmed to play like a GM and a GM who's trained by using computers
agree on how chess should be played is hardly surprising.

> Sure there might be a handful of interesting findings in contrived
> endgames that never occur in reality and which humans had no need to
> analyze, but chess understanding has not been transformed.

KQP vs KQ is not a `contrived endgame that never occurs in reality.'

>> I am not discounting the empirical evidence. I am merely pointing
>> out that no amount of empirical evidence (except for a pruned game
>> tree) is sufficient to justify *knowledge* that chess is drawn,
>> rather than just *belief*. In the case of mathematics, it is
>> precisely proof that separates knowledge from belief.
>
> Do people "believe" that the earth revolves around the sun, or do
> they "know" it?

This is as irrelevant as all your other examples involving physical
systems.

> If you want to get literal, then you should realize that it is
> impossible in principle to "prove" that chess is a draw because the
> game of chess (played by people moving plastic pieces etc.) is not
> the same as its mathematical approximation to which you refer. The
> strongly solved 32-man tablebase won't help you obtain the
> theoretical result if you die in the middle of the game, e.g..

Dying in the middle of the game is not `perfect play'. (Except in a
losing position, where any move, including losing on time, is as good
as any other. ;-) )

Indeed, by raising this issue you're shooting yourself in the foot.
Since any player of chess could die at any time (and, more likely,
receive a call on their mobile phone), it's impossible to `know' the
outcome of any real chess game, regardless of the skill of the
players.

> I don't see anything in your argument that justifies what seems to
> be your position that mathematical "proof" is the only driver of
> knowledge.

I have never asserted any such thing. I have merely asserted that a
statement of mathematics cannot be described as `knowledge' unless it
has been proven. I have also asserted that the statement `chess is a
draw with perfect play' is a statement of mathematics.

> My point was that in the current discussion, mathematical proof is
> little more than a red herring.

No. It is the very essence of the issue.

> Issues should be resolved by examination of the evidence. Period.

I have already explained why examination of the evidence is
insufficient to come to anything more than an educated belief about
the game-theoretic outcome of chess.


Dave.

--
David Richerby Incredible Sushi (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ raw fish but it'll blow your mind!


         
Date: 21 Oct 2008 06:05:13
From: David Kane
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.

"David Richerby" <[email protected] > wrote in message
news:EmF*[email protected]...
> David Kane <[email protected]> wrote:

>>
>> Do people "believe" that the earth revolves around the sun, or do
>> they "know" it?
>
> This is as irrelevant as all your other examples involving physical
> systems.

Not at all. It demonstrates the absurdity of your argument - hinging
as it does on invented nuance in language.



>
>> I don't see anything in your argument that justifies what seems to
>> be your position that mathematical "proof" is the only driver of
>> knowledge.
>
> I have never asserted any such thing. I have merely asserted that a
> statement of mathematics cannot be described as `knowledge' unless it
> has been proven. I have also asserted that the statement `chess is a
> draw with perfect play' is a statement of mathematics.
>
>> My point was that in the current discussion, mathematical proof is
>> little more than a red herring.
>
> No. It is the very essence of the issue.
>
>> Issues should be resolved by examination of the evidence. Period.
>
> I have already explained why examination of the evidence is
> insufficient to come to anything more than an educated belief about
> the game-theoretic outcome of chess.
>

I suppose that it is harmless to allow you your "educated belief" language
in this case. But the danger in doing so is that the essence of your
irrationality
is identical to that which appears in other more weighty debates. For
example, the creationist's primary argument (in one form or another) is that
evidence in favor of theory of evolution can be discarded because it does
not definitively account for the whereabouts of each and every biological
cell in the universe. Hence, it's just a "theory". This mirrors almost exactly
your argument that we can't know whether Queen odds is a win until we've
catalogued the win against each and every possible defense.

Essentially you discredit the scientific process by raising the bar for
evidence
to an unreachable level. An assertion like "maybe the Queen-down side has
a drawing resource, but we (and computers that we've programmed) just
haven't figured it out yet, is *not* evidence. It's just silliness.






 
Date: 16 Oct 2008 12:39:34
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
[Crosspost trimmed -- this has nothing to do with computer chess.
Anything that belongs in a more specific group is off-topic in .misc.]

chessplayer <[email protected] > wrote:
> Also, when they decided to call it a draw it is actually a win for
> white (but a very difficult one). If perfect moves are calculated
> white could actually pull off a very difficult victory from the
> point of calling it a draw on move 32.

In the absence of 19-man tablebases, any comment about what `perfect
moves' can achieve from that position is pure speculation. We suspect
that perfect chess looks a lot like the chess that strong computers
and grandmasters play but there's a lot of conditioning and
reinforcement going on there.

It's still technically possible that chess is a forced win for white
but that the only winning opening move is a3.


Dave.

--
David Richerby Sumerian Umbrella (TM): it's like an
www.chiark.greenend.org.uk/~davidr/ umbrella that's really old!


 
Date: 16 Oct 2008 00:19:34
From: chessplayer
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
On Oct 16, 3:41=A0am, Offramp <[email protected] > wrote:
> On Oct 15, 7:20 pm, chessplayer <[email protected]> wrote:
>
> > Kramnik made a few slight mistakes in game 2 where he did have a
> > slight advantage. In move 26 he should have simply taken the pawn on
> > c4. Even later on move 29 he could have played Rh6 instead of Nd3.
> > Black could have pulled off a victory had he played more aggresively
> > like he started earlier on.
>
> After 26...Qxc4 then 27.Bxh6 would have resulted in a standard Stoke-
> Adams Attack and white would have won as quick as a Rumanian fart.

He should have taken c4 with the knight and not the queen, provided he
had played properly on move 25. Don't criticize unless you understand
what is written. If you don't understand at least ask with what he
should have taken c4.


 
Date: 16 Oct 2008 00:10:38
From: chessplayer
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
On Oct 16, 5:17=A0am, Simon Krahnke <[email protected] > wrote:
> * Offramp <[email protected]> (00:41) schrieb:
>
> > On Oct 15, 7:20 pm, chessplayer <[email protected]> wrote:
>
> >> Kramnik made a few slight mistakes in game 2 where he did have a
> >> slight advantage. In move 26 he should have simply taken the pawn on
> >> c4. Even later on move 29 he could have played Rh6 instead of Nd3.
> >> Black could have pulled off a victory had he played more aggresively
> >> like he started earlier on.
>
> > After 26...Qxc4 then 27.Bxh6 would have resulted in a standard Stoke-
> > Adams Attack and white would have won as quick as a Rumanian fart.
>
> I wonder in which groups these threads are on-topic. Certainly not
> rec.games.chess.computer.
>
> mfg, =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 simon .... l

I did post in on rec.games.chess.computer but it may not have come on
it. I honestly feel (after analyzing the game after move 18 when white
plays Qe2), that it was clearly black's game from there onwards.
However, Kramnik let off the pressure and played it safe. Ironically
when they both agreed to a draw on move 32 it was just about a very
slight edge to Anand. But I guess the mental drain that occurs after
such a game you cannot blame Anand for not trying to push for a win.


 
Date: 16 Oct 2008 00:07:44
From: chessplayer
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
On Oct 16, 3:41=A0am, Offramp <[email protected] > wrote:
> On Oct 15, 7:20 pm, chessplayer <[email protected]> wrote:
>
> > Kramnik made a few slight mistakes in game 2 where he did have a
> > slight advantage. In move 26 he should have simply taken the pawn on
> > c4. Even later on move 29 he could have played Rh6 instead of Nd3.
> > Black could have pulled off a victory had he played more aggresively
> > like he started earlier on.
>
> After 26...Qxc4 then 27.Bxh6 would have resulted in a standard Stoke-
> Adams Attack and white would have won as quick as a Rumanian fart.

After 18 when white played qe2 it was black's game but Kramnik did not
play as he should have.
Even then he could have won it upto move 26 but he made a few minor
mistakes as to what he should have played. Later on when he was a pawn
down he could still have played c5 before white did so. Anyway, its my
opinion that after move 17 Qe3+ it was clearly black's game. If you
want we can go to a chess website and play from thereonwards using
whatever software to aid you.


 
Date: 16 Oct 2008 02:17:42
From: Simon Krahnke
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
* Offramp <[email protected] > (00:41) schrieb:

> On Oct 15, 7:20 pm, chessplayer <[email protected]> wrote:
>
>> Kramnik made a few slight mistakes in game 2 where he did have a
>> slight advantage. In move 26 he should have simply taken the pawn on
>> c4. Even later on move 29 he could have played Rh6 instead of Nd3.
>> Black could have pulled off a victory had he played more aggresively
>> like he started earlier on.
>
> After 26...Qxc4 then 27.Bxh6 would have resulted in a standard Stoke-
> Adams Attack and white would have won as quick as a Rumanian fart.

I wonder in which groups these threads are on-topic. Certainly not
rec.games.chess.computer.

mfg, simon .... l


  
Date: 16 Oct 2008 08:20:50
From: Chess One
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.

"Simon Krahnke" <[email protected] > wrote in message
news:[email protected]...
>* Offramp <[email protected]> (00:41) schrieb:
>
>> On Oct 15, 7:20 pm, chessplayer <[email protected]> wrote:
>>
>>> Kramnik made a few slight mistakes in game 2 where he did have a
>>> slight advantage. In move 26 he should have simply taken the pawn on
>>> c4. Even later on move 29 he could have played Rh6 instead of Nd3.
>>> Black could have pulled off a victory had he played more aggresively
>>> like he started earlier on.
>>
>> After 26...Qxc4 then 27.Bxh6 would have resulted in a standard Stoke-
>> Adams Attack and white would have won as quick as a Rumanian fart.
>
> I wonder in which groups these threads are on-topic. Certainly not
> rec.games.chess.computer.

I beg our pardons, but it is difficult always to assess what /is/ on topic
at chess.computer, since it seems to prefer to discuss abstractions.

That may seem to be a trite comment on the whole realm, but divorcing the
chess from the computer has often sems to me to be a pointless exercise -
for example, if what alan obrien writes about is true - can the computer
evaluate it?

Interestingly the most advanced chess conversations seek to reintegrate
man/machine - such as in Albert Albert's MAMS studies, and I obtained for Dr
A an introduction to Rybka's inventor, who certainly thought so!

From the chessic point of view - if the chess engine cannot see what we
people see, then its evaluation function is insufficient, and MAMS suggests
that this is because the engine still has great difficulty perceiving
exchangesw of material for other advantages. Without solving this aspect of
chess computing, then the subject is as stalled as it was 10 years ago, and
only processor speed and bigger databases and end-tables cause its advance -
in other words, things largely independent of the program intself.

Phil Innes

> mfg, simon .... l




   
Date: 16 Oct 2008 14:52:41
From: David Richerby
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
Chess One <[email protected] > wrote:
> "Simon Krahnke" <[email protected]> wrote:
>> I wonder in which groups these threads are on-topic. Certainly not
>> rec.games.chess.computer.
>
> I beg our pardons, but it is difficult always to assess what /is/ on
> topic at chess.computer, since it seems to prefer to discuss
> abstractions.

It's very simple, Phil. If it's an article *about* computer chess,
it's on-topic in rgc.computer. If it isn't about computer chess, it's
off-topic, even if it uses analysis from a chess computer.

Analysis done by computer is off topic, unless the angle is why the
computer `thought' what it did. So, discussion about why all the
computers think that such and such a move was really bad is on-topic
-- the goal is to better understand computer chess. Discussion about
why Anand or Kramnik did or did not play such and such a move is
off-topic -- the goal is to better understand a game played by humans.

> That may seem to be a trite comment on the whole realm, but
> divorcing the chess from the computer has often sems to me to be a
> pointless exercise - for example, if what alan obrien writes about
> is true - can the computer evaluate it?

There is no attempt to divorce chess from computers. However, if one
is commenting on an Anand-Kramnik game, even using computer analysis
as a tool, this is no more on-topic in rgc.computer (I used Fritz!)
than it is on comp.os.windows (it was running under XP!).

> From the chessic point of view - if the chess engine cannot see what
> we people see, then its evaluation function is insufficient

A discussion along those lines would be entirely on-topic and welcome
in rgc.computer.


Dave.

--
David Richerby Swiss Erotic Chicken (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ farm animal but it's genuinely erotic
and made in Switzerland!


 
Date: 15 Oct 2008 15:41:51
From: Offramp
Subject: Re: In game 2 Kramnik had a slight advantage upto move 25.
On Oct 15, 7:20 pm, chessplayer <[email protected] > wrote:

> Kramnik made a few slight mistakes in game 2 where he did have a
> slight advantage. In move 26 he should have simply taken the pawn on
> c4. Even later on move 29 he could have played Rh6 instead of Nd3.
> Black could have pulled off a victory had he played more aggresively
> like he started earlier on.

After 26...Qxc4 then 27.Bxh6 would have resulted in a standard Stoke-
Adams Attack and white would have won as quick as a Rumanian fart.